Dynamic Feed Control

DYNAMIC FEED CONTROL: A NEW METHOD FOR
INJECTION MOLDING OF HIGH QUALITY PLASTIC PARTS
A DISSERTATION
SUBMITTED TO THE DESIGN DIVISION
OF MECHANICAL ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
IN
MECHANICAL ENGINEERING
David Owen Kazmer

June 1995
© Copyright by David Kazmer 1995
All Rights Reserved
ii
I certify that I have read this dissertation and that in my opinion it is fully adequate,
in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
_______________________________
Philip Barkan (Principal Adviser)
_______________________________
Kosuke Ishii
_______________________________
Friedrich Prinz
Approved for the University Committee on Graduate Studies:
_______________________________
iii
Abstract
The thrust of thermoplastic materials into advanced technical applications

has resulted in industry requirements which exceed the capabilities of available
product development and manufacturing processes. The lack of robustness in
these processes is evidenced by the long product development cycles, excessive
tooling costs, low process yields, and inferior product quality. Dynamic Feed
Control is a major revolutionary concept — an innovative industrial process that
opens up new potential opportunities in the manufacture of plastic parts. With the
cooperation of industry, government, and academia, this concept has been
implemented and validated for the design and production of high quality, molded
plastic parts.
The invention utilizes multiple valves in the feed system of a mold to

selectively regulate the flow to each area of the cavity in response to real-time
feedback from nearby cavity pressure transducers. A minimum complexity
process model was identified for development of a closed-loop control strategy.
Using an adaptive gain scheduling approach, the cavity pressures at each gate
were controlled throughout the filling and packing stages. This enables the
process dynamics to be modified so as to produce parts with the desired part
properties without retooling mold steel.
The flexibility of the process was demonstrated by its ability to control flow
rates and knit-line location in the filling stage as well as pack pressure and part
dimensions in the packing stage. Moreover, the consistency of the proposed
process was compared to the conventional process. This comparison was
achieved by intentionally adding process noise in an experiment designed to
simulate natural material, machine, and operator variation. Analysis of the
iv
experimental results showed an increase in the process capability, Cp, from 0.56
for the conventional molding process to 1.67 for Dynamic Feed Control.
Finally, a methodology for the design of molded parts was introduced

which leverages the degrees of freedom provided by Dynamic Feed Control.
Product robustness was demonstrated by a test series based on the stochastic
(probabilistic) distribution of material properties during the molding process. With
this assumption, the production yield of various design and processing strategies
were evaluated. The methods presented here are directly extensible to arbitrarily
complex applications with a broad range of properties, requirements, and
specifications.
v
Acknowledgments
It is such a fortunate experience that I’ve been able to meet and befriend
Phil Barkan, an extraordinary man. Without his direct involvement, I would not
have returned for this degree, much less achieved the contributions described in
this work. It is not possible to thank him adequately. It was also a pleasure
working with Fritz Prinz and Kos Ishii who provided penetrating examination of
non-trivial issues; Kos also provided support and insight which led to my faculty
appointment at the University of Massachusetts. Many graduate students
provided a constant stream of encouragement and ideas – special thanks to Ron
Worth and Russell Ford.
The work was performed with the cooperation of several industry

sponsors. Gene Yazbak at Dynisco Instruments was the first champion,
providing me with process instrumentation while validating of the importance of
this research. With Dynisco Instrument’s support, Kona Corporation (Paul
Swenson, Chris Lee, and Gloria Hurley) provided the detailed design and
manufacture of the custom feed system which never failed, even with months of
experimental abuse. GE Plastics (Chip Hills, Dick Lassor, Jim Gray, and Phil
McCarthy) donated the material and servo-valve amplifiers used in this
research...not to mention the office, lab, and processing facilities! The United
States Department of Energy also supported this research through the Innovative
Concepts Program.
Finally, I’d like to acknowledge the support of the Stanford Integrated

Manufacturing Association, especially Chuck Holloway and Bill Reynolds for the
creation of the Future Professors of Manufacturing Program as well as Rick Reis
and Susan Hansen for their friendship over the last three years. I’ve enjoyed the
time spent with the other FPMs and look forward to future cooperation.
vi
vii
Dedication
To my wife, Nancy.
I’m so glad you happened to me.
viii
Table of Contents
ABSTRACT ..................................................................................................................................................iv
ACKNOWLEDGEMENTS.........................................................................................................................vi
DEDICATION............................................................................................................................................viii
TABLE OF CONTENTS.............................................................................................................................ix
LIST OF TABLES .....................................................................................................................................xiv
LIST OF FIGURES ....................................................................................................................................xv
NOMENCLATURE .................................................................................................................................xviii
CHAPTER 1: INTRODUCTION ...............................................................................................................2

THE PLASTICS INDUSTRY ........................................................................................................................2
Historical Perspective ...............................................................................................................................2
Current Status ...........................................................................................................................................4
RECENT DEVELOPMENTS ........................................................................................................................5
Advances in the Design Process................................................................................................................6
Example................................................................................................................................................................. 6
Computer Technologies......................................................................................................................................... 7
Advances in the Injection Molding Process ..............................................................................................9
Control Systems .................................................................................................................................................. 10
Alternative Technologies..................................................................................................................................... 13
PROCESS LIMITATIONS ..........................................................................................................................16
Complex Material Behavior ....................................................................................................................16
Sensitivity of Product Quality to Input Variation....................................................................................18
Inability to Adapt Process to Changing Production Requirements ........................................................21
SUMMARY .................................................................................................................................................22
ix
CHAPTER 2: PROPOSED APPROACH.................................................................................................24
A CASE STUDY ..........................................................................................................................................24
DEGREES OF FREEDOM ..........................................................................................................................27
Degrees of freedom in tool design...........................................................................................................28
Degrees of freedom in product design ....................................................................................................29
DYNAMIC FEED CONTROL.....................................................................................................................30
Concept Overview ...................................................................................................................................32
System Overview......................................................................................................................................34
Potential Benefits ....................................................................................................................................35
Compensation for Complex Material Behavior ................................................................................................... 36
Rejection of Input Variation ................................................................................................................................ 36
Ability to Adapt to Changing Requirements ....................................................................................................... 37
SUMMARY .................................................................................................................................................38
CHAPTER 3: PROCESS DYNAMICS.....................................................................................................39

PROCESS DESCRIPTION ..........................................................................................................................39
Plastication Stage ................................................................................................................................................ 40
Filling Stage ........................................................................................................................................................ 40
Packing Stage ...................................................................................................................................................... 41
Ejection Stage...................................................................................................................................................... 42
Assumptions.............................................................................................................................................42
Filling Stage ........................................................................................................................................................ 43
Packing Stage ...................................................................................................................................................... 43
Further Assumptions............................................................................................................................................ 43
Equations for Laminar Flow in a Rectangular Channel.........................................................................45
Machine Dynamics ..................................................................................................................................47
CONTROL MODEL ....................................................................................................................................49
Derivation................................................................................................................................................49
Parameter Estimation..............................................................................................................................52
Rheological Behavior .......................................................................................................................................... 53
Compressibility ................................................................................................................................................... 54
Open Loop Simulation.............................................................................................................................55
Constant Valve Position ...................................................................................................................................... 56
Dynamic Valve Positioning................................................................................................................................. 60
x
CONTROL SYSTEM DEVELOPMENT .....................................................................................................62
Overview of Control Strategies ...............................................................................................................62
Challenges ........................................................................................................................................................... 62
Potential Strategies for Cavity Pressure Control ................................................................................................. 63
Adaptive Gain Scheduling .......................................................................................................................65
Filling Dynamics ................................................................................................................................................. 66
Packing Dynamics ............................................................................................................................................... 71
Adaptive Scheduling ........................................................................................................................................... 76
SUMMARY .................................................................................................................................................83
CHAPTER 4: EXPERIMENTAL VALIDATION OF CONTROL STRATEGY ................................84

IMPLEMENTATION ..................................................................................................................................84
Control System ........................................................................................................................................85
Molding Machine ....................................................................................................................................86
Feed System.............................................................................................................................................88
Mold Cavities ..........................................................................................................................................90
SIMULATION PERFORMANCE ...............................................................................................................91
Actuation Performance............................................................................................................................91
Open Loop Comparison ..........................................................................................................................94
CAVITY PRESSURE CONTROL...............................................................................................................96
Demonstration.........................................................................................................................................97
Controller Tuning....................................................................................................................................98
Filling Stage ........................................................................................................................................................ 99
Packing Stage .................................................................................................................................................... 100
Adaptation ......................................................................................................................................................... 103
SUMMARY ...............................................................................................................................................106
CHAPTER 5: EXPERIMENTAL VALIDATION OF PROCESS CAPABILITY.............................107

PROCESS FLEXIBILITY..........................................................................................................................107
Filling Stage Flexibility.........................................................................................................................108
Packing Stage Flexibility.......................................................................................................................113
PROCESS CONSISTENCY ......................................................................................................................118
Cavity Pressure Response .....................................................................................................................121
Filling Stage Consistency ......................................................................................................................122
Packing Stage Consistency....................................................................................................................125
xi
PROCESS LIMITATIONS ........................................................................................................................128
Limitations from Process Physics .........................................................................................................128
Filling Stage ...................................................................................................................................................... 128
Packing Stage .................................................................................................................................................... 129
Temperature Effects .......................................................................................................................................... 130
Limitations from System Design............................................................................................................131
Response Time .................................................................................................................................................. 131
Hydraulic Pilot .................................................................................................................................................. 133
Valve Interaction ............................................................................................................................................... 133
SUMMARY ...............................................................................................................................................135
CHAPTER 6: DESIGN METHODOLOGY...........................................................................................137

PRODUCT DESIGN PROCESSES ...........................................................................................................138
Iterative Design .....................................................................................................................................138
Optimization ..........................................................................................................................................140
PROPOSED METHODOLOGY................................................................................................................143
Specifications ........................................................................................................................................145
Part Design ...........................................................................................................................................146
Mold Design ..........................................................................................................................................146
Evaluation .............................................................................................................................................148
Process Relations ..................................................................................................................................150
Design Optimization..............................................................................................................................155
Robustness.............................................................................................................................................156
A CASE STUDY ........................................................................................................................................157
Design....................................................................................................................................................158
Objectives and Constraints ...................................................................................................................161
Design Evaluation .................................................................................................................................162
One Center Gate, Constant Pressure.................................................................................................................. 162
One Center Gate, Optimal Pressure................................................................................................................... 164
Three Gates, Optimal Pressure .......................................................................................................................... 165
Two Gates, Optimal Pressure ............................................................................................................................ 166
PROCESS CONTROL FOR MULTIOBJECTIVE OPTIMIZATION.......................................................167
SUMMARY ...............................................................................................................................................170
xii
CHAPTER 7: CONCLUDING REMARKS...........................................................................................172
CONCLUSIONS ........................................................................................................................................172
CONTRIBUTIONS....................................................................................................................................175
Implementation and Validation .............................................................................................................175
Design and Manufacturing Robustness.................................................................................................176
FUTURE WORK .......................................................................................................................................176
Design Methodology .............................................................................................................................176
Control System ......................................................................................................................................177
System Design .......................................................................................................................................177
Commercial Validation .........................................................................................................................178
APPENDICES ...........................................................................................................................................179
A-1 MATLAB ADAPTIVE CODE.............................................................................................................179
A-2 MATLAB CLOSED LOOP SIMULATION........................................................................................180
A-3 MATLAB CODE FOR OPEN LOOP COMPARISON .......................................................................181
B-1 DATA FOR REGRESSION OF CONVENTIONAL PROCESS.........................................................182
B-2 DATA FOR REGRESSION OF DYNAMIC FEED.............................................................................183
C-1 REVERSE TAPER...............................................................................................................................184
C-2 DIGITAL MODULATION ..................................................................................................................185
C-3 ROTARY ACTUATION......................................................................................................................186
D-1 MONTE CARLO SIMULATION........................................................................................................187
D-2 RUNROBUST.XLS .............................................................................................................................188
D-3 ROBUST.XLS .....................................................................................................................................189
REFERENCES ..........................................................................................................................................190
xiii
List of Tables
TABLE 1-1: PROCESS TO MODIFY MOLD............................................................................................ 21

TABLE 2-1: DIMENSIONS FOR MOLDED PART ..................................................................................... 26
TABLE 2-2: TYPICAL CHANGING PRODUCTION REQUIREMENTS ........................................................... 38
TABLE 3-1: LIST OF MODEL PARAMETERS ......................................................................................... 52
TABLE 3-2: PROCEDURE FOR SIMULATION OF NON-LINEAR PROCESS MODEL .................................... 56
TABLE 3-3: GAINS FOR FILLING STAGE CONTROL .............................................................................. 70
TABLE 3-4: GAINS FOR PACKING STAGE CONTROL ............................................................................ 75
TABLE 4-1: INCOMING CONTROL SIGNALS .......................................................................................... 85
TABLE 4-2: PERFORMANCE PARAMETERS OF MITSUBISHI 390 MJ MOLDING MACHINE ........................ 86
TABLE 4-3: MODIFICATIONS PERFORMED TO OBTAIN DYNAMIC FEED SYSTEM .................................... 88
TABLE 4-4: DESIGN PARAMETERS OF FEED SYSTEM ......................................................................... 89
TABLE 5-1: PART DEFECTS CAUSED BY FLOW RATES IN THE FILLING STAGE.................................... 109
TABLE 5-2: PART DEFECTS DETERMINED BY PRESSURES IN THE PACKING STAGE ............................ 114
TABLE 5-3: SOURCES OF VARIATION IN INJECTION MOLDING ............................................................ 118
TABLE 5-4: EXPERIMENTAL DESIGN FOR PROCESS CONSISTENCY.................................................... 120
TABLE 5-5: CONSISTENCY RESULTS FOR CONVENTIONAL PROCESS ................................................. 127
TABLE 5-6: RANGE OF CAVITY PRESSURE SLOPES FOR 2 MM AND 3 MM WALL THICKNESS .............. 128
TABLE 5-7: ACHIEVABLE PACK PRESSURE DIFFERENTIAL ................................................................. 129
TABLE 5-8: RESPONSE TIMES OF SEVERAL CONTROL ELEMENTS ..................................................... 132
TABLE 6-1: TYPICAL PRODUCT SPECIFICATIONS FOR A MOLDED PART ............................................. 146
TABLE 6-2: TYPES OF SPECIFICATIONS ........................................................................................... 147
TABLE 6-3: SPECIFIED DIMENSIONS FOR MOLDED PART .................................................................. 158
TABLE 6-4: PERFORMANCE OF ONE CENTER GATE AT CONSTANT PRESSURE .................................. 163
TABLE 6-5: PERFORMANCE OF ONE CENTER GATE AT OPTIMAL PRESSURE ..................................... 164
TABLE 6-6: PERFORMANCE OF THREE GATES AT OPTIMAL PRESSURES ........................................... 166
TABLE 6-7: PERFORMANCE OF TWO GATES AT OPTIMAL PRESSURES .............................................. 167
TABLE 6-8: PERFORMANCE OF THREE GATES AT NON-OPTIMAL PRESSURES ................................... 170
TABLE 7-1: ALTERNATE SOLUTION APPROACHES ............................................................................. 178
xiv
List of Figures
FIGURE 1-1: CLASSIC DESIGN TO PRODUCTION PROCESS ................................................................... 6

FIGURE 1-2: MA’S DESCRIPTION OF THE INJECTION MOLDING PROCESS ............................................ 11
FIGURE 1-3: MODEL-BASED PROCESS OPTIMIZATION ........................................................................ 13
FIGURE 1-4: PUSH-PULL DEVICE....................................................................................................... 14
FIGURE 1-5: VISCOSITY OF POLYCARBONATE .................................................................................... 17
FIGURE 1-6: PARETO CHART OF COMMON DEFECTS ......................................................................... 19
FIGURE 2-1: TYPICAL MOLDED PART ................................................................................................. 25
FIGURE 2-2: REQUIRED TOLERANCES ............................................................................................... 25
FIGURE 2-3: DEGREES OF FREEDOM IN TOOL DESIGN ....................................................................... 29
FIGURE 2-4: DEGREES OF FREEDOM IN PRODUCT DESIGN ................................................................ 30
FIGURE 2-5: PRODUCT DEVELOPMENT CYCLE ................................................................................... 31
FIGURE 2-6: DYNAMIC FEED CONCEPT .............................................................................................. 32
FIGURE 2-7: SYSTEM OVERVIEW OF DYNAMIC FEED CONTROL .......................................................... 34
FIGURE 2-8: INCREASED DOF THROUGH PROCESS DESIGN ................................................................ 35
FIGURE 3-1: SCHEMATIC OF AN INJECTION MOLDING MACHINE .......................................................... 39
FIGURE 3-2: PRESSURE PROFILE THROUGHOUT THE INJECTION MOLDING PROCESS........................... 41
FIGURE 3-3: CHANNEL FLOW ............................................................................................................ 44
FIGURE 3-4: COORDINATE SYSTEM FOR LAMINAR FLOW .................................................................... 45
FIGURE 3-5: FORCE BALANCE FOR SCREW AND VALVE STEM ............................................................ 47
FIGURE 3-6: STATE SPACE REPRESENTATION OF PLANT ................................................................... 50
FIGURE 3-7: BLOCK DIAGRAM OF PLANT ........................................................................................... 51
FIGURE 3-8: RHEOLOGICAL BEHAVIOR OF UNFILLED, MEDIUM VISCOSITY POLYCARBONATE ................ 53
FIGURE 3-9: COMPRESSIBILITY BEHAVIOR OF MEDIUM VISCOSITY POLYCARBONATE ........................... 54
FIGURE 3-10: SIMULATION RESULTS FOR CONSTANT VALVE POSITION OF 40 MM .............................. 57
FIGURE 3-11:CAVITY FILLING PRESSURE HISTORY AS A FUNCTION OF VALVE POSITION IN MM ........... 58
FIGURE 3-12: CAVITY PACKING PRESSURE AS A FUNCTION OF VALVE POSITION IN MM ...................... 59
FIGURE 3-13: SIMULATION RESULTS WITH OPENING AND CLOSING VALVE STEM ................................ 60
FIGURE 3-14: ADAPTIVE GAIN SCHEDULING....................................................................................... 65
FIGURE 3-15: GAIN SCHEDULING STRUCTURE ................................................................................... 66
FIGURE 3-16: RANDOM PROCESS INPUT AND RESULTING FILLING STAGE RESPONSE .......................... 67
FIGURE 3-17: COMPARISON BETWEEN LINEAR AND PLANT MODELS TO RANDOM INPUT ...................... 68
FIGURE 3-18: ROOT LOCUS OF LINEAR MODEL FOR FILLING STAGE ................................................... 69
xv
FIGURE 3-19: CLOSED LOOP RESPONSE OF PLANT IN FILLING STAGE ................................................ 70
FIGURE 3-20: RANDOM PROCESS INPUT AND RESULTING PACKING STAGE RESPONSE........................ 71
FIGURE 3-21: COMPARISON BETWEEN LINEAR AND PLANT MODELS TO RANDOM INPUT ...................... 72
FIGURE 3-22: ROOT LOCUS OF LINEAR MODEL FOR THE PACKING STAGE.......................................... 73
FIGURE 3-23: ROOT LOCUS OF LINEAR MODEL FOR THE PACKING STAGE WITH LEAD NETWORK ........ 74
FIGURE 3-24: CLOSED LOOP RESPONSE OF PLANT IN THE PACKING STAGE ....................................... 75
FIGURE 3-25: PROCESS CONTROL ALGORITHM.................................................................................. 77
FIGURE 3-26: EFFECT OF TIME SCHEDULE ON PACKING STAGE ......................................................... 79
FIGURE 3-27: EFFECT OF INITIAL VALVE POSITION ON PACKING STAGE ............................................. 80
FIGURE 3-28: CONVERGENCE TO DESIRED PROFILE .......................................................................... 82
FIGURE 4-1: PICTURE OF CONTROL SYSTEM ..................................................................................... 86
FIGURE 4-2: PICTURE OF MOLDING MACHINE .................................................................................... 87
FIGURE 4-3: CROSS-SECTION OF FEED SYSTEM ............................................................................... 89
FIGURE 4-4: MOLD INSERT GEOMETRIES ........................................................................................... 90
FIGURE 4-5: EFFECT OF HYDRAULIC LEAKAGE AND COMPRESSIBILITY ON VALVE POSITION ................ 92
FIGURE 4-6: EFFECT OF CAVITY PRESSURE ON VALVE RESPONSE TIME ............................................ 93
FIGURE 4-7: INPUT PROFILES FOR SIMULATION: VALVE POSITION AND INJECTION PRESSURE .............. 94
FIGURE 4-8: SIMULATION AND EXPERIMENTAL PRESSURE PROFILES FOR CONSTANT VALVE POSITION 95
FIGURE 4-9: TYPICAL PRESSURE PROFILE WITH MEASUREMENT INDICES ........................................... 97
FIGURE 4-10: FILLING STAGE PRESSURE PROFILE WITH NORMAL AND HIGH GAINS ............................ 99
FIGURE 4-11: PACKING STAGE PRESSURE PROFILE WITH NEGATIVE FEEDBACK ............................... 101
FIGURE 4-12: PACKING STAGE PRESSURE PROFILE WITH NEGATIVE FEEDBACK ............................... 102
FIGURE 4-13: PACKING STAGE PRESSURE PROFILES FOR CONSTANT VALVE POSITION .................... 103
FIGURE 4-14: TYPICAL ADAPTIVE CAVITY PRESSURE CONVERGENCE IN FOUR SHOTS ...................... 104
FIGURE 4-15: NUMBER OF SHOTS REQUIRED TO MANUFACTURE GOOD PARTS ................................ 105
FIGURE 5-1: MULTI-CAVITY MOLD WITH UNBALANCED FILL .............................................................. 109
FIGURE 5-2: PRESSURE PROFILES FOR SMALL AND LARGE CAVITIES WITH CONVENTIONAL MOLDING 110
FIGURE 5-3: PRESSURE PROFILES FOR SMALL AND LARGE CAVITIES WITH DYNAMIC FEED............... 111
FIGURE 5-4: EFFECT OF CAVITY PRESSURE SLOPE ON FLOW RATE THROUGH VALVE ...................... 112
FIGURE 5-5: MULTI-GATED PART .................................................................................................... 115
FIGURE 5-6: PRESSURE PROFILES FOR 3MM MULTI-GATED PART .................................................... 115
FIGURE 5-7: EFFECT OF CAVITY PACK PRESSURE ON LINEAR SHRINKAGE ....................................... 117
FIGURE 5-8: RANGE OF CAVITY PRESSURE RESPONSES IN CONVENTIONAL MOLDING ....................... 121
FIGURE 5-9: RANGE OF CAVITY PRESSURE RESPONSES WITH DYNAMIC FEED ................................. 122
FIGURE 5-10: EFFECT OF INPUT NOISE ON FILLING STAGE WITH CONVENTIONAL PROCESS.............. 123
FIGURE 5-11: EFFECT OF INPUT NOISE ON FILLING STAGE WITH DYNAMIC FEED ............................. 124
FIGURE 5-12: EFFECT OF INPUT NOISE ON PACKING STAGE ............................................................ 125
xvi
FIGURE 5-13: EFFECT OF INPUT VARIATION ON PART DIMENSIONS .................................................. 126
FIGURE 5-14: EFFECT OF SMALL DEVIATIONS IN TRANSITION TIME ON CAVITY PRESSURE RESPONSE132
FIGURE 5-15: EFFECT OF VALVE INTERFERENCE ON CAVITY PRESSURE .......................................... 134
FIGURE 6-1: COUPLING IN THE DEVELOPMENT OF INJECTION MOLDED PARTS .................................. 139
FIGURE 6-2: CLASSIC DESIGN TO PRODUCTION DEVELOPMENT PROCESS ........................................ 140
FIGURE 6-3: HIERARCHY OF DESIGN RELATIONSHIPS ...................................................................... 142
FIGURE 6-4: PROPOSED PRODUCT DESIGN METHODOLOGY ............................................................. 144
FIGURE 6-5: DESIGN EVALUATION ................................................................................................... 149
FIGURE 6-6: TYPICAL MOLDED PART AND SPECIFIED DIMENSIONS ................................................... 158
FIGURE 6-7: STOCHASTIC SHRINKAGE BEHAVIOR OF POLYCARBONATE ............................................ 160
FIGURE 6-8: CENTER-GATED BOX................................................................................................... 163
FIGURE 6-9: DYNAMIC FEED CONTROL OF THREE GATES ................................................................ 165
FIGURE 6-10: TWO GATES, OPTIMAL PRESSURES ........................................................................... 166
FIGURE 6-11: PROCESS CONTROL FOR MULTI-OBJECTIVE OPTIMIZATION ......................................... 169
xvii
Nomenclature
PROCESS MODEL
βmelt Bulk modulus of polymeric melt
βT Coefficient of volumetric thermal shrinkage
η Viscosity of polymeric melt
ρ Density of polymeric melt
aram Acceleration of injection cylinder (ram)
Ahyd Cross-sectional area of injection cylinder
Aram Cross-sectional area of ram in contact with polymeric melt
Avalve Cross-sectional area of valve stem
D Mean diameter of valve stem
hvalve Valve gap thickness
h Wall thickness of mold cavity
Lflow Distance from gate to melt front
Lvalve Length of tapered section of valve stem
m& Mass flow rate
Mram Mass of ram
Phyd Hydraulic pressure at injection cylinder
Pinjection Melt pressure at front of ram
Ppack Pack pressure level

Pslope Slope of cavity pressure in filling stage
Pcavity Melt pressure at gate of cavity
Rvalve Resistance to flow through valve
Rcavity Resistance to flow through cavity
tp Control estimate of transition between filling and packing stages
tstart Start of rise of cavity pressure for filling stage
xviii
u Flow velocity, function of thickness position
vram Velocity of injection cylinder (ram)
Vmelt Volume of polymeric melt in cavity
xp Control estimate of valve stem position for packing stage
w Width of mold cavity
CONTROL DEVELOPMENT
A Square (4 by 4) matrix representing the internal system dynamics
B Rectangular (4 by 2) matrix representing the effect of inputs on system
dynamics
Kd Derivative feedback gain for pressure
Kp Proportional feedback gain for pressure
Kx Proportional feedback gain for position
t Time at a given point, k
k Discrete time step
u Column vector (2 by 1) of system inputs: Phyd, vvalve
x System states: vram, Pcavity, xvalve, Lflow
xi Column vector (4 by 1) of system state at a given discrete time step, k=i
y Output of closed loop control system
ye Error between output and input of closed loop control system
yr Input to closed loop control system
OPTIMIZATION
G Objective function, relation sought to be optimized
Π Performance function, measure of the overall product performance
ℜ Product robustness, a measure of product performance relative to
range of acceptable product performance
xi Independent, adjustable design variables, set by designer
yi Product characteristics, outcome of design variables on performance
wi Value functions: a comparator of importance
xix
OPTIMIZATION EXAMPLE
Li Molded part’s length of dimension i
L iÆ j Distance from point i to j
LSL Lower specification limit of product requirement
n Number of performance parameters to be satisfied
Pi Pressure around point i in cavity
Φ Normal probability density function
Φ-1 Normal probability density function
s$ max Non-deterministic, maximum linear shrinkage
ds$ Non-deterministic relationship of pressure to linear shrinkage
dPi
si Linear shrinkage of dimension i
τ Target, or mean, of product requirement
Ti Temperature around point i in cavity
µx Mean of molded part properties
~
Ui Stochastic base value of property i
USL Upper specification limit of product requirement
~
Vi Stochastic temperature to property i correlation matrix
~
Wi Stochastic pressure to property i correlation matrix
xi Desired length of dimension i, or value of other part attribute
yi Expected yield of parts within tolerance of dimension i
xx
DYNAMIC FEED CONTROL: A NEW METHOD FOR
INJECTION MOLDING OF HIGH QUALITY PLASTIC PARTS
David Kazmer, Ph.D.

Stanford University, 1995
Adviser: Philip Barkan
The thrust of thermoplastic materials into advanced technical applications has

resulted in industry requirements which exceed the capabilities of available product
development and manufacturing processes. The lack of robustness in these processes
is evidenced by the long product development cycles, excessive tooling costs, low
process yields, and inferior product quality. Dynamic Feed Control is a major
revolutionary concept — an innovative industrial process that opens up new potential
opportunities in the manufacture of plastic parts. With the cooperation of industry,
government, and academia, this concept has been implemented and validated for the
design and production of high quality, molded plastic parts.
The invention utilizes multiple valves in the feed system of a mold to selectively
regulate the flow to each area of the cavity in response to real-time feedback from
nearby cavity pressure transducers. A minimum complexity process model was
identified for development of a closed-loop control strategy. Using an adaptive gain
scheduling approach, the cavity pressures at each gate were controlled throughout the
filling and packing stages. This enables the process dynamics to be modified so as to
produce parts with the desired part properties without retooling mold steel.
The flexibility of the process was demonstrated by its ability to control flow rates
and knit-line location in the filling stage as well as pack pressure and part dimensions in
the packing stage. Moreover, the consistency of the proposed process was compared
to the conventional process. This comparison was achieved by intentionally adding
process noise in an experiment designed to simulate natural material, machine, and
operator variation. Analysis of the experimental results showed an increase in the
process capability, Cp, from 0.56 for the conventional molding process to 1.67 for
Dynamic Feed Control.
Finally, a methodology for the design of molded parts was introduced which
leverages the degrees of freedom provided by Dynamic Feed Control. Product
robustness was demonstrated by a test series based on the stochastic (probabilistic)
distribution of material properties during the molding process. With this assumption, the
production yield of various design and processing strategies were evaluated. The
methods presented here are directly extensible to arbitrarily complex applications with a
broad range of properties, requirements, and specifications.
Approved for publication:
By: _________________________
For Mechanical Engineering
ii
To the University Committee on Graduate Studies:
I certify that I have read the dissertation of David Kazmer, Dynamic Feed Control:
A New Method for Injection Molding of High Quality Plastic Parts, in its final form for
submission and have found it to be satisfactory.
_________________________________
___
(Signature) (Date)
Kos Ishii
ME Design Division
iii
_________________________________
___
(Signature) (Date)
Phil Barkan
ME Design Division
iv
_________________________________
___
(Signature) (Date)
Fritz Prinz
ME Design Division
v
Chapter 1: Introduction
TABLE OF CONTENTS
CHAPTER 1: INTRODUCTION ................................................................................................................2

THE PLASTICS INDUSTRY ........................................................................................................................2
Historical Perspective ...............................................................................................................................2
Current Status............................................................................................................................................4
RECENT DEVELOPMENTS ........................................................................................................................5
Advances in the Design Process................................................................................................................6
Example................................................................................................................................................................. 6
Computer Technologies......................................................................................................................................... 7
Advances in the Injection Molding Process ..............................................................................................9
Control Systems .................................................................................................................................................. 10
Alternative Technologies .................................................................................................................................... 13
PROCESS LIMITATIONS ..........................................................................................................................16
Complex Material Behavior ....................................................................................................................16
Sensitivity of Product Quality to Input Variation ....................................................................................18
Inability to Adapt Process to Changing Production Requirements.........................................................21
SUMMARY .................................................................................................................................................22
TABLE OF FIGURES
FIGURE 1-1: CLASSIC DESIGN TO PRODUCTION PROCESS ..................................................................................6
FIGURE 1-2: MA’S DESCRIPTION OF THE INJECTION MOLDING PROCESS........................................................11
FIGURE 1-3: MODEL-BASED PROCESS OPTIMIZATION .....................................................................................13
FIGURE 1-4: PUSH-PULL DEVICE.....................................................................................................................14
FIGURE 1-5: VISCOSITY OF POLYCARBONATE ................................................................................................17
FIGURE 1-6: PARETO CHART OF COMMON DEFECTS ........................................................................................19
TABLE OF TABLES
TABLE 1-1: PROCESS TO MODIFY MOLD ........................................................................................................21
START OF CHAPTER:1
START OF ENDNOTES: 1
1
Injection molding of thermoplastics has emerged as the premier vehicle for

delivering high quality, value added commercial products. Continued global
competitiveness has increased standards for product capability and quality while
requiring reduced product development time and unit cost. Despite advanced
design methods and new process technologies, it is becoming apparent that the
injection molding process is neither flexible nor robust.
THE PLASTICS INDUSTRY

While modern commercial products frequently utilize thermoplastic
components, the widespread penetration of plastics into technical and
demanding applications is a recent occurrence. In the past, plastics were
regarded as cheap imitations for more expensive materials. This perception has
changed with the development of improved materials and their extensive use in
demanding technical applications.
Historical Perspective
The first plastic compound, celluloid, was created in 1869 by John Hyatt to
win a $10,000 prize from the Pheland and Collander Company for replacement of
ivory in the production of billiard balls.1 Once crude pressure forming techniques
were developed, this material was also used to create inexpensive products such
as buttons, combs, and thimbles. Interestingly, significant effort was taken to
make these products appear like ivory or tortoise shell even though they could
have been more easily molded to yield vibrant colors or unique textures.2
Derivatives of celluloid were soon introduced as replacements for cotton and
leather, finding immediate use in commodity goods such as baby pants,
2
footwear, golf balls, and gloves.3 As implied by these early applications, plastics
were regarded solely as replacements for traditional materials.
Fundamental research in the 1930’s led to the development of nylon and

polyethyelene. Again, the initial uses of these new polymers were for commodity
goods. With the coming of World War II and developments in other technical
fields, plastics were introduced in technical applications which exploited their
unique material properties. By 1941, nylon was being used in self-lubricating
bearings, wire insulation, and surgical devices. Polyethylene was used as
insulation in submarine and radar cable applications where low water absorption
and high dielectric strength are essential.4 For these applications, thermoplastics
provided properties superior to any other natural or synthetic material.
Even though plastics continued to be used in technical applications after

the war, public awareness of their capabilities remained limited. Poor designs
and trivial uses of plastics perpetuated the notion that plastic products were
merely cheap alternatives for more expensive goods. With the baby boom after
the war, inexpensive plastic toys were imported in enormous quantities from
Japan. The term “Made in Japan” became synonymous with “cheap plastic” — a
misleading assessment which perpetuated for several decades.5
Due to the cold war and the space race, the 1960’s were a period of rapid
developments for the plastics industry. New applications required material
properties which existing commodity resins could not deliver. Hence, new
polymers were invented with significantly improved mechanical, electrical, and
thermal properties. These engineering thermoplastics became replacements in
structural applications previously utilizing aluminum, brass, and steel.
Automobile manufacturers began to utilize plastics to increase component
functionality, reduce part count and vehicle weight, and simplify vehicle
assembly. These new materials quickly penetrated most commercial markets as
they displaced traditional materials in general applications and opened up entirely
new applications which leveraged their unique material properties.
3
Current Status
By 1985, the volume of plastics produced in the United States exceeded

that of steel, aluminum, and copper combined.*,6 Domestic production in 1992
exceeded 20 million cubic meters of plastic resin, accounting for sales of $30
billion in raw materials and $71.3 billion in plastics products.7 With new material
and further process development, the plastics industry is expected to maintain its
historical growth of several percentage points above the general manufacturing
sector.
The injection molding process, described later in Chapter 3, is the most

common method of forming thermoplastic materials into commercial goods. This
process allows large volume, economical production of arbitrarily complex, three-
dimensional geometries with no post-molding operations. Due to its process
capability, injection molding of engineering thermoplastics has emerged as the
prevalent method for production of high quality, value added components.
As applications become increasingly technical, competitive pressures are

placing new demands on the development and production of plastic parts
requiring greater precision, reduced unit cost, and shorter development times.
Moreover, the continuing drive to integrate maximum functionality into a single
molded component has vastly increased the number of product requirements and
process specifications. Simultaneously, these specifications are requiring tighter
dimensional control with stringent quality assurance. As plastics continue to
penetrate into increasingly technical and demanding applications, many critical
commercial opportunities are limited by current process capabilities.
Manufacturers of plastic parts are constantly striving to reduce unit cost.

This has been achieved by optimizing the molding process to reduce cycle time
and increase process yield. Alternatively, lower cost commodity resins have
*
1985 Production of steel, aluminum, copper, and plastic were 76, 7.7, 2.0, and 20 million metric
tons, respectively. With specific gravities of 7.9, 2.7, and 8.7 for the metals vs. plastic’s specific
gravity of 1.0, the total volume of metals is approximately 15 million cubic meters, less than the
20 million cubic meters of plastic produced.
4
begun to displace engineering thermoplastics in less demanding technical

applications. Also, new designs utilizing engineering thermoplastics are lowering
costs by utilizing thinner wall thicknesses to reduce molding cycle time and part
weight – where flow length to wall thickness ratios of 50:1 were standard, ratios
of 100:1 are now common and quickly approaching 200:1. However, shifting to
commodity resins with inferior properties while reducing part wall thickness
makes parts more difficult to produce due to increased susceptibility to aesthetic
and dimensional defects.
With a greater number of product manufacturers competing for a given

market, global competitiveness also demands shorter time to market. Shorter
time to market requires shorter product development time, fewer tooling
iterations, and increased manufacturing productivity. Thus, market forces are
demanding increased product quality and design sophistication with reduced unit
costs. These conflicting requirements can not be fully resolved within the current
capabilities of the injection molding process.
RECENT DEVELOPMENTS
Due to the intrinsic limitations of the injection molding process, the
industry has historically utilized an iterative development cycle for part design
and manufacture as shown in Figure Chapter 1: -1.8 In this scenario, the quality
of the design decisions and process capability can only be determined after the
mold is tooled and the plastic product is sampled and measured. While costly,
this development process was adequate for large production volumes of a few
product designs. Such applications could rationalize the long lead times and
development costs. However, iterative development cycles have proven
excessively costly for shorter production volumes with increasing product variety.
5
Specifications
Part Design
Prototype Mold Design
Rework
Evaluation
Accept
Production
Figure Chapter 1: -1: Classic design to production process
Advances in the Design Process
Continued competitive pressures during a recessionary period have driven

firms to seek ways to reduce product costs without compromising product quality.
In the last decade, significant advances have occurred in the plastic part
development cycle to reduce development risk while improving product quality.
Additionally, modifications of the injection molding process have enhanced the
manufacturing capability.
Example
In the short life cycle typical of computer and communication markets,

each month of delayed product development time may represent a loss in 10% of
cumulative life-time sales for the product.9 Similar conditions have fueled the
drive to reduce product development times in most industries. Occasionally, the
product launch or product shipment may be delayed due to the unavailability of
acceptable plastic components. This can be devastating to a company’s profits if
not its future solvency. Due to problem’s with the computer’s housing, for
instance, Apple’s announcement of its first PowerPC based notebook was
6
delayed from Christmas, 1994 to second quarter, 1995 in a volatile period in the
company’s history.10
Product development times have been reduced significantly in the last

decade. The time required to manufacture commercial plastic parts from a
design concept has been reported to be as little as three weeks.11 This has
become possible due to three factors: increased knowledge regarding plastics,
improved design processes, and integration of computer technologies.
Computer Technologies
The utilization of computer technologies for Computer Aided Design (CAD)

has had a dramatic impact on the plastic part design process. CAD has
permitted quick, accurate drawings and effortless maintenance of product
concepts. The development and integration of standard design and
manufacturing data bases have greatly reduced engineering design times.12,13
With its internal design and tooling functions, Cap-Snap (San Jose, CA) is able to
re-design and re-tool a fairly simple product in one day – reduced from three
weeks just four years ago.14 Moreover, CAD has become a common form of
communication between design firm, tooling house, and molder. With the
parallel evolution of automatic tool path generation through Computer Aided
Manufacturing, average reductions of 30% in the manufacturing development
times have been achieved in the last five years.15 Roughly 90% of all design
firms and tool shops are anticipated to rely upon computer aided design and
manufacturing systems by 1999.16
The evolution of CAD has also facilitated the development and integration
of new rapid prototyping methods. Rapid prototyping allows product designers
and tool engineers to review full scale, detailed models of the plastic parts before
tooling begins – and hopefully avoid unforeseen design or tooling problems.
Moreover, new tooling techniques have been created to use the full scale models
as patterns for the casting of molds to reduce tool development time.
7
Concurrently, process simulation has become an integral tool in the

design cycle of most commercial thermoplastic parts. Fundamental research of
the injection molding process began with Spencer and Gillmore in the 1950’s with
an empirical investigation of melt flow advancement in injection molding. 17 Harry
and Parrott were the first to introduce process simulation for injection molded part
design in 1970.18 Their numerical implementation utilized a finite difference form
of the heat equation to predict the melt flow advancement down a long, narrow
strip for a specific material and injection pressure. Although their results were
inaccurate, the analysis could be utilized to qualitatively compare candidate
designs and materials for commercial applications.
Williams and Lord advanced the simulation of the injection molding

process by discretizing both the length and thickness dimension to track the melt
front propagation while simultaneously performing heat transfer calculations.19
This was the first analysis to consider the dynamic build-up of a solidified skin
layer as well as the polymer’s complex non-Newtonian (shear dependent)
rheological behavior. This analysis tool was applied to molding equipment
design and problem diagnosis.
By 1980, commercial simulations had been introduced by Moldflow Ptd20

and the Cornell Injection Molding Program21 for use in part design and process
troubleshooting. These simulations represented two-dimensional geometries by
a series of one-dimensional flow segments; each segment could represent a
circular runner, end-gated strip, or center gated disk. Once these “lay-flat”
techniques had been developed, simulations became a major tool in evaluating
product design and ensuring manufacturability. Fueled by lucrative commercial
prospects, research into more advanced numerical schemes such as the hybrid
finite element/finite difference method enabled tracking the melt propagation in
arbitrarily complex three-dimensional geometries.22 These simulations became
prevalent in the mid-1980’s due to their more accurate prediction and rendering
of flow fields and cavity pressure distributions. Product designers and process
8
engineers became accustomed to performing multiple analyses to avoid costly

design mistakes and optimize existing molding processes.
Mold-filling analyses are now a formal tool in many firms’ plastic part
design processes.23,24 Researchers have continued the development of more
sophisticated injection molding simulations and fundamental examination of the
complex rheological and viscoelastic behavior of thermoplastic materials.25,26
Research on the simulation of the plastication, packing, and cooling stages27,28
seeks to predict the residual stresses,29 fiber orientation,30 and birefringence31 of
the final molded product. The ultimate goal of these simulations is not to predict
the flow fields and pressure distributions generated in the molding process but
rather to yield exact estimates of the molded part’s final end-use characteristics
such as appearance, final dimensions, structural stiffness, long-term creep, etc.
Analysis techniques have already produced fair estimates of many of these
properties and will continue to improve with further research, modeling, and
validation. 32,33
However, computer simulations will always be, at best, an approximation

of the actual molding process and the candidate designs the molding process
can produce. Simulation can be used to optimize a part design and
manufacturing, but the actual manufacture of the product will be constrained by
the inherent limitations of the material and molding process. As such, the
injection molding process is evolving, attempting to provide greater process
repeatability, flexibility, and robustness.
Advances in the Injection Molding Process
After the modern reciprocating screw machine was developed in 1972,

there were relatively few technical changes until the mid-1980’s. Driven by
increased competition and advances in the electronics industry, the injection
molding process has recently undergone a series of significant advances in
process capability. While many of these developments can be classified as
9
incremental changes to the molding machine’s control systems, some

adaptations include the addition of major subsystems which significantly alter the
functionality and capability of the molding process.
Control Systems
Prior to 1972, injection molding machines controlled the injection pressure

through the use of a pressure relief valve between a hydraulic accumulator and
the injection unit’s cylinder. At the start of the cycle, hydraulic flow from the
accumulator would push the ram forward at an arbitrary rate. After the cavity
was filled, the pressure relief valve would maintain a maximum holding pressure
until the part was packed and began solidification. Machines with advanced
control systems could use two relief valves to enable a time-based switchover
from higher injection pressures to lower holding pressures during solidification.
Electrically modulated pressure relief valves were introduced in 1975 to provide
solid state control of the injection pressure throughout the molding cycle.34
The first modern computer-controlled injection molding machine was first

described by Carl Ma in 1974 while employed by Cincinnati Milacron.35 In this
work, Ma described the molding process as shown in Figure Chapter 1: -2,
identified the critical process variables, and designed a control system for each
process stage. This work led to the development of modern control systems for
injection molding machines and enabled current closed loop control systems for
ram velocity and injection pressure.35,36
10
EXTERNAL INPUT VARIABLES
Machine Wear, Hydraulic Leakage,

Raw Material Ambient Temp., Cycle Time Interrupt
MACHINE
INPUT
VARIABLES
PROCESS
Barrel Temp. PLASTICATION Melt Pressure OUTPUT
PROCESS VARIABLES
Extruder RPM Melt Temp.
Back Pressure Melt Volume INJECTION Melt Pressure Cav. Pressure

PROCESS
Injection Velocity Melt Temp. Cav. Temp.
PACKING
Injection Pressure Melt Volume & Part Dimensions
COOLING
Packing Pressure PROCESS Part Strength
Mold Temp. Part Appearance
Figure Chapter 1: -2: Ma’s Description of the Injection Molding Process
Control systems using servo-valves to obtain closed loop control of ram

position and hydraulic pressure were commercially available by 1980. While
closed loop ram velocity and injection pressure control significantly improved the
process capabilities of injection molding, the output state variables in the cavity
which Ma identified to determine product quality remain uncontrolled. In this
sense, injection molding is an open loop process once the material leaves
the barrel of the molding machine. When quality issues arise, there are no
available process parameters which may be used to directly control state
variables and resolve the problem.
By monitoring cavity pressure as an input signal, closed loop cavity

pressure control could automatically compensate for variations in melt viscosity,
injection pressure, and hydraulic subsystems to achieve a consistent process
and uniform product attributes. Mann first described this control scheme in 1974
and employed an iterative algorithm to achieve a functional process using
modulated pressure relief valves.37
The use of closed loop cavity pressure control has not yet become
common even though significant improvements in process repeatability have
38
been reported. Since the characteristics of the mold are fixed once tooled,
11
closed loop cavity pressure control provides control at only a single point – an
expensive and limited form of quality control to reduce the effects of input
variation. With more robust sensor technology and less expensive computer
control systems, however, closed loop cavity pressure control is beginning to be
used in applications demanding guaranteed process repeatability.
More sophisticated model-based control systems have yielded only

marginal improvements in process capability and product quality. In these
control schemes, analytical models are used to generate real-time estimates of
unobservable process variables such as residual stress, shear rate, or fiber
orientation. Computations are based on measurable inputs such as injection
velocity, melt temperature, and cavity pressure. The control system can then use
the analytical models in its control law to determine the molding machines
dynamics to optimize part properties, many of which could only be ascertained
after molding, if at all.
Many model-based controllers have been developed. Shankar was first to

develop a non-linear model-based control system in 1978 to optimize ram
velocity with a discrete iterative control method.39 Agrawal et. al. have reviewed
several recent process control strategies and proposed a system which
maintains consistent density throughout the part based on an equation of state
for the material’s compressibility behavior.40 Based on this work, several
adaptive model-based controllers have recently been developed to remove
machine dependency and increase the product quality and consistency.41,42,43,44
Another potential benefit of model-based control is the resolution of

multiple conflicting goals given the constraints of the molding process. In 1991,
Seaman devised a machine controller to monitor the molded part’s quality and
maintain the process on an optimal trade-off boundary to avoid performance
degradation while trying to improve one of the part’s characteristics.45 For
instance, Seaman utilized the procedure shown in Figure Chapter 1: -3 for a
spiral mold tool to prevent flashing while minimizing the cycle time and part to
12
part deviation in flow length. Similar research has integrated the analysis
capabilities of process simulations to optimize process control.46
Process
Disturbances
Controller
Machine Performance
Inputs Machine Performance Measure
+
Dynamics Evaluation
Automatic
Controller
Multi-Obj
Optimization
Figure Chapter 1: -3: Model-based process optimization
Alternative Technologies
Process control systems for injection molding have become substantially

more repeatable and capable of producing consistent products. A consistent
product, however, does not infer high levels of quality or desirable product
attributes. The most sophisticated control systems will ultimately be able to
eliminate variation while making trade-offs between multiple production goals.
However, the control system will not be able to resolve multiple conflicting goals
since most of the process dynamics are embedded in the mold steel. This
couples the process dynamics between different areas of the mold – any process
change imposed to improve one part property can inadvertently reduce quality
elsewhere in the part.
New process technologies have been incorporated into the injection

molding process to address specific quality issues. Ibar et. al. has recently
developed a device which utilizes reciprocating action of one or more melt-
accumulator pistons adjacent to the flow path to induce melt-flow oscillation in
13
the post-filling stages of the molding process.47,48,49 These flow-fields can be

used to alter and improve the extent of orientation in amorphous plastics and the
morphology of semi-crystalline plastics. By orienting and vibrating the material,
the ‘rheometric’ process forces flow fronts to intermingle at the molecular level,
diffusing knit-lines and increasing part strength. The process is also claimed to
control shrinkage and internal stresses to reduce part warpage and birefringence.
Mold Cavity
Melt Pistons
Figure Chapter 1: -4: Push-pull device
Several recent efforts have described different methods of accomplishing

similar results. Gardner and Malloy utilized ejector pins on a cam to induce melt
oscillation after the filling stage.50 Becker et. al. utilizes two injection units to
‘push-pull’ the melt.51 Grossman et. al. accomplishes melt oscillation by inserting
pistons into a multi-branched runner system as shown in Figure Chapter 1: -
4.52,53 Finally, Kazmer and Roe increased knit-line strength in a conventional hot
runner system simply by closing one valve gate in the post-filling stage and
allowing the melt solidification and shrinkage to induce flow across the cavity.54
Knit-lines are one common structural defect which may be addressed with any of
the above techniques.
14
Dimensional stability has become an increasingly significant issue as

plastics penetrate further into technical applications with significant levels of
functional integration. To achieve dimensional stability, cavity pressures during
molding must be uniform throughout the cavity to assure low levels of molded-in
residual stress and avoid post-molding warpage. Uniform pressure distributions
are difficult to achieve given the high viscosity of the plastic melt and the flow
resistance of most designs’ thin wall sections and long flow lengths.
Low pressure foam molding has been developed as a method to produce

parts with uniform melt flow and reduced residual stresses.55 In this process, the
part is molded at lower pressures by introducing a blowing agent into the resin
during injection which creates a low viscosity foam in the cavity. With lower,
more evenly distributed cavity pressures, shrinkage and residual stresses are
fairly uniform. While foam molding is especially suited to manufacture of parts
with thick wall sections, the chaotic nature of foam molding makes tight
tolerances impossible.
Gas-assisted injection molding was developed to enable hollow molding of

thick wall-sectioned parts while delivering uniform cavity pressures in the packing
stage. The process differs from injection molding in that a gas is introduced into
the molten core of a partially filled cavity at the end of the injection stage. With
negligible viscosity, the gas transmits the gas pressure uniformly throughout the
cavity which results in reduced residual stress and part distortion.56
Unfortunately, gas-assist molding generates a new set of unfamiliar problems, for
instance: blow through, in which the gas penetrates through the surface of the
part; fingering, in which the gas escapes the designed gas channels and
penetrates into thin areas of the part. Moreover, gas assist reduces the
dimensional stability when compared to molding with closed loop cavity pressure
control.57
If low pressure molding does not relieve many of injection molding’s

limitations, work in high pressure molding promises to deliver a different set of
15
process capabilities. In high pressure molding, the injection unit is redesigned to

produce and deliver injection pressures above 400 MPa (60,000 psi). This
enables very short fill times and rapid transition into the packing stage. While
specialty polymers must be designed to withstand the shear rates and pressures,
high pressure molding does yield a fairly uniform cavity pressure distribution and
almost zero part shrinkage.58,59 Unfortunately, mold development for this
process is costly and many issues regarding control of residual stresses have yet
to be resolved.
PROCESS LIMITATIONS
In spite of these many advances, the injection molding process is not
staying abreast of the advancing needs of industry. Limitations of the injection
molding process stem from the inherent properties of the materials being molded
and the characteristics of the manufacturing process itself. While these
limitations are well understood, they often can not be avoided and result in costly
failures during product development and production.
Complex Material Behavior
Polymers exhibit very complex material behavior due to their molecular

structure. Difficulties with predicting a material's performance are further
exacerbated by the inherent nature of the molding process with rapid fluctuations
in temperature, shear rates, and pressures. The viscosity of a medium grade
amorphous polycarbonate is shown in Figure Chapter 1: -5. The complexity of
this material's behavior impedes the product development process in two ways.
First, product and tool engineers can provide only poor estimates of
manufacturability – small changes in design detail can prove critical. Second,
small changes in material temperature or shear rates due to varying process
conditions can produce large deviations in the melt viscosity.60 Product
development is then forced to be an iterative trial and error process as previously
shown in Figure Chapter 1: -1.
16
3
10
280C
Viscosity (PaSec)
300C
320C
2
10 2 3 4
10 10 10
Shear Rate (1/sec)
Figure Chapter 1: -5: Viscosity of polycarbonate
Unfortunately, the production of high quality plastic parts is further

complicated by thermoplastic materials’ complex thermal expansion and
compressibility properties. This behavior may lead to a 20% volumetric
expansion due to an increase in melt temperature as the material is plasticized.
Once the cavity is filled, the material undergoes significant contraction due to a
simultaneous increase in pressure and decrease in temperature. Since the
pressure and temperature history vary across a part during molding, the
compressibility behavior of polymers will cause the part to shrink non-uniformly.
Part dimensions frequently do not meet expected tolerances – sometimes
warpage is so severe that the part is rendered unusable.
Mathematical modeling of the rheological and compressibility behavior of

polymers has been an active field of research for forty years.61 Even if more
accurate characterization methods and material models were available, the
physics of the molding process are so complex that they could not be exploited
without more sophisticated numerical methods.
17
As a result, mold designers have historically resorted to overly

approximate formulas for balancing flow and have assumed familiar part
shrinkages in setting tool dimensions. Combined with intuition, these estimates
are used to generate an initial set of specifications which must be later be
modified as the mold is built and tested. Even with numerous design iterations,
however, the final molded product may not meet the original design intent and
frequently is far from desirable quality. A recent example in the automotive
industry occurred despite significant expertise with the plastic materials, the
injection molding process, and formal design methodologies:
“Chairman Eaton wasn’t smiling the first time he drove the

Cirrus. When Robert J. Eaton drove a prototype Cirrus sedan in March,
the Chrysler Corp. chairman knew the car needed lots of improvement
before it could roll into dealer showrooms. In his two days behind the
wheel at Chrysler’s proving ground north of Phoenix, Eaton noticed the
poor fit of dashboard pieces and other plastic interior trim...To help hide
unsightly gaps, engineers added a five millimeter lip to the joint where
the Stratus’ dashboard meets a piece of trim just under the windshield.”62
Tool changes late in the product development cycle may cost an

automobile manufacturer hundred of millions of dollars in the development of a
new model.63
Sensitivity of Product Quality to Input Variation
Even after the mold design has been tuned by successive design-build-
test iterations, quality issues frequently arise due to the sensitivity of the molding
process to input variation and random noise. Frequently, these quality problems
are only identified after the manufacturing process is complete since the defects
are not obvious to the process operators and are difficult to measure. Figure
Chapter 1: -6 is a Pareto chart from a molder – the most ‘common’ defects are
those which are visually obvious. Unfortunately it is the non-visual defects which
result in later failure during assembly or end-use.
18
45
40
35
30
25
20
15
10
5
Figure Chapter 1: -6: Pareto chart of common defects
These quality problems are most often attributed to fluctuations in material

properties. While variations in molecular weight and composition will effect the
rheological and compressibility behavior of the material being molded,64
variability can also be introduced through the practice of the process operators or
the performance of the injection molding machine. For instance, manufacturers
might be given a set of process conditions to mold a given component. Process
operators, however, can vary many of the machine settings significantly while
seeking to deliver the ambient process conditions. These operators can cause
significant part variations while believing they are optimizing the process.
Deviations in part consistency are common between different shifts and different
manufacturing sites.
Even if variation due to the inconsistencies of process operators could be

removed, similar variation can be introduced from the varying behavior of
injection molding machines – the machine settings are not necessarily
representative of the conditions of the plastic melt during the molding
process. Melt temperature and injection pressure, for instance, may deviate
19
from the controlled barrel temperature and hydraulic pressure. Moreover,

differences in machine design (not to mention wear of mechanical and hydraulic
components) can induce significant variations in residence time, melt
temperature, injection pressure dynamics, etc. and are the source of variations in
part consistency between machines and manufacturing sites. For instance, a
printer housing might be run on either a 300 or 400 ton press due to machine
availability. The 400 ton machine would typically induce longer residence times
and slower response times – this will alter the material and process behavior and
lead to product variations.
To achieve consistent products it might seem the optimal method would

be to manufacture all desired parts using one operator on one machine with a
single, huge lot of material. This is not economically feasible or even physically
possible. Moreover, there can still be significant deviations in quality between
consecutive molded parts as will be shown later. The effects of product variation
can be daunting, as shown by one illustrative example which occurred in 1994 at
a world famous printer manufacturer, renown for its product innovativeness and
quality:
A printer manufacturer assembles and distributes roughly

100,000 ink jet printers per month with three primary subassemblies:
print engine, electrical components, and plastic housing. The plastic
housing consists of upper and bottom halves which are fit together
during final assembly. To meet production quantities, there are five
identical molds for each side of the housing which are sent to different
molders who manufacture the parts. When some of the parts reach final
assembly, however, the two halves don’t fit together, the printer can not
be assembled, and the production line stops.65
These difficulties arose despite the use of sophisticated flow analysis,

documented design and manufacturing processes, and stringent quality
assurance methods at the manufacturing site. The sensitivity of product quality
to input variation resulted in lengthy delays and excessive costs in the
20
manufacturing and assembly of these products. Unfortunately, there might be

additional undetected quality costs associated with part quality which are
discovered by the end user resulting in repair charges, damaged reputation, or
lost future revenue. To ensure product consistency, this manufacturer is
considering utilizing closed loop cavity pressure control in all their molded
housings.
Inability to Adapt Process to Changing Production Requirements
Commercial products are often subject to a prolonged development cycle

and excessive tooling costs because of the inflexibility of injection molding
process. This inability to adapt is intrinsic to injection molding since the process
characteristics are immutable once design parameters have been locked into the
mold steel. Process attributes can only be changed by reshaping the mold steel
by machining (or welding and then machining) the needed changes as shown in
Table Chapter 1: -1 for a typical commercial part. Given that the design
parameters were based only on approximate estimates and tool modifications
are likely, development of commercial plastic components is inherently a lengthy
and costly process.
Table Chapter 1: -1: Process to Modify Mold

Process Step Minimum Time
Required
Remove mold from injection molding machine 2 hours
Ship mold to tooling shop 1 day
Disassemble mold 2 hours
Make modifications 4 hours
Assemble mold 2 hours
Ship mold to molder 1 day
Hang mold in press and start production 2 hours
Design and production goals often change due unforeseen aesthetic

defects, performance issues, or variations in material pricing. Moving knit-lines,
adjusting the cavity pressure distributions, or utilizing different viscosity resins,
are typical instances requiring costly modifications of the feed system to achieve
21
these production goals – the injection molding process does not have the
flexibility required to achieve different flow patterns and pressure distributions.
These changing production requirements may necessitate several cycles of
tooling modifications during the development and production of a commercial
product.
The Jeep Grand Cherokee's air conditioner is contained within

two halves of a plastic housing. The two large halves must maintain
tight dimensions at several points for fixturing to the other half as well as
for inlet and exhaust pipes to the vehicle's ventilation system. To hold a
vacuum within the unit, both housings must also remain flat around the
perimeter.
Advanced process simulations were utilized to optimize part

design and predict the final part dimensions. Tooling was built based
upon the recommendations and produced parts of excellent dimensional
accuracy. Unfortunately, the parts were not stiff enough to accommodate
the loading of the air conditioner unit so the wall thickness and ribs were
subsequently altered to increase the unit's stiffness. This tooling change
caused significant warpage of the housing such that neither the flatness
nor position tolerances could be attained and the part was rendered
unusable.66
SUMMARY
In summary, the plastics industry is increasingly hard pressed to improve
performance, reduce cost, and shorten the product development time. To meet
these difficult goals, recent developments have attempted to reduce the inherent
limitations of the injection molding process but have produced mixed results.
Simulation has reduced difficulties in product design and processing of complex
materials but has not increased the capability of the actual molding process.
Similarly, closed loop feedback control reduces the sensitivity of the process to
input variation but does not introduce extra process flexibility. Finally, process
22
optimization techniques can only keep the process on a restrictive trade-off

boundary, a constraint that is far below current industry needs.
Despite these numerous technological advances, commercial applications

continue to experience significant difficulties in plastic part production, delaying
product time to market while incurring excessive costs. The industry needs a
method to increase the flexibility of the injection molding process to compensate
for unforeseen changes in material properties and production requirements while
also ensuring a high level of product consistency. One method for achieving
these goals is next introduced.
23
1
Hyatt, J. W., A Celluloidal Material for Ivory Replacement, United States Patent #892,300, 1869.
2
Friedel, R., The First Plastic, American Heritage of Invention and Technology, v. 2, n. 2, pp. 18 (Summer,
1987).
3
Time magazine, Plastics in the War, v. 40, n. 9, pp. 72 (31 August, 1942).
4
McCrum, N.G., Buckley, C.P., Bucknall, Principles of Polymer Engineering, Oxford Science Publications,
Oxford, 1989.
5
Foy, G. F., Engineering Plastics and Their Commercial Development, American Chemical Society,
Washington D. C., 1969.
6
Sheets, K. R., US News and World Report, v. 52, n. 2 (1986)
7
Brown, R.H., Garter, J.E., 1994 United States Industrial Outlook, United States Department of
Commerce/International Trade Administration, Washington D. C., 1994.
8
Muccio, E.A., Plastic Part Technology, ASM International, Material Park, Ohio, 1992, pp. 176.
9
Rosenthal, S. R., Tatticonda, M. V., Time Management in New Product Development: Case Study Findings,
IEEE Engineering Management Review, v. 21, n. 3, pp. 13, 1993.
10
Halper, M., Apple Delays Adding PowerPC to Portables, Computerworld, v. 28, n. 47, pp. 6 (1994).
11
Beal, C. I., Concept to Production in Three Weeks - Is it Possible?, Proceedings from the 1993 Annual
Technical Meeting of the Society of Plastics Engineers, v. 51, pp. 3158 (1993).
12
Groover, M. P., Zimmers, E. W., Computer-Aided Design and Manufacturing, Prentice-Hall, Englewood
Cliffs, NJ, 1984.
13
Bernhardt, E. C., Computer Aided Engineering for Injection Molding, Hanser Publishers, Munchen,
Germany, 1983.
14
Personal communication with Jack Watts, President and CEO of Cap-Snap Moldings, San Jose, California
(April 11, 1994).
15
Noaker, P. M., Mirror Image Moldmaking, Manufacturing Engineering, v. 113, n. 3, pp. 36-41 (September,
1994).
16
Muccio, pp. 181.
24
17
Spencer, R. S., Gilmore, G. D., Some Flow Phenomena in the Injection Molding of Polystyrene, Journal of
Colloidal Science, v. 6, pp. 118 (1951).
18
Harry, D., Parrot, R., Numerical Simulation of Injection Mold Filling, Polymer Engineering and Science, v.
10, pp. 209 (1970).
19
Williams, G., Lord, H., Mold-Filling Studies for the Injection Molding Process with PVC in Injection Molding,
Polymer Engineering and Science, v. 15, pp. 553 (1975).
20
Austin, C., Chapter 9 of CAE-Computer Aided Engineering for Injection Molding, E.C. Berhnhardt (ed.),
Hanser Publishers, Munchen, Germany, 1983.
21
Hieber, C.A., Shen, S. F., Flow Analysis of the Non-Isothermal Two-Dimensional Filling Process in
Injection Molding, Israel Journal of Technology, v. 16, pp. 248 (1978).
22
Wang, V. W., Hieber, C. A., Wang, K. K., Filling of an Arbitrary Three Dimensional Thin Cavity, Journal of
Polymer Engineering, v. 7, pp. 21 (1986).
23
Personal Communication with Suresh Shah, Project Manager, GM Inland-Fisher Guide, Troy, Michigan
(November 11, 1994).
24
Austin, C., Industrial Metamorphosis, Proceedings from the 1994 Annual Technical Meeting of the Society
of Plastics Engineers, v. 52, pp. 1626 (1994).
25
Kamal, M., Goyal, S., Chu, E., Simulation of Injection Mold Filling of a Viscoelastic Polymer with Fountain
Flow, The American Institute of Chemical Engineers, v. 34, pp. 94 (1988).
26
Cross, M., Relation between Viscoelasticity and Shear Thinning Behavior in Liquids, Rheological Acta, v.
18, pp. 609 (1979).
27
Hu, J., Vogel, J.H., Dynamic Modeling and Control of Packing Pressure in Injection Molding, Journal of
Engineering Materials and Technology, v. 116, n. 2 (April 1994).
28
Chiu, W.-Y., Wang, C., Wang, D.-C., Analysis of Filling, Packing and Cooling Stages in Injection Molding of
Disk Cavities, Journal of Applied Polymer Science, v. 43, n. 1 (July 1991).
29
Rezayat, M., Stafford, R.O., A Thermoviscoelastic Model for Residual Stress in Injection Molded
Thermoplastics, Polymer Engineering and Science, v. 31, n. 6 (March 1991).
30
Akbar, S., Altan, M. C., On the Solution of Fiber Orientation in Two-Dimensional Homogeneous Flows,
Polymer Engineering and Science, v. 32, n. 12, (June 1992).
31
Greener, J., Pearson, G., Orientation, Residual Stresses, and Birefringence in Injection Molding, Journal of
Rheology, v. 27, pp. 115 (1983).
32
Tseng, S.C., Osswald, T.A., Predicting Shrinkage and Warpage of Fibre-Reinforced Composite Parts,
Polymer Composites, v. 15, n. 4 (August 1994).
25
33
Matsuoka, T., Takabatake, J.I., Koiwai, A., Integrated Simulation to Predict Warpage of Injection Molded
Parts, Polymer Engineering and Science, v. 31, n. 14 (July 1991).
34
Stempnik, L., J., Automatic Control for Plastic Machinery, Proceedings of the National Conference on Fluid
Power, v. 28, pp. 539 (1972).
35
Ma, C., Y., A Design Approach to A Computer-Controlled Injection-Molding Machine, Polymer Engineering
and Science, v. 14, n. 11, pp. 768 (1974).
36
O’Bryan, J. E., Proportional Valves, Microprocessors, and Closed-Loop Control Keeps Plastics Molders
Competitive, Hydraulic and Pneumatics, v. 42, n. 3, pp. 95 (1989).
37
Mann, J., W., Process Parameter Control: The Key to Optimization, Plastics Engineering, pp. 25 (January,
1974).
38
Okeke, E. J., Cosma, L., Dimensional Repeatability of Gas Assisted Injection Molding and Cavity Pressure
Controlled Closed Loop Injection Molding for Structural Parts, Proceedings from the 1993 Annual
39
Shankar, A., Dynamic Modeling and Control of Injection Molding Machines, Doctoral Dissertation submitted
to Carnegie-Mellon University Department of Mechanical Engineering, 1978.
40
Agrawal, A. R., Pandelidis, I. O., Pecht, M., Injection-Molding Process Control — A Review, Polymer
Engineering and Science, v. 27, p. 18 (1987).
41
Nunn, R. E., Grolman, C. P., Adaptive Process Control for Injection Molding, Journal of Reinforced Plastics
and Composites, v. 9, pp. 2121 (1991).
42
Gao, F., Patterson, W. I., Kamal, M. R., Self-Tuning Cavity Pressure Control of Injection Mold Filling,
Advances in Polymer Technology, v. 13, n. 2, pp. 111 (1994).
43
Smud, S. M., Harper, D. O., Deshpande, P. B., Advanced Process Control for Injection Molding, Polymer
Engineering and Science, v. 31, n. 15 (1991).
44
Chiu, C. P., Wei, J. H., Shih, M. C., Adaptive Model following Control of the Mold Filling Process in an
Injection Molding Machine, Polymer Engineering and Science, v. 31, n. 15 (1991).
45
Seaman, C., M., A Multiple Objective Optimization Approach to Quality Control, Doctoral Dissertation
submitted to Rensselaer Polytechnic Institute Department of Electrical, Computer, and Systems
Engineering, 1991.
46
Rowland, J. C., Ho-Le, K., Process Quality Assurance for Injection Molding of Thermoplastic Polymers
Proceedings from the 1994 Annual Technical Meeting of the Society of Plastics Engineers, v. 52, pp. 342
(1994).
47
De Gaspari, J. D., Melt-Flow Oscillation Improves Part Properties, Plastics Technology, pp. 78 (March,
1994).
26
48
Ibar, J. P., Rheomolding – A New Process to Mold Polymeric Materials, Proceedings from the 1994 Annual
49
Vernyi, B., New Method Improves Plastics Mechanically, Plastics News, pp. 33 (October 18, 1993).
50
Gardner, G., Malloy, R., A Moving Boundary Technique to Strengthen Weld Lines in Injection Molding,
Proceedings from the 1994 Annual Technical Meeting of the Society of Plastics Engineers, v. 52, pp. 2794
(1994).
51
Becker, H., Fischer, G., Muller, U., Push-Pull Injection Moulding of Industrial Products, Kunstoffe German
Plastics, v. 83, n. 3 (1993).
52
Parekh, S., Desai, S., Brizzolara, J., Monitoring the Multi-Live Feed Injection Molding Process with Cavity
Instrumentation, Proceedings from the 1994 Annual Technical Meeting of the Society of Plastics
Engineers, v. 52, pp. 621 (1994).
53
Ogbonna, C., I., Production of Self-Reinforced Polyethylene using the Multi Live-Feed (Injection) Moulding
Technique, Doctoral Dissertation submitted to Brunel University Department of Materials Technology,
1989.
54
Kazmer, D. O., Roe, D. S., Increasing Weld-Line Strength through Dynamic Control of Volumetric
Shrinkage, Proceedings from the 1994 Annual Technical Meeting of the Society of Plastics Engineers, v.
52, pp. 631 (1994).
55
De Gaspari, J. D., Low-Pressure Alternatives for Molding Large Automotive Parts, Plastics Technology, v.
39, n. 10, (September 1, 1993).
56
Rusch, K. C., Gas-Assisted Injection Molding — A New Technology is Commercialized, Plastics
Engineering, v. 35, (July, 1989).
57
Okeke, E. J., Cosma, L., Dimensional Repeatability of Gas Assisted Injection Molding and Cavity Pressure
Controlled Closed Loop Injection Molding for Structural Parts, Proceedings from the 1993 Annual
58
Molding Impossible Possibilities, Injection Molding Magazine, October, 1994.
59
Binary Injection Molding Machine, Article received from Dan Furlano, Applications Engineer, GE Plastics,
Portland, Oregon (May 15, 1994).
60
Trantina, G. G., Ysseldyke, D. A., An Engineering Design System for Thermoplastics, 1989 Society of
Plastics Engineers Annual Technical Conference Proceedings, v. 48, pp. 635-639 (1989).
61
Spencer, R., Dillon, R., The Viscous Flow of Molten Polystyrene, Journal of Colloidal Science, v. 3, pp. 163
(1948).
62
Woodruff, D., Bug Control at Chrysler, Business Week, August 22, 1994, pp. 26.
27
63
Personal communication with Ken Debronsky, Vice President, Mustang Division of Ford Motor Company,
Palo Alto, California (December, 1993).
64
Poslinski, A. J., Aslam, S., Kazmer, D. O., The Effects of Viscosity Variation on Injection Molding, GE
Research & Development Technical Information Series, 92CRD146, 1992.
65
Personal communication with Dan Furlano, Design Engineer, GE Plastics, Portland, Oregon (May 15,
1994).
66
Personal communication with Kosuke Ishii, Associate Professor, Stanford University (October 11, 1994).
28
Chapter 2: Propo
TABLE OF CONTENTS
CHAPTER 2: PROPOSED APPROACH.................................................................................................24

A CASE STUDY ..........................................................................................................................................24
DEGREES OF FREEDOM ..........................................................................................................................27
Degrees of freedom in tool design...........................................................................................................28
Degrees of freedom in product design ....................................................................................................29
DYNAMIC FEED CONTROL.....................................................................................................................30
Concept Overview ...................................................................................................................................32
System Overview......................................................................................................................................34
Potential Benefits ....................................................................................................................................35
Compensation for Complex Material Behavior ................................................................................................... 36
Rejection of Input Variation ................................................................................................................................ 36
Ability to Adapt to Changing Requirements ....................................................................................................... 37
SUMMARY .................................................................................................................................................38
TABLE OF FIGURES
FIGURE 2-1: TYPICAL MOLDED PART ..............................................................................................................25
FIGURE 2-2: REQUIRED TOLERANCES ............................................................................................................25
FIGURE 2-3: DEGREES OF FREEDOM IN TOOL DESIGN .....................................................................................29
FIGURE 2-4: DEGREES OF FREEDOM IN PRODUCT DESIGN ...............................................................................30
FIGURE 2-5: PRODUCT DEVELOPMENT CYCLE ................................................................................................31
FIGURE 2-6: DYNAMIC FEED CONCEPT ..........................................................................................................33
FIGURE 2-7: SYSTEM OVERVIEW OF DYNAMIC FEED CONTROL .....................................................................34
FIGURE 2-8: INCREASED DOF THROUGH PROCESS DESIGN .............................................................................36
TABLE OF TABLES
TABLE 2-1: DIMENSIONS FOR MOLDED PART ..................................................................................................26
TABLE 2-2: TYPICAL CHANGING PRODUCTION REQUIREMENTS ......................................................................38
START OF CHAPTER:24
1
Chapter 2: Proposed Approach
As introduced in Chapter 1, the advancing requirements of technical

applications are tending to consistently exceed the capabilities of the injection
molding process. Recent developments in molding technology have failed to
deliver the process flexibility and consistency required to produce the desired
high quality plastic parts. The approaches which have been reviewed have had
only limited success since the process dynamics are largely determined by the
geometry of the flow channels in the mold and can not be easily changed once
tooled in hard steel.
In this chapter, an alternative approach will be proposed which introduces

additional degrees of freedom by dynamically changing the mold characteristics
during the molding cycle. An overview of this system and its potential benefits
will then be presented before investigating the process dynamics in Chapter 3.
A CASE STUDY
A common need in the injection molding of complex parts is the resolution
of multiple conflicting goals without incurring excessive costs or production
delays. Conflicting goals which frequently arise include, for example, the
reduction of sink near a rib without increasing flash at the edge of the part, or
obtaining satisfactory dimensions without inducing warpage or excessive residual
stresses. Figure Chapter 2: -1 shows a typical molded housing with four bosses
uses to attach a cover to the base.
24
Figure Chapter 2: -1: Typical molded part
In this application, three critical dimensions are specified to provide a

proper fit with a mating part as shown in Figure Chapter 2: -2. In this particular
application, the mold utilized a center-gated sprue to provide a balanced fill with
uniform cavity pressures – a constant linear shrinkage of 0.5% was assumed in
tooling.
L1
L2
L3
Figure Chapter 2: -2: Required Tolerances
Upon production of the part, the absence of the fourth side wall caused
higher cavity pressures in the area around L3 while the rest of the part filled. As
a result, L3 experienced less shrinkage than the rest of the part. Table Chapter
2: -1 lists the required specifications and measured part dimensions. Due to the
non-uniform cavity pressures, the part has shrunk non-uniformly, and does not
25
meet all the required specifications. It should be noted that non-uniform

shrinkage will frequently result in part warpage and misalignment of the boss'
vertical axis, further exaggerating the dimensions listed in Table Chapter 2: -1.
Table Chapter 2: -1: Dimensions for molded part

Dimensions (in) Required Measured "Optimal"
L1 6.00 ± 0.010 5.995 ± 0.002 5.992 ± 0.003
L2 10.00 ± 0.015 9.995 ± 0.003 9.992 ± 0.005
L3 6.00 ± 0.010 6.011 ± 0.002 6.008 ± 0.003
As a first step to increase the shrinkage of L3 and bring the part within
specification, the process engineer might attempt to reduce the cavity pressure
around L3 by increasing the injection speed and reducing the packing pressure.
Since the flow dynamics are determined by the mold geometry, however, a
decrease in the cavity pressure around L3 will also reduce the cavity pressure
throughout the part. While increasing the part shrinkage is beneficial to L3, this
process change will have a negative impact on the other part dimensions. The
shaded column in Table Chapter 2: -1 lists the subsequent part dimensions
achieved in the "optimal" process. Though the nominal part dimensions are
within tolerance, the multiple requirements would force a process yield of only
42% – 100% inspection techniques would be required to ensure acceptable
quality.
If this yield is unacceptable, the molder must resort to stopping production,

pulling the tool, and re-cutting the tool steel to move each boss on L3 0.005"
closer to the center line. Alternatively, the tool engineer might increase the wall
thickness to L1 in an attempt to lead flow away from the area around L3 and
provide more uniform cavity pressures. In either case, production is delayed with
no guarantee of acceptable part dimensions upon return to production. Once this
mold has been tuned for a given material and set of process conditions,
moreover, the introduction of different resins and process conditions could result
in lower yields or unacceptable parts and necessitate further tooling changes.
26
No currently available technology can resolve these conflicts. Simulation

might help in part or tool design, but would be ineffective when material
properties, part geometry, or process conditions differ from those assumed in the
analysis. Closed loop process control might reduce the variation of the part
dimensions, but is not capable of changing multiple dimensions independently.
Similarly, adaptive control methods might increase the process yield by
automatically controlling the process conditions to improve one dimension at the
cost of another and ensure the process remains on the optimal trade-off surface.
Nor do any of these techniques provide the process flexibility required to adapt to
changes in product specifications when, for instance, product assembly requires
L2 to be increased 0.020".
All these technologies fall short of resolving multiple conflicting goals –

they do not possess the required degrees of freedom. In the injection
molding process, there is only one degree of freedom: ram hydraulic pressure,
which controls the magnitude of the cavity pressure distribution and subsequent
part characteristics. In fact, to ensure that all of the part’s dimensions are within
tolerance, three degrees of freedom are required in the manufacturing process –
one degree of freedom to control the cavity pressure and shrinkage in each
critical area of the part.
DEGREES OF FREEDOM
Suh et. al. hypothesized that there exists a small set of global principles,
or axioms, which can be applied in the synthesis of a design which aid the
designer in reaching “correct” decisions.67 The first of their two axioms (not to be
considered a full set) is that independence of functional requirements should be
maintained, i.e. parts of a product should be physically separated if functional
requirements become coupled in the manufacturing process. There should be
an independent, adjustable parameter for each critical feature or specification in
the design. In the molding of a complex part, all of the part specifications
27
become interdependent due to the dynamics of the injection molding process.

Unless these specifications are very loose, it may be impossible to manufacture
acceptable parts.
As applications have become more complex, the number of functional

requirements have increased without relaxation of part specifications. As
previously shown, the degrees of freedom necessary to manufacture these parts
do not exist in the injection molding process. When problems are encountered
late in the development process, this limitation means that the engineer must
utilize degrees of freedom available only through tooling and conceptual design
changes. These late changes are costly and time consuming while providing no
guarantee of problem resolution.
Degrees of freedom in tool design
Increasing the quality of injection molded products is possible by

recognizing the constraints imposed by the molding process and designing the
tool to be easily modified to obtain the desired product attributes. This method
has been recently developed and successfully utilized in design of tight-tolerance
parts which may have several critical specifications such as computer chassis
and automotive components.68 Using this philosophy with the previous example,
a gate would be placed near each critical feature or dimension as shown in
Figure Chapter 2: -3.
L1
L2
L3
28
Figure Chapter 2: -3: Degrees of freedom in tool design
The pressure drop through each gate can then be modified by changing
the runner dimensions to control the local cavity pressures and part properties.
For this example, the diameter of the runner into L3 is reduced to restrict the
amount of flow into that area of the cavity. If L3 or other dimensions were not
satisfactory upon molding, the runners could be redesigned to obtain the desired
flow. One degree of freedom is provided for each gate into the part, but
adjustments are only possible by stopping production and re-cutting the tool
steel. This can be a costly and lengthy process since several adjustments may
be necessary before obtaining acceptable dimensions.
Degrees of freedom in product design
In the design stage of the development process, latitude exists to provide

nearly unbounded design freedom. One design philosophy to eliminate functional
inter-dependencies between multiple part tolerances is to increase the number of
parts. As shown in Figure Chapter 2: -4, the same structure could formed of two
separate parts. The redesign provides three degrees of freedom utilizing two
molded pieces. Each piece can be manufactured separately with a center-gated
sprue to provide independent control of dimension L1 and L3; the components
would later be assembled with a slide for adjustment of dimension L2.
29
L1
L2
L3
Figure Chapter 2: -4: Degrees of freedom in product design
Unfortunately, increasing the part count violates of Nam Suh's other

axiom: minimize information content or “keep it simple, stupid.”69 While providing
the necessary degrees of freedom, this design will increase the total product cost
and sacrifice the product performance, thereby reducing the benefits originally
sought in utilizing the injection molding process.
DYNAMIC FEED CONTROL

Due to the absence of degrees of freedom in the manufacturing process,
industry has utilized the iterative development cycle shown in Figure Chapter 2: -
5 to leverage the degrees of freedom provided in the tooling and conceptual
design stages.
30
Typical
Time For Quality Target
Corrections Part Specification
weeks Product Design
days Tool Design
hours Molding Process
Examine
Quality IMPROVE
OK
Figure Chapter 2: -5: Product development cycle
There are several limitations which this development process inflicts upon the
plastic part designer and manufacturer. First, the molding process does not
possess multiple degrees of freedom, thus it provides only for trade-offs between
multiple requirements. To achieve the desired quality, the degrees of freedom in
the tooling and design stages must be utilized. These changes, unfortunately,
require lengthy production delays and costly tooling changes. Moreover, there is
no guarantee that tooling or design changes will produce the desired effect until
molding is actually performed and samples are tested. This development cycle
becomes excessively prohibitive for very demanding or complex applications,
leading to the failures described in Chapter 1.
31
If extra degrees of freedom could be embedded within the injection

molding process itself, then many quality issues could be resolved at the
production level, without incurring significant production delays due to tooling or
design changes. This need has motivated the development of Dynamic Feed
Control for injection molding.
Concept Overview
The proposed approach, Dynamic Feed Control, is a new method to

increase the number of degrees of freedom available in the injection molding
process. There are many possible concepts but the most generic approach is to
provide a means for instantaneously modifying the diameters in each branch of a
runner system. As shown in Figure Chapter 2: -6, this is accomplished by driving
a tapered valve stem with a hydraulic actuator. Since the resistance to flow is
determined by the gap thickness between the valve stem and the mold wall, axial
displacement of the valve stem can be used to dynamically vary the flow rate and
pressure drop through each valve. If used in a closed loop control system,
Dynamic Feed can provide direct control of either valve positions or cavity
pressures.
Pinjection
P1 P2
32
Figure Chapter 2: -6: Dynamic Feed Concept
Dynamic Feed Control embeds three new characteristics into the molding
process. First, the independent control of each valve allows the pressure and
flow in multiple regions of the cavity to be decoupled. Previously, changes aimed
at improving an area of the part could result in other detrimental effects since the
effect of process changes are transmitted throughout the cavity. With Dynamic
Feed, the flow through each valve can be controlled independently, bringing extra
degrees of freedom to the molding process.
Independent control of multiple valves does not necessarily leverage all

the capabilities of this dynamic system. The benefits of Dynamic Feed can be
multiplied by dynamic re-positioning of the valve within the molding cycle. This
strategy can be used, for instance, to specify one set of valve positions to profile
flow rates in the filling stage followed by a completely different set of valve
positions to profile pack pressures. Using the dynamic capabilities of the new
system, it will be shown that each valve brings several degrees of freedom to the
molding process.
Finally, the dynamic capabilities of this process allow the valves to be

instantaneously controlled in response to feedback from process sensors in the
mold cavity, thus providing closed loop control of the cavity state variables which
directly determine the product quality. Variation in molding machine input
parameters, machine behavior, or material properties can be dynamically
compensated to produce consistent parts. Moreover, the control of cavity
variables directly enables the use of pressure measurements as a process
control technique for automated detection of quality problems. This could
eliminate the need for manual inspection of part quality in many circumstances.
Since the dynamics of the molding machine are completely decoupled from the
cavity, details of molding machine performance now become insignificant!
33
System Overview
The Dynamic Feed Control system utilizes an external process controller

outside the injection molding machine as shown in Figure Chapter 2: -7. In this
ideal representation, the control system receives a quality target and part
specifications for the application being molded such as part dimensions and
allowable tolerances. Using simple empirical and analytical relations, these
specifications are converted to local cavity pressure distributions to be produced
during molding. The control system then compares the desired and observed
cavity pressures, calculating and generating command signals for real time
control.
Quality Target
Part Specification
x, σ, ...
Control
System
Control signals
Injection pressure Molding

Machine
Hydraulic flow & pressurized melt
Valve positions Feed

System
Metered flow
Cavity pressures Mold

Cavity
Plastic parts
Figure Chapter 2: -7: System Overview of Dynamic Feed Control
The command signals generated by the control system are transmitted to

multiple hydraulic servo-valves on the molding machine. These servo-valves
provide pressurized melt to the inlet of the feed system as well as hydraulic flow
to each of the feed system’s actuators shown in Figure Chapter 2: -6. These
actuators in the feed system move the valve stems to meter the pressurized melt
34
to each area of the cavity. As the process continues, process sensors provide
the control system with the feedback from the current valve positions and cavity
pressures.
There are two items not shown in Figure Chapter 2: -7. First, the control
system will attempt to produce the desired cavity pressure profiles. If there is
substantial variation between the desired and observed process behavior, the
control system can generate a signal to an operator or robot to discard the plastic
part. Secondly, there is feedback from the operator to the control system
regarding part quality. Upon examination of molded parts, the operator can
provide a new set of part specifications which will be incorporated into the
process dynamics of subsequent molding cycles.
Potential Benefits
The potential benefits of Dynamic Feed Control will now be discussed.

Figure Chapter 2: -8 illustrates the use of three dynamic valves in the production
of the housing previously shown. It is clear from the figure that each valve can
be controlled to selectively meter the polymer flow and cavity pressure at each
gate. If the dimension L3 was over-sized, then one valve can be closed to reduce
cavity pressure and increase shrinkage in that region, thereby meeting the part
specifications without adversely effecting other part properties.
L1
L2
Time
Adjustable
Valves
L3
35
Figure Chapter 2: -8: Increased DOF through process design
There are similarities between Dynamic Feed Control and the tight
tolerance strategy shown in Figure Chapter 2: -3 – both utilize runner sizing to
provide degrees of freedom to obtain the desired part quality. However, Dynamic
Feed Control provides additional benefits which can resolve the limitations
described in Chapter 1 by: compensating for complex material properties,
eliminating the sensitivity of product quality to input variation, and quickly
adapting to changing production requirements.
Compensation for Complex Material Behavior
Thermoplastics’ complex material properties coupled with the injection

molding process’ complex dynamics impede efforts to predict the flow behavior
and subsequent part properties. This difficulty can be resolved in the product
development by introducing sufficient degrees of freedom in the tooling and
design stages to achieve acceptable part quality. By providing extra degrees of
freedom in the injection molding process, Dynamic Feed enables rapid
compensation for complex material properties not foreseen in design and tooling.
If, for instance, material orientation caused anisotropic behavior in the

housing's base and dimension L2 or part warpage was unacceptable, the process
dynamics could be adjusted to reduce orientation and correct the problem
without delaying production. The freedom exists to adjust the process dynamics
to compensate for unknown material properties or to successfully mold
completely different materials.
Rejection of Input Variation
Even with a well-tuned mold, quality issues may continue to arise in

conventional injection molding due to the sensitivity of the molded part quality to
variation in the molding machine dynamics, material properties, and operator
technique. Even with the most advanced control systems, the capability of the
36
injection molding process has been limited. Closed loop cavity pressure control
in conventional molding has been limited by the intrinsic process dynamics and
system design. The ram's injection cylinder is physically distant from the cavity,
separated by 20 cm to 60 cm of feed system which induces a lengthy time lag.
Moreover, the significant mass of the ram and volume of the hydraulic cylinder
limit the response time. Finally, the system provides control at only a single point
without ensuring acceptable levels of cavity pressure elsewhere in the mold.
Dynamic Feed Control provides the degrees of freedom to dynamically

control the cavity pressure at each valve independently. With minimal mass and
actuator volume, the response dynamics of the valves are significantly faster
than those of the molding machine. Moreover, the lag time is negligible due to
the valves' location directly adjacent to the point being monitored. In a closed
loop process, Dynamic Feed can eliminate the effects of variation in the molding
machine dynamics, material properties, and operator techniques. By using
closed loop cavity pressure control, the Dynamic Feed system can automatically
compensate for these sources of variation and provide consistent cavity
pressures – or provide an output signal when the controller was incapable of
providing a consistent pressure.
Ability to Adapt to Changing Requirements
Prior technologies have attempted to control the molding process to

maintain an optimal trade-off between several objectives. These techniques
have had limited success due to the constraints of the molding machine and
process dynamics which are immutable within the mold steel. With the additional
degrees of freedom provided by Dynamic Feed, each objective may be controlled
independently – moving the injection molding process beyond previous trade-off
surfaces. Since process dynamics can be altered without tooling or design
changes, Dynamic Feed enables rapid process adaptation to meet changing
37
production requirements shown in Table Chapter 2: -2 without adversely affecting

other part properties.
Table Chapter 2: -2: Typical changing production requirements

Change specific part dimensions
Different resin, composition, or color
Move/eliminate knit-lines
Reduce sink/flash in problem areas
Compensate for warpage
SUMMARY
Dynamic Feed Control introduces multiple degrees of freedom to the
injection molding process. These degrees of freedom can be used to
compensate for complex material properties, reject input variation, and adapt to
changing production requirements. With this production stage flexibility, the
product time to market will inevitably be reduced while ensuring acceptable levels
of product quality and process yields. More significantly, however, is that the
improved process flexibility and capability permit greater risk in the conceptual
design stages which may ultimately result in previously unattained product
capabilities.
In a final representation, Dynamic Feed Control provides the product

designer additional freedom while simplifying the tasks of the tooling engineer
and machine operator. To accomplish these feats, Dynamic Feed requires
minor changes to conventional tooling, process instrumentation embedded within
the mold, and a control system to bring it all together. The logic within the control
system is the most critical component – its development will next be discussed.
38
67
Suh, N.P., Wilson, D.R., Bell, A.C., Van Dyck, F., Tice, W.W., Manufacturing Azioms and Their Corollaries,
Presented at the Society of Manufacturing Engineer’s Seventh North American Metalworking Research
Conference, pp. 113 (May, 1979).
68
Burgeson, J., Tight Tolerance Design, Plastics Engineering, v. 47, n. 5, pp. 23, (1991).
69
Suh, N.P., Bell, A.C., Gossard, D.C., On an Axiomatic Approach to Manufacturing and Manufacturing
Systems, Presented at the American Society of Mechanical Engineers Winter Annual Meeting, pp. 124
(December 1977).
39
TABLE OF CONTENTS
CHAPTER 3: PROCESS DYNAMICS.....................................................................................................39

PROCESS DESCRIPTION ..........................................................................................................................39
Plastication Stage ................................................................................................................................................ 40
Filling Stage ........................................................................................................................................................ 40
Packing Stage ...................................................................................................................................................... 41
Ejection Stage...................................................................................................................................................... 42
Assumptions.............................................................................................................................................42
Filling Stage ........................................................................................................................................................ 43
Packing Stage ...................................................................................................................................................... 43
Further Assumptions............................................................................................................................................ 43
Equations for Laminar Flow in a Rectangular Channel.........................................................................45
Machine Dynamics ..................................................................................................................................46
CONTROL MODEL ....................................................................................................................................49
Derivation................................................................................................................................................49
Parameter Estimation..............................................................................................................................52
Rheological Behavior .......................................................................................................................................... 53
Compressibility ................................................................................................................................................... 54
Open Loop Simulation.............................................................................................................................56
Constant Valve Position ...................................................................................................................................... 56
Dynamic Valve Positioning................................................................................................................................. 60
CONTROL SYSTEM DEVELOPMENT ....................................................................................................62
Overview of Control Strategies ...............................................................................................................63
Challenges ........................................................................................................................................................... 63
Potential Strategies for Cavity Pressure Control ................................................................................................. 64
Adaptive Gain Scheduling .......................................................................................................................65
Filling Dynamics ................................................................................................................................................. 67
Packing Dynamics ............................................................................................................................................... 72
Adaptive Scheduling ........................................................................................................................................... 77
SUMMARY .................................................................................................................................................84
TABLE OF FIGURES
FIGURE 3-1: SCHEMATIC OF AN INJECTION MOLDING MACHINE ....................................................................39
FIGURE 3-2: PRESSURE PROFILE THROUGHOUT THE INJECTION MOLDING PROCESS ........................................41
FIGURE 3-3: CHANNEL FLOW .........................................................................................................................44
1
FIGURE 3-4: COORDINATE SYSTEM FOR LAMINAR FLOW ................................................................................45

FIGURE 3-5: FORCE BALANCE FOR SCREW AND VALVE STEM .........................................................................47
FIGURE 3-6: STATE SPACE REPRESENTATION OF PLANT ..................................................................................50
FIGURE 3-7: BLOCK DIAGRAM OF PLANT ........................................................................................................51
FIGURE 3-8: RHEOLOGICAL BEHAVIOR OF UNFILLED, MEDIUM VISCOSITY POLYCARBONATE .........................53
FIGURE 3-9: COMPRESSIBILITY BEHAVIOR OF MEDIUM VISCOSITY POLYCARBONATE .....................................55
FIGURE 3-10: SIMULATION RESULTS FOR CONSTANT VALVE POSITION OF 40 MM ...........................................57
FIGURE 3-11:CAVITY FILLING PRESSURE HISTORY AS A FUNCTION OF VALVE POSITION IN MM ......................59
FIGURE 3-12: CAVITY PACKING PRESSURE AS A FUNCTION OF VALVE POSITION IN MM ..................................60
FIGURE 3-13: SIMULATION RESULTS WITH OPENING AND CLOSING VALVE STEM............................................61
FIGURE 3-14: ADAPTIVE GAIN SCHEDULING ...................................................................................................66
FIGURE 3-15: GAIN SCHEDULING STRUCTURE ................................................................................................67
FIGURE 3-16: RANDOM PROCESS INPUT AND RESULTING FILLING STAGE RESPONSE .......................................68
FIGURE 3-17: COMPARISON BETWEEN LINEAR AND PLANT MODELS TO RANDOM INPUT .................................69
FIGURE 3-18: ROOT LOCUS OF LINEAR MODEL FOR FILLING STAGE ................................................................70
FIGURE 3-19: CLOSED LOOP RESPONSE OF PLANT IN FILLING STAGE ..............................................................71
FIGURE 3-20: RANDOM PROCESS INPUT AND RESULTING PACKING STAGE RESPONSE .....................................72
FIGURE 3-21: COMPARISON BETWEEN LINEAR AND PLANT MODELS TO RANDOM INPUT .................................73
FIGURE 3-22: ROOT LOCUS OF LINEAR MODEL FOR THE PACKING STAGE........................................................74
FIGURE 3-23: ROOT LOCUS OF LINEAR MODEL FOR THE PACKING STAGE WITH LEAD NETWORK .....................75
FIGURE 3-24: CLOSED LOOP RESPONSE OF PLANT IN THE PACKING STAGE......................................................77
FIGURE 3-25: PROCESS CONTROL ALGORITHM ..............................................................................................78
FIGURE 3-26: EFFECT OF TIME SCHEDULE ON PACKING STAGE .....................................................................80
FIGURE 3-27: EFFECT OF INITIAL VALVE POSITION ON PACKING STAGE........................................................81
FIGURE 3-28: CONVERGENCE TO DESIRED PROFILE .......................................................................................83
TABLE OF TABLES
TABLE 3-1: LIST OF MODEL PARAMETERS .....................................................................................................52
TABLE 3-2: PROCEDURE FOR SIMULATION OF NON-LINEAR PROCESS MODEL .................................................56
TABLE 3-3: GAINS FOR FILLING STAGE CONTROL ...........................................................................................71
TABLE 3-4: GAINS FOR PACKING STAGE CONTROL .........................................................................................76
START OF CHAPTER:39
2
Chapter 3: Process Dynamics
The development of an effective control system requires an understanding

of the process physics to be controlled. Unfortunately, the injection molding
process is intrinsically complex and strongly non-linear, rendering many classical
techniques for control system design useless. To aid in control system
development, a simplified process model is developed to adequately capture the
vital dynamics of the molding process – numerous assumptions are utilized to
determine the general behavior of the polymer physics and machine elements
which comprise the process. The performance of several control methods are
then evaluated using simulation to create a fairly robust control strategy to be
validated in Chapter 4.
PROCESS DESCRIPTION
Injection molding is a dynamic, non-linear process consisting of four

sequential stages: plastication, filling, packing, and ejection. Figure Chapter 3: -1
illustrates the principal machine elements involved in the process.
Mold Cavity Feed System Injection Unit
Screw
Injection Cylinder
Figure Chapter 3: -1: Schematic of an Injection Molding Machine
39
Plastication Stage
In the plastication stage, solid thermoplastic pellets are heated and

sheared to form a viscous melt which is accumulated in front of the screw. After
plastication, the thermoplastic melt is composed of an amalgam of molecular
strands of varying length and orientation dependent upon compounding and the
recent temperature and stress fields in the screw. This complex molecular
structure results in extremely complex temperature, shear, and orientation-
dependent rheological and elongational properties which determine the part's
manufacture and end use performance.
Filling Stage
In the filling stage, the screw is driven forward to propel the accumulated
melt through the feed system and into the mold cavity. Flow through the feed
system is primarily fully developed Pouiselle flow,70 generally shear driven with
significant pressure drops occurring in the flow direction. As the melt propagates
in the mold cavity, heat is conducted from the hot polymer melt, through the mold
steel, to the cooler water being circulated through the cooling lines. A solidifies
layer quickly develops along the mold walls, narrowing the effective flow channel
across the cross-section but also reducing the cooling heat transfer away from
the moving melt. The injection pressure required to maintain desired flow rates
increases with the increasing flow length as shown in Figure Chapter 3: -2.
40
Filling Packing Cooling Ejection

100
Injection
75
Pressure (MPa) Cavity

50
25
0
0 3 6 9 12 15 18
Time (sec)
Figure Chapter 3: -2: Pressure profile throughout the injection molding process
As time progresses, further cooling causes the solidified layer to build, locking in
the residual stress and orientation in the ‘skin’ of the part. The progressively
narrower effective wall thickness forces the pressure distribution in the cavity to
become increasingly non-uniform.
Packing Stage
Once the mold cavity is filled, the molding machine maintains a constant
hydraulic pressure in the injection cylinder in an attempt to produce more uniform
pressures in the mold cavity as shown Figure Chapter 3: -2. The material in the
cavity is compressed under this packing pressure as the melt cools and
volumetrically contracts. Even though flow rates in the cavity are small
compared to flow rates in the filling stage, the increasing flow resistance usually
prevents a uniform cavity pressure distribution from developing. The polymer
freezes, starting at the mold walls and propagating towards the center, locking in
the orientation and residual stresses. Eventually, the material near the gate
solidifies, preventing further material from flowing into the mold cavity. The
molding machine controller removes the hydraulic pack pressure and plasticizes
more melt for the next molding cycle.
41
Ejection Stage
Once the gate has frozen shut, the part is cooled within the mold allowing
the material to fully solidify and become rigid. During this time, the rate of cooling
and decay of cavity pressure are determined solely by the process dynamics in
the cavity. When the part is sufficiently rigid, the mold opens and ejector pins
behind the part are actuated to remove the part.
Released from the geometric constraints of the mold cavity, the residual
stresses from the molding process are now free to relax and the part's geometry
is immediately distorted. As the part cools to room temperature, further thermal
contraction and molecular relaxation occurs until the part's final geometry and
performance characteristics are obtained. While most of the shrinkage will occur
within a few minutes of molding, the part may not reach its final dimensions until
days or weeks after molding.71
Assumptions
It is not possible to fully characterize the macroscopic behavior of the melt

over the necessary range of processing and end use conditions, or to fully
simulate the actual process physics in injection molding. Fortunately, this is not
crucial – many simplifying assumptions have historically been involved in
obtaining useful simulations as previously described in Chapter 1. Assumptions
will now be utilized to simplify the material behavior and process dynamics to
obtain a process model useful in control system development.
The goal of the process model is to capture the essential process

dynamics of the filling and packing stages of the molding process: the melt
propagation in the filling stage; the melt compressibility in the packing stage; the
flow resistance due to the valve position; and, the dynamics of cavity pressure
dependent upon the previous effects. The following simplifying assumptions are
utilized to obtain a model of minimal complexity.
42
Filling Stage
The process model is based upon laminar Hele-Shah flow72 for a purely
viscous, Newtonian fluid under isothermal conditions. The Hele-Shah
approximation is applicable to the momentum equation due to the relatively long
characteristic length of the flow direction compared to the thickness direction.
The characteristic Reynolds numbers are very small, ϑ(10-3), so inertial effects
may also be omitted from the momentum equation. The flow regions are
considered fully developed and both the unsteady state and gravitational force
73
effects are ignored due to negligible local acceleration. As a result of these
assumptions, the flow dynamics are governed solely by the shear-stress effects
of the polymer melt so the pressure varies only in the predominate flow direction.
Packing Stage
The model assumes isothermal conditions to simplify the numerical

solution for control purposes, otherwise fairly complex finite element/finite
difference methods would be necessary in simulation. This assumption,
however, will result in an underestimation of the effects of thermal volumetric
shrinkage as the polymer cools and contracts in the packing stage. To provide
for an accurate model of the post-filling stage flow rates, the specific volume of
the material throughout the cavity is modeled with an empirical equation of state
dependent upon cavity pressure and cavity temperature; the cavity temperature
is estimated without performing heat transfer calculations by assuming a
constant empirical cooling rate. Again, the goal is to develop a minimal
complexity control model which captures the vital dynamics of the system – such
assumptions will provide reasonable estimates of aggregate process dynamics.
Further Assumptions
The model also assumes that there is no interaction between multiple

valves, i.e. that the dynamics of one valve do not affect the injection pressure or
cavity pressures in other areas of the mold. The model will also assume that the
43
servo-valves have negligible response time compared to the process dynamics,

that there is negligible leakage and friction in the control actuators, and that the
hydraulic fluid is incompressible. The correctness of these assumptions will be
re-examined later.
D valve
L valve
πD
hvalve L valve
hc avity
wcavity
hvalve
L flow L flow hc avi ty
Figure Chapter 3: -3: Channel flow
Finally, the process model utilizes a series of rectangular channels to

model the flow dynamics through various mold elements, as shown in Figure
Chapter 3: -3. In this model, an average diameter for the valve has been used to
represent the constant width of the channel; the thickness of the channel is
proportional to the valve position. Similarly, the mold cavity is represented by a
rectangular channel whose length increases as the melt propagates in the filling
stage. Both of these assumptions are not strictly correct – as the valve tapers,
there is less cross sectional area so the fluid velocity and pressure drop will
increase proportionately, a significant effect not modeled by a uniform
rectangular channel of constant width.
All these assumptions could be relaxed by developing a more complex

process model. However, the resulting model would induce further non-linear
44
and time-dependent behavior which increases the difficulty of interpretation and

control system development without substantially increasing the value of the
model results. These assumptions lead to a simple process model which
captures the aggregate process dynamics and is useful in control system
development.
Equations for Laminar Flow in a Rectangular Channel
Utilizing the above assumptions, the governing mass and momentum

equations for laminar flow in a rectangular strip are respectively given by:
∂ρ ∂
+ ( ρu ) = 0 (3-1)
∂t ∂x
∂ p ∂ ⎛ ∂ u⎞
= ⎜η ⎟ (3-2)
∂x ∂z ⎝ ∂z ⎠
for the coordinate system described in Figure Chapter 3: -4. As described in the
nomenclature, p is the melt pressure, ρ is the melt density, and η is the fluid
viscosity.
h/2 z
u
0 x
-h/2
Figure Chapter 3: -4: Coordinate system for laminar flow
Assuming no fluid slip at the mold walls, the velocity, u, is 0 at the side walls;
∂u
symmetry at the center-line infers that ∂z
is 0 at z=0.74 Utilizing these
assumptions to integrate the momentum equation provides the common relation
between mass flow rate and pressure drop:
45
h
h3 ∂ p
m& = w ⋅ ∫− h u dz = 12 ⋅ w ⋅ ρ ⋅
2
. (3-3)
2 η ∂x
Integrating the mass equation (3-1) in the z direction, the transient effects
of compressibility can be included:
∂ ∂
w ( ρ h ) + m& = 0 (3-4)
∂t ∂x
Since the density has been modeled as a function of temperature and the
pressure is assumed constant across the cross-section, the volumetric effect of
compressibility can be estimated as:
∂ ⎛ ∂ ρ ∂ p ∂ρ ∂T ⎞
( ρh ) = h ⎜ + ⎟ (3-5)
∂t ⎝∂ p ∂t ∂T ∂t ⎠
∂ρ ∂ρ
The terms ∂p and ∂T are approximated by empirical material properties which
will be estimated later. Since the analysis does not consider heat loss from the
∂T
melt to the mold, ∂t will be assumed constant as an input to the control model.*
Substituting (3-3) and (3-5) into (3-4), a simple unified model is developed for
viscous, ‘compressible’ flow in the filling and packing stages of the injection
molding process:
⎛ ∂ ρ ∂ p ∂ ρ ∂T ⎞ ∂ ⎛ h3 ∂ p⎞
wh ⎜ + ⎟ + ⎜ 12 ⋅ w ⋅ ρ ⋅ ⎟ =0 (3-6)
⎝∂ p ∂t ∂T ∂t ⎠ ∂x⎝ η ∂x⎠
Machine Dynamics
The described laminar flow model must be coupled with the molding
machine dynamics to develop a useful model of the process. Again, the goal is to
*
the rate of cooling may be estimated by heat conduction for an infinite plane of finite thickness with known
initial mold and melt temperatures. Alternatively, a simple knowledge of the process' melt temperature,
ejection temperature, and cooling time will provide a similar result of approximately 10C/sec (geometry and
material dependent).
46
develop a model which captures the most significant relationships between the
machine dynamics and cavity pressure. As previously mentioned, second order
effects (such as servo-valve dynamics, hydraulic leakage, etc.) are not
considered. The principal dynamics being modeled include:
• the acceleration of the ram, dependent upon input hydraulic

pressure and transient injection pressure;
• the effect of ram velocity on flow rate and cavity pressure; and
• the effect of valve stem volumetric displacement on cavity
pressure.
There are two primary forces acting on the screw and valve stems which
determine the motion of these machine elements as shown in : (1) hydraulic
pressure within the hydraulic actuator forcing the desired motion against (2) the
pressure in front of the ram or valve stem in contact with the melt.
Ah yd
Pinj Aram Phyd
Fmelt F hyd
Mram
Figure Chapter 3: -5: Force balance for screw and valve stem
Assuming an ideal hydraulic control system, the force balance on the ram and
valve stem becomes:
Fram = M ram a ram = Phyd Ahyd − Pinj Aram (3-7)
Fvalve = M valve a valve = Phyd Avalve − Pcav A valve (3-8)
where Pinj and Pcav are the injection pressure and cavity pressure, respectively. It
should be noted that if the product of hydraulic pressure and area of the hydraulic
actuator is large compared to product of melt pressure and area of the valve,
47
then the actuation force is much greater than the reaction force due to the melt
pressure. In this case, the ram or valve would accelerate quickly to a velocity
limited by the volumetric capacity of the machine's hydraulic pump. Since the
valve stem area is very small, ~1 cm2, compared to its hydraulic actuator, ~30
cm2, the process model simplifies and the valve stem velocity becomes a
specified input to the system with negligible acceleration dynamics. The validity
of this assumption will be verified in Chapter 4.
Once the acceleration of the ram and specified velocity of the valve stem
are known, the resulting melt flow rates may be determined. The ram
displacement is the principal source of flow but the valve stem displacement
affects the flow dynamics through two mechanisms. First, the location of the
valve stem determines the resistance to flow which subsequently determines the
injection pressure and the ram velocity. Secondly, volumetric displacement of
the valve stem is also a source (or sink) of flow – in the packing phase, the
volume displaced by the valve stem can be considerable compared to the
volumetric changes due to compressibility in the cavity. These effects make
valve stem displacement an effective, though difficult to regulate, method for
direct control of cavity pressures. The conservation of mass then states:
m& ram = m& flow + m& valve , (3-9)
which is the primary linkage between the machine dynamics and flow physics.
48
CONTROL MODEL
The control model is next developed using the governing equations for
laminar flow and machine dynamics previously described. The parameters for
the process model are then evaluated from the rheological and compressibility
properties of a typical thermoplastic material. The open loop behavior of the
control model will then be simulated to provide insight into the process dynamics.
Derivation
Substituting the results from the momentum equation for laminar flow (3-6)
into the conservation of mass (3-9) and adding terms for the mass flow rate due
to displacement of the ram and valve stem provide the process model:
Pinj ⎛ ∂ ρ ∂ p ∂ ρ ∂T ⎞
Aram v ram = +⎜ + ⎟ V cavity − Avalve v valve (3-10)
R total ⎝∂ p ∂t ∂T ∂t ⎠
where Pinj represents the injection pressure in front of the ram, Rtotal is a total
measure of resistance to flow (sum of the feed system and cavity resistance),
and Vcavity is the volume of the material in the cavity. Rtotal is dependent on the
valve stem position as well as the melt front position, making the process model
non-linear.
In (3-10), vvalve is an input while vram is calculated from (3-7) given the
current injection pressure, Pinj, from the plant model and the hydraulic pressure,
Phyd, as an input. The resulting system of equations is:
v ram − R val ⋅ A ram

2
A ram A hyd
= v ram − Pcav + Phyd
dt M ram M ram M ram
(3-11)
Pcav A ram ⋅ β R cav ⋅ β −β T ⋅ β ⋅ w ⋅ h − A val ⋅ β
= v ram − Pcav − L flow − v valve
dt V cav V cav V cav V cav
49
where:
∂ρ
β=
∂p
∂ ρ ∂T
βT = ⋅
∂T ∂t
η
R cav = 12 ⋅ w ⋅ ⋅ (3-12)
h3
η
R val = 12 ⋅ 2 ⋅ π ⋅ rval ⋅
3
x val
Q Pcav
L flow =∫ dt = ∫ dt
w⋅h w ⋅ h ⋅ R cav
The plant model (3-11) can be arranged in a state space representation as

shown in Figure Chapter 3: -6. As previously mentioned, there are two inputs:
Phyd and vvalve which directly influence the ram acceleration and valve position.
The plant has been augmented to include four states: vram, Pcav, xval, and Lflow.
The last two states, valve position and flow length, were made explicit to simplify
computation.
⎡ − R val ⋅ Aram
2
− A ram ⎤ ⎡ A ⎤
⎡ ram ⎤ ⎢
&
v 0 ⎥ ⎡ v ram ⎤ ⎢ hyd
0
⎢ ⎥ ⎢
0 ⎥
M ram M ram ⎥⎢ ⎥ M
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ram ⎥
⎢ & ⎥ ⎢ A ram ⋅ β − R cav ⋅ β −β T ⋅β ⋅ w⋅h ⎥⎢ ⎥ ⎢ ⎥
− Aval ⋅ β ⎥ ⎡ P
⎢ Pcav ⎥ ⎢ 0 ⎢ Pcav ⎥ ⎢ 0 ⎤
⎥ hyd
⎥+⎢ V cav ⎥ ⎢ ⎥
V cav V cav V cav
⎢ ⎥=⎢ ⎥⎢
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ x& val ⎥ ⎢ ⎢ ⎥⎢ ⎥
0 0 0 0 ⎥ ⎢ x val ⎥ ⎢ 0 ⎥ ⎣ v val ⎦
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ 1
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 1 ⎥⎢ ⎥ ⎢ ⎥
⎢⎣ L& flow ⎥⎦ ⎢ ⎢ ⎥
0 0 0 ⎥ ⎢⎣ flow ⎥⎦ ⎣ 0
L 0 ⎦
⎣ w ⋅ h ⋅ R cav ⎦
Figure Chapter 3: -6: State space representation of plant
The plant model may also be represented by a block diagram as shown in

Figure Chapter 3: -7. As shown more clearly in the diagram, the ram
50
acceleration, aram, is determined solely by the input hydraulic pressure and the
current injection pressure. The flow rate into the cavity is determined primarily by
the ram velocity. However, valve stem displacement can either increase or
decrease the melt flow rate into the cavity depending upon the direction of
movement; solidification of the melt can only reduce the melt flow. The sum of
these three flows advances to the melt front.
Vmelt Lflow
1/wh
−βT 1/s
-1/Rcavity
Phyd Fhyd Fram aram vram Q Pcavity

Ahyd + 1/Mram 1/s Aram + β melt/Vmelt 1/s
Fmelt
Pinjection ∆Pvalve
-Aram + Rvalve
xvalve
1/s
v valve
Avalve
Figure Chapter 3: -7: Block diagram of plant
There is a significant change in the system dynamics when the melt flow
reaches the end of the cavity. At this point, an impermeable boundary condition
is applied at the melt front, i.e. R cavity → ∞ , and the melt front is not permitted to
advance. The process dynamics in the packing stage are quite different than
those of the filling stage – the flow due to the ram displacement becomes
negligible, increasing the significance of flow due to compressibility and valve
displacement.
The system described in Figure Chapter 3: -7 is clearly not linear. In fact,

the valve position, xvalve, only affects the system through the non-linear term Rval
(3-12). For a small incremental time step compared to the process dynamics, the
51
change in valve position and flow length is small, i.e. ∆xvalve/xvalve<<1 and
∆Lflow/Lflow<<1. This allows calculation of the flow resistance in the valve and
cavity at that moment. The matrix elements are then updated and the resulting
flow rates and melt pressures determined. Numerous iterations are performed
until the simulation is finished.
Parameter Estimation
The model has a total of nine parameters as listed in Table Chapter 3: -1:
three parameters are related to material properties and six constants are related
to physical dimensions of the molding machine and mold cavity. While the
physical parameters will vary with the molding machine and mold geometry, they
can be easily measured with confidence. The values listed in Table Chapter 3: -
1 are for the material, molding machine, and mold geometry used in the
experimental portion of this research.
Table Chapter 3: -1: List of Model Parameters

Constan Description Estimated Value
t
βmelt Bulk modulus of melt -4.2e-4 cm3/g/MPa
βT Coefficient of thermal contraction 4.1e-4 cm3/g/C
η Shear viscosity of thermoplastic melt 400 Pa Sec
Ahyd Cross-sectional area of hydraulic 100 cm2
cylinder
Aram Cross-sectional area of ram 10 cm2
Avalve Cross-sectional area of valve stem 1 cm2
h Wall thickness of mold cavity 0.2 cm
Mram Mass of ram 60 kg
w Width of mold cavity 5 cm
This research utilizes the material properties of an unfilled, medium

viscosity polycarbonate (LEXAN™ 121 resin, GE Plastics). Polycarbonate is an
amorphous thermoplastic with well understood rheological and compressibility
behavior. The material properties listed in Table Chapter 3: -1, however, are
gross estimates of this material's actual behavior in molding – experimental
52
validation described in Chapter 4 demonstrates how significantly these rough

estimates have affected the accuracy of the control model.
Rheological Behavior
The material was extruded through a laboratory capillary shear rheometer

(Materials Characterization Lab, GE Plastics, Pittsfield, MA) with a
length:diameter ratio of 25:1 across a typical range of processing conditions. The
resulting temperature and shear rate dependence of this material's rheological
behavior is shown below in Figure Chapter 3: -8. The viscosity exhibits
Newtonian behavior at lower shear rates and transitions to an asymptotic power
law behavior at higher shear rates. Additionally, the viscosity is extremely
temperature dependent – in fact, both analytical and empirical studies have
shown Arrhenius (exponential) functions describe the temperature dependence
quite well.75,76
3
10
280C Range of
process
conditions in Newtonian
molding viscosity
assumed in
analysis
Viscosity (PaSec)
300C
320C
Figure Chapter 3: -8: Rheological behavior of unfilled, medium viscosity polycarbonate
53
The dark line in Figure Chapter 3: -8 represents the constant Newtonian

viscosity assumed in the analysis. It appears to be an invalid assumption since
the viscosity is strongly shear rate and temperature dependent. Fortunately,
shear rates encountered in the mold filling are generally less than 500 sec-1 –
well within the Newtonian range of the material's behavior. The temperature
dependence, however, is a less trivial assumption. Fortunately, the thermal
convection in the filling stage tends to make the melt during the filling stages
almost isothermal for reasonable fill times. However, intuition must be applied
when simulating slow flow rates which may induce excessive cooling, significant
variations in melt viscosity, and deviation from the simulation predictions.
Compressibility
The compressibility of the material was tested at GE Plastics using a PvT

apparatus designed by the University of Colorado at Boulder.77 The specific
volume for the amorphous polycarbonate is plotted against temperature for a
variety of processing pressures in Figure Chapter 3: -9. The material changes
from a solid to a liquid between 150 and 200°C (the transition between melt and
solid is pressure dependent and related to the molecular free volume). Both the
thermal expansion and compressibility of the polymeric melt are substantial.
54
0.98
0 MPa
0.96
Range of process
0.94 conditions in
50 MPa
Specific Volume (g/cc)

molding
0.92
0.9 100 MPa
0.88 150 MPa
200 MPa
0.86
0.84
0.82
0.8 Bi-linear model

0 50 100 150 200 250 300 350 used in
Temperature (C) analysis
Figure Chapter 3: -9: Compressibility behavior of medium viscosity polycarbonate
Very sophisticated models have been developed to accurately model the

compressibility behavior of both amorphous and crystalline thermoplastic
materials.78,79 To simplify numerical methods, a bi-linear model was fit to the
measured data in the liquid-phase above the melting point:
ρ ( P,T ) = ρ 0 + β ⋅ P + β T ⋅ T (3-13)
The resulting fit is indicated by the parallel lines in Figure Chapter 3: -9. The
constants β and βT were estimated as -4.9e-4 cc/g/MPa and 4.1e-4 cc/g/ C ,
respectively, with a coefficient of correlation of 0.9, indicating a fair fit to the data.
The plant model considers the effect of compressibility and thermal

volumetric shrinkage during the filling and packing stages of the injection molding
process utilizing this bi-linear model. In these stages, the thermoplastic melt is
primarily liquid – the packing stage ends when flow has solidified. By this time,
both the rheological and compressibility assumptions are clearly invalid. This
simple model, however, provides an acceptable model for estimating the
55
aggregate dynamics of the process in the filling and packing stages. More
complex material and process models could be utilized, but would further
complicate the numerics and non-linearities seeking to be understood.
Open Loop Simulation
The simulation of the plant normally begins with the cavity completely
empty, the valve in a fully closed position, and the screw at rest. The only inputs
are hydraulic pressure to the injection cylinder and valve stem velocity, both of
which may be profiled throughout the filling and packing stages. At the start of
the simulation, hydraulic pressure is supplied to the ram's hydraulic actuator and
the valve begins to open. At each time step, the ram velocity, valve position, flow
resistance, and cavity pressure are calculated for the process dynamics from the
previous time step. The resulting data are then input to the plant model and the
calculations performed for the next time step. This sequence is repeated through
the filling and packing stages, until finished. Table Chapter 3: -2 summarizes this
procedure with the variables defined in the nomenclature.
Table Chapter 3: -2: Procedure for simulation of non-linear process model

Initialize variables at k=0
At t=k,
A(t) = A(k)
xk+1 = A(k)·xk + B·uk
k=k+1
Repeat until finished
Constant Valve Position
The dynamics of the plant may be more easily understood by first

neglecting the effects of valve displacement. In this simulation, the cavity is
initially empty, the valve is at a specified initial condition, and the ram is at rest.
At the start of the process, hydraulic pressure is supplied to the ram's hydraulic
actuator and the ram starts to move; the valve stem is maintained at its initial
open position. In each successive step, the flow length and corresponding flow
56
resistance in the cavity is calculated and the results are inserted into the plant
dynamics as previously described.
Cavity Pressure (MPa)

80
60 Filling Stage
40
20 Packing Stage
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Ram Velocity (mm/sec)
20
15
10
5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (sec)
Figure Chapter 3: -10: Simulation results for constant valve position of 40 mm
Figure Chapter 3: -10 graphs the cavity pressure and ram velocity
predictions for a constant valve position of 40 mm and a supplied hydraulic
pressure of 8 MPa (1200 psi). As shown in the figure, the ram starts at rest and
accelerates to a maximum velocity of 20 mm/sec, limited solely by the pressure
drop through the valve. As the melt advances in the cavity for times less than 1.4
sec, both the cavity pressure and flow resistance increase. Due to the increasing
injection pressure, the ram velocity decays slightly which slows the rate of rise of
the cavity pressure. If ram velocity were desired to be held constant, the
hydraulic pressure to the injection cylinder could be increased slightly using a
feedback loop to automatically adjust for the increasing injection pressure.
At 1.4 seconds, the cavity is completely filled and flow into the cavity is
driven solely by compressibility and volumetric shrinkage. This forces the flow
57
rate through the valve to quickly decay. Since the valve is maintained in a semi-
open position, the pressure drop through the feed system decreases and the
cavity pressure quickly reaches the correct asymptotic pressure of 80 MPa. (In
the processing of thermoplastics, most molding machines hydraulic cylinders are
approximately ten times the area of the ram.) Lower cavity packing pressures
may be induced by either reducing the supplied hydraulic pressure or more fully
closing the valve in the packing stage.
Valve position has a dramatic effect upon both the flow rates and injection
pressures in the filling stage. Figure Chapter 3: -11 graphs cavity pressure
traces for several different valve positions. According to the results, the slope of
the injection pressure can be specified anywhere in the range of 0 MPa/sec to
100 MPa/sec by changing the valve position. At a fully open valve positions
(greater than 50 mm), there is negligible pressure drop through the valve and the
slope of cavity pressure is determined by the hydraulic capabilities of the molding
machine. At nearly closed valve positions, the pressure drop through the valve is
so great that negligible flow occurs – a valve position of 25 mm would likely result
in a short shot due to slow flow rates and melt solidification in the filling stage.
58
80
70
60

50 60
40 50
30 40
35
20
30
10 25
0
0 1 2 3 4 5
Time (sec)
Figure Chapter 3: -11:Cavity filling pressure history as a function of valve position in mm
The effect of valve position on flow rates in the filling stage may be roughly
deduced from the fill times in Figure Chapter 3: -11 (the time at which the cavity
pressure begins a steep ascent toward the asymptotic packing pressure). The
3 3
flow rates may be controlled from 40 cm /sec to 5 cm /sec, though very low flow
rates entail greater heat loss and poor model prediction. Interestingly, the
dynamics of switchover from the filling to the packing stage become slower as
the valve closes and restricts flow, clearly due to the longer time required to
deliver the volume of melt necessary to obtain a full compression of the melt
cavity.
The effect of valve position on cavity packing pressures is shown in Figure

Chapter 3: -12. At more open positions, the pressure drop through the valve in
the packing stage is negligible due to the reduced flow rates – the hydraulic
pressure of the injection cylinder drives the pressure to 80 MPa. However, the
valve position significantly effects the level of packing pressure at more closed
positions, between 10 mm and 25 mm. In these partially closed positions, the
59
flow resistance through the valve is so great that even small flow rates can
induce significant pressure drops and provide reasonable control of packing
pressures.
80
70
Cavity Pack Pressure (MPa)
60
50
40
30
20
10
0
10 15 20 25 30 35 40
Valve Position (mm)
Figure Chapter 3: -12: Cavity packing pressure as a function of valve position in mm
Dynamic Valve Positioning
Simulation has shown that the valve position determines the flow rates,
filling pressures, and packing pressures in the molding process. Unfortunately, it
is unlikely that the desired filling and packing dynamics will be achieved through
the use of a single constant valve position – excellent control of the filling
dynamics requires fairly open valve positions while control of the packing
dynamics requires nearly closed valve positions. As such, dynamic re-
positioning of the valve is required to provide more degrees of freedom and
adequate control of both the filling and packing phases of the molding process.
At the start of this simulation, the cavity is initially empty, the valve is fully
closed, and the screw is at rest. The process begins when hydraulic pressure is
supplied to the ram's hydraulic actuator while the valve begins its prescribed
60
motion. In each successive step, the valve stem position and flow resistance
through the valve is updated along with the flow length and corresponding flow
resistance in the cavity, after which the process dynamics are updated. Figure
Chapter 3: -13 graphs the predicted cavity pressures for dynamic opening of the
valve to a position of 40 mm followed by fully closing the valve.
150
100 Filling
50
Packing
0
0 1 2 3 4 5 6
Valve shuts
causing
Valve Position (mm)
40 pressure spike
30
20
10
0
0 1 2 3 4 5 6
Time (sec)
Figure Chapter 3: -13: Simulation results with opening and closing valve stem
As shown in Figure Chapter 3: -13, the general filling and packing

behavior between 0.5 and 3.5 seconds is identical to that of Figure Chapter 3: -
10 – it should be as the valve's open positions are identical. However, the
dynamic positioning of the valve causes a delay in the start of the cavity filling.
The initially closed position fully restricted flow and acceleration of the ram; as
the valve opened, the flow restriction through the valve remains enormous,
forcing the ram to accelerate slowly. Moreover, the valve's backwards
displacement creates an empty vacuum – this volume has to be filled before the
melt can enter the cavity. The 0.5 second time delay does not limit the
capabilities of the process but needs to be considered in the control system
development.
61
A more significant effect of dynamic valve movement is the pressure spike

which occurs when quickly closing the valve with a filled cavity. In this instance,
the flow rates due to compressibility and volumetric shrinkage can be small
compared to the amount of volume displaced by the moving valve stem. As
shown in Figure Chapter 3: -13, closing the valve stem at 4.0 sec forces material
into the cavity while simultaneously increasing the resistance to flow of material
out of the cavity. This considerably increases the cavity pressure – in this
instance to 150 MPa from 75 MPa, nearly double. The magnitude of this effect
can be reduced by 1) reducing the closing velocity of the valve, 2) reducing the
distance the valve must close, or 3) closing the valve after the injection pressure
has been removed. This effect must also be considered in subsequent control
system development.
Once the valve is fully closed, the cavity pressure begins to decay due to
volumetric shrinkage during cooling. This simulation predicts a decay of roughly
10 MPa/sec (1400 psi/sec) which will be compared to the observed behavior of
the molding process in Chapter 4. This result also implies a strategy in control
system development. Since cavity packing pressures can decay at a maximum
rate of 10 MPa/sec, the control system should never overshoot the desired cavity
pressure.
CONTROL SYSTEM DEVELOPMENT
As previously described in Chapter 1, current methodologies for

development of injection molded parts require iterative tool changes to obtain the
desired filling patterns and pressure distribution – a lengthy and costly tooling
development process. As shown in the previous section with Figure Chapter 3: -
12, hard-tooled flow paths also implies a lack of control in the packing stage.
The flexibility of the molding process would be greatly increased by being able to
change the filling patterns and pressure dynamics without re-tooling. The
62
consistency of the molding process would be greatly increased if the process

dynamics were made independent of material, machine, and operator variation.
The goal of the control system development is to control the delivery of the
melt through each of the multiple valves thereby varying cavity pressure profiles
as required in different areas of the mold – independent of material, machine,
and operator variation. If successful, Dynamic Feed Control would provide the
means to specify the process dynamics in the cavity around each gate without
re-tooling.
Overview of Control Strategies
Challenges
The physics of the molding process provides formidable challenges to

control system development. Even with all the assumptions regarding material
properties and process physics, the plant model remains highly non-linear. The
primary input to the system, valve stem velocity, only indirectly effects the plant
behavior. Moreover, the dynamics of the process (Figure Chapter 3: -6) are time
dependent – not only do they change with valve position but with time as the
polymer flows through the cavity. Indeed, the filling and packing stages of the
process exhibit very different process dynamics, each stage must be addressed
in control system development. Some necessary performance characteristics of
the control system are:
• tracking of time-varying cavity pressure profile

• rejection of sensor noise and actuator hysteresis
• compensation for machine, material, operator variation
• guaranteed stability
• robustness for unknown inputs and unmodeled plant dynamics
• readily transferable to other molding machines, feed system
designs, and mold geometries
63
Potential Strategies for Cavity Pressure Control
Dynamic Feed Control aims to provide the degrees of freedom and

process control needed to vastly improve the capabilities of the molding process.
However, this goal requires that some expert knowledge be integrated within the
control system. There are several potential strategies for accomplishing this
goal.
As shown in Figure Chapter 3: -11, valve position determines the cavity

pressure response in the filling and packing stages. An effective method for
controlling cavity pressure, then, would be to specify a varying valve position
profile across the filling and packing stages. For instance, a valve position of 35
mm could be utilized in the filling stage to achieve a cavity pressure profile
around 10 MPa/sec (Figure Chapter 3: -11) followed by a change to 18 mm to
obtain a packing pressure of 30 MPa (Figure Chapter 3: -12). Without an
effective control algorithm, this would involve direct operator intervention and
expert knowledge of the process physics to achieve and maintain the desired
process performance. Moreover, all effects of machine, material, and operator
variations would be transmitted through the feed system to the cavity. Since
cavity pressure is not directly controlled, it is evident that the consistency of the
injection molding process would not be improved.
Proportional, integral, derivative (PID) controllers are ubiquitous in the

process control industry. In fact, PID controllers are currently used to control ram
velocity, barrel temperature, coolant temperature – nearly every process variable
within the injection molding process. Unfortunately, PID control is unsuitable for
cavity pressure control due to the rapidly varying, non-linear behavior of the
cavity pressure dynamics. As implied in Figure Chapter 3: -11, a constant set of
gains might prove acceptable at the start of fill only to become unstable
throughout the packing stage. A stable response could be ensured with a very
64
conservative set of gains, but then the system performance would be very slow
resulting in unacceptable response and defective parts (short shots).
The alternative to controllers with fixed feedback gains is to utilize

adaptive control techniques. Many adaptive techniques have been developed for
control system design – reviews are available in the literature.80,81,82,83 The
distinguishing feature of adaptive control systems is that the control behavior
adapts to changing plant dynamics, i.e. the control system redesigns itself as the
plant changes. Adaptive control schemes were first developed in the 1950's to
improve aircraft flight dynamics at varying altitudes. Model reference adaptive
control (MRAC) is currently the most prevalent research area within the adaptive
control field and may be positioned in various forms such as H∞, optimal control,
self tuning controllers,84 variable structure systems.85
Unfortunately, all accessible adaptive control techniques require the

process dynamics to be mostly linear with a well-defined structure. Less
significant non-linear dynamics are ignored and later compensated by a robust
controller design. Control design of nonlinear systems with invertible
nonlinearities are currently an active field of research.86 Since cavity pressures
within the injection molding process are nonlinear and these dynamics can not be
linearized, most MRAC and optimal control techniques do not apply.
The control design problem is not intractable, however. Adaptive gain

scheduling is the oldest and simplest form of adaptive control – it is also the most
widely used of all adaptive algorithms by far. 87 Moreover, the plant model lends
itself well to this form of control.
Adaptive Gain Scheduling
Gain scheduling permits the use of several different control algorithms

based on the current operating conditions as shown in Figure Chapter 3: -14.
65
The main burden of developing a system which successfully utilizes gain

scheduling is identifying the proper control design to be used in each stage and
effecting a smooth transfer between stages during operation. Within a stage, the
controller properties are fixed – this enables the control system to be designed
off-line and then tuned on-line for each stage before production use.
Controller Auxiliary measurement

parameter
selection
yr ye u Linear y
+ Controller
plant
Figure Chapter 3: -14: Adaptive gain scheduling88
As previously described, the plant model exhibits very different process

dynamics in the filling and packing stages which demand different types of valve
actuation. In the filling stage, excellent control of flow rates are provided with the
valve in primarily open positions. The flow through the valve is much greater
than the flow displaced by the valve movement, so excellent dynamic response
should be provided by rapid valve actuation. In the packing stage, however, the
flow rates are small compared to the flow displaced by valve movement. Control
of packing pressures necessitates the valve to be in a semi-closed position at the
start of the packing stage and respond slowly to changes in cavity pressure.
With adaptive gain scheduling, the control system can utilize completely different
control algorithms in the filling and packing stages, as shown in Figure Chapter 3:
-15.
66
Filling Stage Packing Stage

-Aram -Aram
Vmelt Lflow Vmelt Lflow
1/ wh 1/ wh
βT 1/s βT 1/s
-1/Rcavity -1/Rcavity
Start Phyd
Ahyd
Fhyd
Fmelt
+
Fram
1/ Mram
aram
1/s
vram
Aram +
Q
βmelt/Vmelt 1/s
Pcavity
Switch? Phyd
Ahyd
Fhyd
Fmelt
+
Fram
1/ Mram
aram
1/s
vram
Aram +
Q
βmelt/Vmelt 1/s
Pcavity
End?
∆Pvalve ∆Pvalve
-Aram Rvalve -Aram Rvavl e
xvalve xvalve
1/s 1/s
No No
vvalve vvalve
Avavl e Avavl e
u(t)=10P(t)-P'(t)-X(t) u(t)=P(t)-P'(t)-10X(t)
Figure Chapter 3: -15: Gain scheduling structure
It will next be shown that PD control with moderate gains provides

excellent response and stability in the filling stage. In the packing stage, PID
control with much lower gains can be used to specify the cavity packing
pressures. The most difficult task involves estimation of the transition from the
filling to the packing stage. Each of these elements in the control system design
will now be presented.
Filling Dynamics
To achieve some estimation of the system stability and performance, each

stage of the molding process was linearized about a very narrow operating point.
The filling stage was linearized about a valve position of 40 mm and a ram
velocity of 20 mm/sec with the cavity 40% full. For small changes in valve stem
position, the plant dynamics should not vary significantly from this point. A
stream of random input (valve velocities oscillating with a magnitude of ± 0.02
mm/sec) were then generated and input to the model; Figure Chapter 3: -16
graphs the random input and resulting process output.
67
OUTPUT #1
60
40
20
0
0 50 100 150 200 250 300
INPUT #1
0.02
0.01
0
-0.01
-0.02
0 50 100 150 200 250 300
Figure Chapter 3: -16: Random process input and resulting filling stage response
Model identification was then performed using the Identification Toolbox in

Matlab.89 Linear models from ϑ(1) to ϑ(10) were fit to Figure Chapter 3: -16.
The final model was selected which produced the closest approximation to the
non-linear plant. In all cases, a model with four poles proved closest to the plant
response. The selected model is:
108 s 2 + 5290 s + 45200

Y ( s) = (3-14)
s 4 + 119 s 3 + 3820 s 2 + 53000 s + 1160
The same stream of random input was then provided to the linear model – Figure
Chapter 3: -17 compares the results from the linear model (dotted line) to that of
the plant (solid line). The correlation of the responses are excellent. It is
interesting to note that the linear model does not predict the slight spikes due to
valve movement in the plant model. Since this is not a significant effect in the
filling stage, the linearized model is a good approximation for that region of the
filling phase.
68
50
45
40

35
30
25
20
15
10
0 1 2 3 4 5 6
Time (sec)
Figure Chapter 3: -17: Comparison between linear and plant models to random input
The closed loop response of this linear model is shown by the root locus
plot of Figure Chapter 3: -18. In this root locus, the poles and zeros are indicated
by x and o, respectively. There is a second zero far off on the positive real axis.
This plot indicates that the closed loop response will be stable for very small
gains. (The closed loop response is stable when the locus lies entirely on the
negative real axis.90) As the gain increases, the system will respond more
quickly – some oscillation will occur before achieving a steady state value. For
larger gains, the system response becomes more oscillatory, the dynamics
become slower, and the system goes unstable.
The objective in developing the linear model is to understand the behavior

of the Dynamic Feed process, develop an estimate of its performance, and
foresee difficulties in control of the real process. Figure Chapter 3: -18 implies
that cavity pressures in the filling stage may be controlled with a constant
feedback gain but that the system can easily become unstable. This information
must be further discounted since the linearized model is only an estimate of the
69
non-linear plant about one region within the filling stage as the process dynamics
may be significantly different as the cavity continues to fill or the valve moves to
different positions.
200
150
100
50
Imag Axis
-50
-100
-150
-200
-100 -50 0 50 100
Real Axis
Figure Chapter 3: -18: Root locus of linear model for filling stage
As shown in Figure Chapter 3: -18, the closed loop response is marginally

stable for proportional control. It is well known that derivative control will
increase the response time and stability of the control system at the cost of
reduced steady state accuracy.90 Since stability is the primary concern, the
control algorithm developed for the filling stage utilizes proportional, derivative
control of pressure coupled with weak proportional control for valve position:
U ( t ) = K p ( Pdes − Pout ) + K d ( Pdes ′ ) + K x ( X des − X out ) ,

′ − Pout (3-15)
where the subscripts des and out represent the desired input and process output
signals. This will tend to keep the valve in a mostly open position and adjust the
valve position to obtain the desired cavity pressure. Because of the uncertainties
70
in the system dynamics, a conservative set of gains must be used in the control
system. The gains chosen are listed in Table Chapter 3: -3.
Table Chapter 3: -3: Gains for filling stage control

Kp Kd Kx
0.0175 mm/MPa 0.0005 mm/MPa/sec 0.01 mm/mm
This control algorithm was implemented for the non-linear plant model to obtain
the results shown in Figure Chapter 3: -19. The three straight lines correspond
to the desired input cavity pressures with slopes of 6, 18, and 54 MPa/sec – a
typical range of cavity pressures utilized in the injection molding process. The
closed loop response (shown by the dotted lines) is excellent. For low slopes of
cavity pressure, the closed loop control delivers the desired response – the
output of the plant just leads the 6 MPa/sec input. At 54 MPa/sec and higher
slopes, the plant slightly lags the input but maintains an acceptable response; the
response is more oscillatory since the slopes are higher, resulting in an
effectively greater Kd.
60
54 18 MPa/sec
40
20 6 MPa/sec
0
0 0.5 1 1.5 2 2.5 3 3.5 4
60
tion (mm)
54 MPa/sec 18 MPa/sec
40
6 MPa/sec
Figure Chapter 3: -19: Closed loop response of plant in filling stage
71
Packing Dynamics
The control system development for the packing stage will follow the same
methodology used in the filling stage. To begin, the packing stage was linearized
about an initial valve position of 15 mm and a slow ram velocity of 2 mm/sec –
the cavity was completely full and pressurized to 80 MPa (the steady state
packing pressure). The packing dynamics are very different than those in the
filling stage; small changes in valve stem will significantly effect the cavity
pressure. A stream of random input (valve velocities oscillating with a small
magnitude of ± 0.01 mm/sec) were then generated and input to the non-linear
plant model. The resulting output is shown in Figure Chapter 3: -20.
OUTPUT #1
4
-2
0 50 100 150 200
INPUT #1
0.01
0.005
0
-0.005
-0.01
0 50 100 150 200
Figure Chapter 3: -20: Random process input and resulting packing stage response
Model identification was again performed using the Identification Toolbox

in Matlab as previously described.89 In all cases, a model with four poles and
three zeros reproduced the plant response. The selected model for the packing
stage is:
72
s 3 + 15.6 s 2 + 3250 s − 9410000

Y ( s) = 4 (3-16)
s + 1890 s 3 + 50700 s 2 + 4570000 s + 4830000
The same stream of random input was then provided to the linear model – Figure
Chapter 3: -21 compares the results from the linear model (dotted line) to that of
the non-linear plant. The correlation of the responses is fairly good. The linear
model predicts the general shape of the plant response due to valve movement
but does not accurately reflect the volumetric shrinkage of the plant model
(compare the slopes between 1 and 3 seconds).
84
83.5
83
82.5
82
81.5
81
80.5
80
79.5
79
0 1 2 3 4 5 6 7 8
Time (sec)
Figure Chapter 3: -21: Comparison between linear and plant models to random input
The closed loop response of this linear model is shown by the root locus
plot of Figure Chapter 3: -22. Even without looking at the plot, equation 16
indicates that the linear model exhibits an unstable zero – the root locus confirms
this (the closed loop response of the system is unstable when any part of the
locus lies on the right half plane). Not shown in the figure is a fourth pole far off
on the negative real axis which does not effect the system dynamics.
73
100
80
60
40
20
Imag Axis
0
-20
-40
-60
-80
-100
-100 -50 0 50 100
Real Axis
Figure Chapter 3: -22: Root locus of linear model for the packing stage
This root locus plot of Figure Chapter 3: -22 indicates that the closed loop
response is inherently unstable, even for very small proportional gains. This
instability is due to the effect of valve stem movement on cavity pressure in the
packing phase as can be demonstrated by an example. If the cavity pressure is
above the desired packing pressure, for instance, the control system will close
the valve to decrease flow to the cavity. However, the movement of the valve
stem will force additional material into the cavity, increasing the cavity pressure
and driving the proportional control system to further close the valve. Eventually,
the valve will be fully closed and the cavity pressure will decay below the desired
level. The control system will then open the valve. In doing so, a vacuum will be
created and flow will be pulled from the cavity. The cavity pressure will decrease
rapidly, causing the control system to further open the valve. The unstable
behavior will continue throughout the packing stage.
74
As the proportional gain increases, the system will become more unstable
(the valve movement will increase for small errors in cavity pressures). A lead
compensator may be utilized to attempt to stabilize the system:
s +1
G ( s) = . (3-17)
s + 30
Similar to derivative control, a lead network will raise bandwidth, lower the rise
time, and decrease the transient overshoot. The root locus for this the linearized
model with the described lead network is shown in Figure Chapter 3: -23.
100
80
60
40
20
Imag Axis
0
-20
-40
-60
-80
-100
-100 -50 0 50 100
Real Axis
Figure Chapter 3: -23: Root locus of linear model for the packing stage with lead network
Figure Chapter 3: -23 infers that the control system is marginally stable.
To further increase the system stability, the proportional and derivative pressure
control will be coupled with proportional control of valve position, as previously
described in equation (3-15). The gains utilized for the packing stage are listed
in Table Chapter 3: -4. Even though Kd and Kx are identical to those listed in
75
Table Chapter 3: -3, their relative impact on system performance is greatly

increased since Kp has been reduced by an order of magnitude.
Table Chapter 3: -4: Gains for packing stage control

Kp Kd Kx
0.0015 mm/MPa 0.0005 mm/MPa/sec 0.01 mm/mm
Since proportional control of valve position plays a critical role in the packing
stage, a good estimate of the valve position is critical to obtaining an acceptable
response. An adaptive mechanism is utilized to find the correct valve position as
will be described in detail in the next section.
This control algorithm was implemented for the non-linear plant model to
obtain the results shown in Figure Chapter 3: -24. There are three levels of
desired packing pressure: 20, 40, and 60 MPa which correspond to a broad
range of packing pressures common in industry. In each case, the cavity
pressure was set to 10MPa below the desired pressure with a valve position of
15 mm. The closed loop response is fairly good as indicated by the dotted
curves.
80
60
60
40
40
20
20
0
0 1 2 3 4 5 6
25
m)
60
40
20
MPa
76
Figure Chapter 3: -24: Closed loop response of plant in the packing stage
The rise time of cavity pressure (time at which the value reaches 95% of
the desired signal) varies between 1 and 3 sec, increasing with the desired level
of packing pressure. This occurs as the valve opens to increase the flow to the
cavity and, in doing so, creates a vacuum which decreases the cavity pressure
and must be filled before the cavity pressure begins to increase. There is some
slight steady state error caused by the non-linear effect of valve displacement on
cavity pressure at the end of the packing stage. This error could be eliminated
with better estimates of initial valve position.
To ensure good performance in the packing stage, the transition from the
filling stage must be smooth. This requires 1) a transition between stages timed
precisely with the end of fill as well as 2) placement of the valve to an
advantageous position for the start of the packing stage. An adaptive
mechanism will next be developed to ferret out these parameters and obtain the
desired process behavior.
Adaptive Scheduling
Individual control algorithms have been developed for the filling and
packing stages using the control structure shown in Figure Chapter 3: -15. As
previously implied, a smooth transition from the filling stage to the packing stage
necessitates moving the valve to a semi-closed position just prior to the end of
fill. Unfortunately, the end of fill is not known but rather depends on the cavity
pressure and flow rate histories of the filling stage. An adaptation mechanism is
necessary to detect the end of the filling stage and move the valve to a position
from which stable control of the packing stage may begin. The process for
achieving this is shown in Figure Chapter 3: -25.
77
Start
Estimate tp & xp
Filling Stage
Move to xp at tp
Packing Stage
Unacceptable
Examine Error Reject Part
Acceptable
Accept Part
Figure Chapter 3: -25: Process Control Algorithm
At the start of the process, the time for end of fill, tp, and initial packing
valve position, xp, are estimated from the empirical equations:
t p = t start + 2 .5 − 0.03 ⋅ Pslope

, (3-18)
x p = 1.0 + 0.05 ⋅ Ppack
where tstart represents the start of fill, Pslope is the desired slope of cavity pressure
in the filling stage, and Ppack is the desired level of cavity pressure in the packing
stage. These equations are rough models of the molding process and provide
time and position estimates from which to start the adaptive algorithm. For
instance, the greater the slope of pressure in the filling stage, the higher the flow
rates and shorter the fill time, tp.
The closed loop control of the filling stage is then initiated and continues
until time tp. At tp, closed loop control of cavity pressure is briefly released. The
valve is moved (under closed loop position control) to an initial valve position for
78
the start of the packing stage. Once the valve is at the desired position, closed
loop control of cavity pressure is restored with the packing stage control
algorithm. At the end of the packing stage, the valve is slowly moved to a fully
closed position; a closing valve speed of 0.6 mm/sec matches the flow displaced
by the valve movement with volumetric shrinkage in the packing stage and
avoids adverse increases in cavity pressure.
The error between the desired and measured cavity pressure traces are
then examined. If the error is significant, a message is sent to the operator that
the part should be discarded. The control system then examines the pressure
history to determine corrective action in the subsequent molding cycle to obtain
the desired response. As previously introduced, there are two adjustable
parameters: tp and xp.
Errors in the estimated transition time, tp, are readily observable in the
cavity pressure profile shown in Figure Chapter 3: -26. If the control system
switches to the packing stage before the cavity is full, the pressure history shown
by the dotted curve results – the cavity pressure decays rapidly as the closing
valve restricts flow and the melt in the cavity decompresses. At 2.5 sec, the
closed loop control of the packing stage is initiated. The valve opens slightly to
allow additional flow into the cavity and the pressure eventually obtains the
desired level. However, the hesitation in flow may result in the cavity not filling
completely, a reduction in part properties, or numerous aesthetic defects.
79

100
Late Switchover
50
Early Switchover
0
0 1 2 3 4 5 6 7
10
Valve Position (mm)
Late
5
Early
0
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 3: -26: Effect of Time Schedule on Packing Stage
The dashed curve in Figure Chapter 3: -26 indicates the cavity pressure
history for a transition begun at 2.5 seconds, after the cavity is full. There are
two spikes in the cavity pressure history, the first occurring when the valve
moves to a semi-closed position before initiation of the packing stage. The
second spike occurs when closed loop control of the packing stage begins and
the controller fully closes the valve in attempt to reduce the cavity pressure.
Thermal volumetric shrinkage does not reduce the cavity pressure to the desired
levels for three seconds – the residual stress levels and part properties may be
significantly affected.
Fortunately, these errors may be used to adjust the timing estimates and
achieve the desired performance in subsequent iterations. If the cavity pressure
prior to the start of the packing stage exceeds the desired packing pressure, the
fill time should be reduced. As with any closed loop control system, large
corrections in fill time (corresponding to high feedback gains) increase the risk of
an unstable adaptive response with the packing pressures perpetually oscillating
80
above and below the desired levels in consecutive molding cycles. Conservative
adaptive gains must be used. Moreover, timing needs to be precise: simulation
indicates the end of fill estimate must be accurate to within 50 mSec to obtain an
acceptable process response.
The second adaptive variable, xp, is an estimate of the initial valve position
at the start of the packing stage. The starting position effects the cavity pressure
history since large valve displacements in the packing phase will non-linearly
effect the cavity pressure being controlled. This effect is shown in Figure
Chapter 3: -27:
80
60
40
20
0
0 1 2 3 4 5 6 7
10
Valve Position (mm)
0
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 3: -27: Effect of Initial Valve Position on Packing Stage
An appropriate valve position for this level of packing pressures is 15 mm. The
dotted and dashed curves represent initial valve positions of 25 and 10 mm,
respectively. As shown in the dotted curve, an initial valve position of 25 mm
permits too much flow into the cavity and higher than desired cavity pressures; a
valve position of 10 mm forces a small rise early in the packing stage due to flow
81
displaced by the valve, then too little flow due to the restricted valve position. In
both cases, the closed loop control of cavity pressure attempts to adjust the valve
position to obtain the desired cavity pressure response – the valve position
moves to approximately 15 mm starting from different initial conditions.
However, the valve movement (after 3.0 sec) causes slight flow in the cavity
which affect the cavity packing pressures.
The cavity pressure at the end of the packing stage may be used to
determine the necessary corrective action. If the packing pressures are too low,
the valve should not be as restricted. Fortunately, small changes in the valve's
starting position do not significantly affect the required start of transition
(otherwise a recursive technique varying both parameters simultaneously must
be utilized). Simulation runs indicate that the estimate of valve position should
converge to within 1 mm to obtain acceptable process behavior.
Between three and seven molding cycles are needed to converge to an

acceptable response using conservative adaptation mechanisms. A typical
cavity pressure profile and the process' adaptation is shown in Figure Chapter 3:
-28. In the figure, the control system switched from the filling to the packing
stage too late – the cavity was already full and a large pressure spike occurred.
In the second iteration, the control system switched slightly early resulting in a
hesitation in the filling and unacceptable delay in the packing stage. By the third
stage, the system had converged to an excellent response.
82
100
90 1
80

70
60
3
50
40
2
30
20
10
0
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 3: -28: Convergence to Desired Profile
In Figure Chapter 3: -28, only three iterations were needed to obtain a

desired response. This may vary significantly, however, with the mold geometry,
material properties, machine behavior, or desired process response.
Unfortunately, no formal methodologies exist to estimate the performance or
prove the stability of these adaptive mechanisms. Conservative adaptation gains
must be used to ensure convergence; the rate of convergence can be increased
only with increased chance of instability.
This simulated control system has been tested with a variety of cavity
pressure profiles, input hydraulic pressures, and mold geometries. In most
cases, the process exhibited convergence and acceptable response. There were
some instances at the borders of the process envelope, however, where the
process could not converge to the desired response. These situations have not
been presented due to space considerations. However, the performance and
limitations of the actual molding process will be investigated in Chapters 4 and 5.
83
SUMMARY
In this chapter, simplifying assumptions have been utilized to develop a

minimum complexity process model which effectively predicts the aggregate
dynamics of cavity pressure in the injection molding process. While non-linear,
the model provided a means to understand the process dynamics and revealed
issues relevant to control system development.
An adaptive gain scheduling approach was employed in controlling the

cavity pressure throughout the filling and packing stages. Linearizing the plant
about these stages provided a means for estimating the capabilities and
robustness of the control system. An adaptive mechanism was developed to
identify the point of switchover from the filling to the packing stage and provide a
smooth and stable transition between stages. For reference, the Matlab listings
for the adaptive algorithm and closed loop simulation have been included in
Appendix A.1.
The linearized models are valid approximations for the non-linear plant
model about a narrow operating range only. Moreover, the non-linear plant is
only an approximation of a more complex physical process. As such, it is
extremely unlikely that the linearized models will provide good prediction of the
closed loop response of the physical process – the physical implementation and
experimental validation of this approach must now be presented.
84
70
O’Neill, M. E., Chorlton, F., Viscous and Compressible Fluid Dynamics, John Wiley & Sons, New York,
1989.
71
Nagy, M. R., Long-Term Shrinkage of Polypropylene, Proceedings from the 1993 Annual Technical
Meeting of the Society of Plastics Engineers, v. 51, pp. 2155 (1993).
72
Schlitling, H., Boundary Layer Theory, Seventh Edition, McGraw-Hill, New York, New York, 1979.
73
Lee, H.S., Thin-Cavity Filling Analysis using the Finite Element Method with Control Volume Techniques,
Doctoral Dissertation to Rensselaer Polytechnic Institute Department of Mechanical Engineering, 1989.
74
Chiang, H.H., Simulation and Verification of Filling and Post-Filling Stages of the Injection Molding
Process, Doctoral Dissertation to Cornell University Sibley School of Mechanical Engineering, 1989.
75
Kazmer, D. O., Willey, S. J., An Examination of Material Characterization and Modeling Techniques for
Injection Molding Simulation, Proceedings from the 1992 Annual Technical Meeting of the Society of
Plastics Engineers, v. 50, pp. 958 (1992).
76
Williams, M. L., Landel, R. F., Ferry, J. D., A Model for the Viscosity of Molten Polystyrene, Journal of
American Chemical Society, v. 77, n. 8, pp. 3701 (1955).
77
Zoller, P., Bolli, P., Pahud, V., Ackermann, H., Apparatus for Measuring Pressure-Volume-Temperature
2
Relationships of Polymers to 350 °C and 2200 kg/cm , Review Scientific Instrumentation, v. 47, pp. 948
(August 1976).
78
Zoller,P., Pressure-Volume-Temperature Relationships of Solid and Molten Polypropylene, Journal of
Applied Polymer Science, v. 23, pp. 1057 (1979).
79
Kamal, M.R., Lafleur, P.G., Computer Simulation of Injection Molding, Polymer Engineering and Science,
v. 22, n. 17, pp. 1069 (1982).
80
Seborg, D.E., Edgar, T.F., Shah, S.L., Adaptive Control Strategies for Process Control: A Survey, AICheJ,
A Publication of the American Institute of Chemical Engineers, v. 32, pp. 881-913 (June 1986).
81
Sanschagrin, B., Process Control for Injection Molding, Polymer engineering and Science, v. 23, n. 8, pp.
431 (1983).
82
Korem, Y., Adaptive Control Systems for Machining, Manufacturing Review, v. 2, n. 1, pp. 6 (March 1989).
83
deVries, B. A. J., Verbruggen, H. B., Multivariable Process and Prediction Models in Control, International
Journal of Adaptive Control, v. 8, n. 2, pp. 261, (1994).
85
84
Roffel, B., Vermeer, P.J., Chin, P.A., Simulation and Implementation of Self-Tuning Controllers, Prentice
Hall, Englewood Cliffs, New Jersey, 1989.
85
Butler, H., Model Reference Adaptive Control, Prentice Hall-Hemel Hapstead, Hertfordshire, 1992.
86
Sastri, S.S., Isidori, A., Adaptive Control of Linearizable Systems, IEEE Transactions on Automatic Control,
v. AC-34, pp. 1123 (1989).
87
Franklin, G.F., Powell, J.D., Workman, M.L., Digital Control of Dynamic Systems, Addison-Wesley, New
York, New York, 1992, pp. 462.
88
Franklin, G.F., Powell, J.D., Workman, M.L., Digital Control of Dynamic Systems, Addison-Wesley, New
York, New York, 1992, pp. 468.
89
Ljung, L., System Identification Toolbox, The Math Works, Cambridge, Massachussets, 1988.
90
Franklin, G. F., Powell, J. D., Emami-Naeini, A., Feedback control of Dynamic Systems, Addison-Wesley,
New York, New York, 1988.
86
TABLE OF CONTENTS
CHAPTER 4: EXPERIMENTAL VALIDATION OF CONTROL STRATEGY ................................84

IMPLEMENTATION ..................................................................................................................................84
Control System ........................................................................................................................................84
Molding Machine ....................................................................................................................................86
Feed System.............................................................................................................................................88
Mold Cavities ..........................................................................................................................................90
SIMULATION PERFORMANCE ...............................................................................................................91
Actuation Performance............................................................................................................................91
Open Loop Comparison ..........................................................................................................................94
CAVITY PRESSURE CONTROL...............................................................................................................97
Demonstration.........................................................................................................................................97
Controller Tuning....................................................................................................................................99
Filling Stage ........................................................................................................................................................ 99
Packing Stage .................................................................................................................................................... 101
Adaptation ......................................................................................................................................................... 104
SUMMARY ...............................................................................................................................................107
TABLE OF FIGURES
FIGURE 4-1: PICTURE OF CONTROL SYSTEM ...................................................................................................86
FIGURE 4-2: PICTURE OF MOLDING MACHINE .................................................................................................87
FIGURE 4-3: CROSS-SECTION OF FEED SYSTEM ...............................................................................................89
FIGURE 4-4: MOLD INSERT GEOMETRIES ........................................................................................................91
FIGURE 4-5: EFFECT OF HYDRAULIC LEAKAGE AND COMPRESSIBILITY ON VALVE POSITION ..........................92
FIGURE 4-6: EFFECT OF CAVITY PRESSURE ON VALVE RESPONSE TIME ...........................................................93
FIGURE 4-7: INPUT PROFILES FOR SIMULATION: VALVE POSITION AND INJECTION PRESSURE .......................95
FIGURE 4-8: SIMULATION AND EXPERIMENTAL PRESSURE PROFILES FOR CONSTANT VALVE POSITION ...........95
FIGURE 4-9: TYPICAL PRESSURE PROFILE WITH MEASUREMENT INDICES ........................................................98
FIGURE 4-10: FILLING STAGE PRESSURE PROFILE WITH NORMAL AND HIGH GAINS .......................................100
FIGURE 4-11: PACKING STAGE PRESSURE PROFILE WITH NEGATIVE FEEDBACK ............................................102
FIGURE 4-12: PACKING STAGE PRESSURE PROFILE WITH NEGATIVE FEEDBACK ............................................103
FIGURE 4-13: PACKING STAGE PRESSURE PROFILES FOR CONSTANT VALVE POSITION ..................................104
FIGURE 4-14: TYPICAL ADAPTIVE CAVITY PRESSURE CONVERGENCE IN FOUR SHOTS ...................................105
FIGURE 4-15: NUMBER OF SHOTS REQUIRED TO MANUFACTURE GOOD PARTS ..............................................106
1
TABLE OF TABLES
TABLE 4-1: INCOMING CONTROL SIGNALS ......................................................................................................85
TABLE 4-2: PERFORMANCE PARAMETERS OF MITSUBISHI 390 MJ MOLDING MACHINE ..................................87
TABLE 4-3: MODIFICATIONS PERFORMED TO OBTAIN DYNAMIC FEED SYSTEM ..............................................89
TABLE 4-4: DESIGN PARAMETERS OF FEED SYSTEM .......................................................................................90
START OF CHAPTER:84
2
Chapter 4: Experimental Validation of Control Strategy
Chapter 4: Experimental Validation of
Control Strategy
The experimental validation of Dynamic Feed Control is comprised of two

parts. First, this chapter presents experimental data which verifies the simplifying
assumptions utilized in development of the process model. Then, the molding
process is correlated to the minimum complexity process model developed in
Chapter 3. After the open loop behavior of the molding process has been
described, further experimental work validates closed loop cavity pressure
control in both the filling and packing stages of injection molding. Such closed
loop cavity pressure control has never before been accomplished.
Following validation of the control strategy, Chapter 5 describes the

process capability of the Dynamic Feed Control process – demonstrating its
extraordinary process flexibility and enhanced process consistency.
IMPLEMENTATION
A system schematic for Dynamic Feed was presented in Figure 13 of

Chapter 2. There are four major sub-systems shown in the figure: control
system, molding machine, feed system, and mold cavity. Each sub-system will
now be described in detail.
Control System
The control system was programmed in the C programming language,

utilizing a 90 Mhz Pentium® processor (PCI5/90, Tangent Computer, Burlingame,
84
CA). An internal data acquisition board (AT-MIO-16L-9, National Instruments,

Austin, TX) monitored the nine incoming signals listed below in Table Chapter 4:
-1. All of the instrumentation was selected to provide 0 to 10 volt modulated
analog signals at a sampling frequency of 100Hz. As will later be demonstrated,
this range of signal voltage and data sampling rate provided ample process detail
without excess sensitivity to signal noise. The control system employed a
software development kit (Lab Windows for DOS version 2.3a, National
Instruments) to provide high level graphical interface and data I/O subroutines. A
photograph of the physical embodiment of the control system is shown in Figure
Chapter 4: -1.
Table Chapter 4: -1: Incoming control signals

molding machine start
injection pressure
melt temperature
four cavity pressures
two valve positions
During operation, the control system idles until it receives a start of

injection signal from the molding machine. The digital timer on the I/O board is
then reset and the cycle begins using the algorithm previously described in
Figure 15 of Chapter 3. As the molding cycle progresses, the control system
analyzes the process data and issues -10 to +10 volt analog signals through the
data acquisition board to the two hydraulic servo-valve controllers (Vickers EEA-
PAM-535-A-30, Paul-Munroe Sweetland, Whittier, CA). These controllers
provide 0 to 10 amp currents at 24 VDC to the proportional hydraulic valves on
the molding machine to alter the flow through the feed system.
85
Figure Chapter 4: -1: Picture of control system
Molding Machine
The molding trials were performed at GE Plastics’ Commercial

Development Center in Pleasanton, CA utilizing the 400 ton molding machine
(390MJ-100, Mitsubishi Heavy Industries Ltd., Tokyo, Japan) pictured in Figure
Chapter 4: -2. Key performance parameters for this machine are listed in Table
Chapter 4: -2. The machine’s hydraulic system provides a source of hydraulic
power to the feed system control. Proportional servo-valves (Vickers KFDG-4-V-
5, Paul-Munroe Sweetland) were mounted on the machine's hydraulic manifold
and replaced the two directional control valves normally used for core pulls
during the molding process.
86
Table Chapter 4: -2: Performance parameters of Mitsubishi 390 MJ molding machine

Parameter Value
type of machine hydraulic
clamp tonnage 390 metric tons
injection pressure 200 MPa (30,000 psi)
shot size 200 cc (20 ounce)
control modes closed loop ram velocity
closed loop injection
pressure
# of moving cores two
Since the molding machine's injection hydraulics were not directly

controlled by the control system, the ram velocity and injection pressure were
specified by the process technician. Generally, an injection profile was utilized
which provided consistent, excessively high injection pressures to the polymer
melt. The Dynamic Feed control system then varied the valve resistance to
obtain the desired cavity pressures. Better performance could have been
obtained if the Dynamic Feed control system encapsulated the molding
machine's controller, but this task is well outside the scope of this research.
Figure Chapter 4: -2: Picture of molding machine
87
The use of a 390 ton molding machine was necessitated by the

packaging of the feed system, mold, and instrumentation in a mold base
requiring 75 cm (30 inches) of daylight (defined as the clearance between the
molding machine’s plattens). Since only 10% of the machine's rated shot size is
utilized during the molding process, residence times for the polymer melt
approached five minutes. As such, the potential for material degradation was
severe and restricted the use of resins to thermally stable materials. The risk of
degradation was further minimized by operating in the lower range of melt
temperatures and avoiding excessively long delays between cycles. No signs of
degradation (black specs, splay, etc.) were apparent in any of the molded parts
while running on a continuous cycle.
Feed System
The feed system shown in Figure Chapter 4: -3 is a custom hot manifold

developed by Kona Corporation (Gloucester, MA) specifically for this research. A
base manifold (VG-33SR single bar manifold, Job #10120) was modified as listed
in Table Chapter 4: -3. Melt temperature and pressure transducers (MTX-935
and PT465XL-30M, respectively, Dynisco Instruments, Sharon, MA) were
installed in a heated sprue bushing to obtain the melt inlet conditions. The most
significant modifications involved the accommodation of miniature position
transducers (HydraStar HS-1000, Data Instruments, Boston, MA) to monitor each
of valve stem’s changing position. Together with their signal amplifiers (SP-50,
Data Instruments), these inductance transducers provided position data to 0.01
mm with a response time less than 1 mSec. Accommodating the position
transducers required a re-design of the hydraulic actuator as well as the addition
of an auxiliary plate behind the hydraulic plate as shown in Figure Chapter 4: -3.
88
Table Chapter 4: -3: Modifications performed to obtain Dynamic Feed system

Area Modifications
valve stem increased diameter
lengthened taper
hydraulic cylinder increased bore and length of travel
added auxiliary plate for transducers
melt inlet manufactured heated sprue bushing
installed melt and pressure transducers
Figure Chapter 4: -3 illustrates the resulting design. This particular

manifold design utilizes heat pipes imbedded in the manifold and drops to obtain
a uniform melt temperature within 3°C of nominal.91 Altogether, the design
incorporated five heater zones which were externally controlled (SmartSeries
CSS15G, D-M-E Corporation, Madison Heights, WI).
Pr essure
Tr ansducer
Posi ti on
Tr ansducer Auxili ary
Plate
Act uator
Zone 1
Hydrauli c
Plate
Zone 2
Mani fold
Bar
Zone 3
Tapered Valve Stem Tapered Valve

Stem
Zone 4 Zone 5
Figure Chapter 4: -3: Cross-section of feed system
89
Several critical design parameters of the feed system are listed in Table
Chapter 4: -4. With the given bore of the hydraulic actuators and flow rate
capacity of the hydraulic proportional valves, the maximum speed of the valve
stem is approximately 30 cm/sec. Including the acceleration dynamics of the
valve, the valve stem can move across its full range of motion in approximately
0.150 sec – much faster than will generally be required during the molding
process.
Table Chapter 4: -4: Design parameters of feed system

Parameter Value
actuator area 15 cm2
valve stem diameter 1 cm
length of valve taper 1.5 cm
angle of valve taper 8°
length of valve travel 2 cm
Mold Cavities
A standard 30 cm by 50 cm mold base (1220A-37-13, D-M-E-

Corporation) was machined to accept a series of 20 cm x 30 cm x 3 cm
rectangular inserts. Each P-20 steel insert was designed to accommodate four
pressure transducers (PT4656XL-30M, Dynisco Instruments) for feedback of
cavity pressures during the molding process. The pressure transducers were
located very close to each gate as well as at the end of flow; these pressure
transducers provided a precision of approximately 0.1 MPa (20 psi) and a
response time of 15 mSec.92
The inserts permitted the molding of several different plaque geometries,

shown in Figure Chapter 4: -4. Altogether, three inserts were tooled: a one cavity
insert with a 2 mm wall thickness, a one cavity insert with a 3 mm wall thickness,
and an unbalanced two cavity insert with a wall thickness of 2 mm. Temperature
control of the mold and mold inserts was achieved through two external water
heaters (MC Jet III 15-30, Matsui Co. Ltd., Tokyo, Japan).
90
Gates
Pressure Transducers
12.5 cm
5 cm
5 cm
9.5 cm 6 cm 5 cm
10 cm
12.5 cm
11.3 cm
Figure Chapter 4: -4: Mold insert geometries
SIMULATION PERFORMANCE
The process model developed in Chapter 3 was validated with the

hardware implementation just described. The validation involved comparison of
the simulation’s open loop prediction to data collected in the injection molding
process. If the simulation closely models the process dynamics, then the control
strategies developed in Chapter 3 should be successful in closed loop control of
cavity pressure.
Actuation Performance
Several simplifying assumptions regarding the hydraulic actuators were

utilized in the development of the process model shown in Figure 6 of Chapter 3.
In particular, the simulation assumed an ideal hydraulic system, i.e. that the
hydraulic fluid was incompressible, that there was no leakage in the hydraulic
actuators, and that the valve stem response was independent of the magnitude
of cavity pressures. Figure Chapter 4: -5 shows the combined effects of
hydraulic leakage and fluid compressibility on the valve position. In the figure,
the valve stem is moved to a position of 35 mm using closed loop feedback. At
91
1.0 seconds, after the valve has reached its desired position, the feedback loop
is turned off and the hydraulic servo-valve is provided zero current –
corresponding to a blockage of flow into and out of the hydraulic actuator. During
this time, molten plastic is flowing past the valve into the cavity, raising the
pressure in front of the valve stem as the cavity fills and packs.
40
39
38
37
Valve Position (mm)
36
35
34
33
32
31
30
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 4: -5: Effect of hydraulic leakage and compressibility on valve position
During this experiment, the melt pressure near the valve exceeded 100
MPa (14,000 psi). However, Figure Chapter 4: -5 shows that the valve stem
largely maintains its position. This demonstrates that the hydraulic leakage and
fluid compressibility is not significant. A displacement of 0.5 mm represents
roughly 1% of the axial travel of the valve stem, not enough to significantly alter
the process dynamics. This finding suggests that displacement transducers for
acquisition of valve stem position may not be necessary in future feed system
designs since the valve position may be adequately controlled via open loop
control.
92
The analysis also assumed that the valve stem response was independent
of the melt pressure in front of the valve stem. To investigate this effect, the
valve was moved from a 75% open position to a 25% open position when the
melt pressure reached 5, 30, 55, and 80 MPa. The response times for the
varying melt pressures are shown in Figure Chapter 4: -6.
300
250
Response Time (ms)
200
150
100
50
0
0 10 20 30 40 50 60 70 80
Cavity Pressure (MPa
Figure Chapter 4: -6: Effect of cavity pressure on valve response time
These results indicate that the melt pressure does, in fact, have an effect
on the time response of the actuator. As the melt pressure increases, the valve
stem responds more slowly due to the increased force of the melt pressure
acting on the valve stem. This effect, however, is not particularly significant since
the response time is an order of magnitude faster than the time dynamics of the
molding process. Additionally, the valve stem response times may be further
improved by increasing the proportional gains in the control system (currently
30% of maximum) or by increasing the hydraulic supply pressure (currently 6
MPa of a 20 MPa maximum). Experimental results show that the valve stem
response for hydraulic supply pressures of 8 MPa (1,000 psi) are acceptable to
cavity pressures of 140 MPa (20,000 psi) which is representative of the
93
maximum pressure used in commercial molding. At higher cavity pressures, the

melt forces on the valve stem approach the actuation forces and the hydraulic
supply pressures must be increased to 12 MPa (1,500 psi).
Open Loop Comparison
Additional assumptions utilized in the model include a Newtonian viscosity

( µ = constant ), a bi-linear compressibility model for the polymer melt
( ρ ( P , T ) = ρ 0 + β ⋅ P + β T ⋅ T ), and a constant rate of cooling in the packing stage
( dT
dt
= constant ). An open loop comparison between predicted and observed
cavity pressures can validate not only our model of the process dynamics but
also our strategies to control those dynamics. The term ‘open loop’ indicates that
the cavity pressure is not being directly controlled through the valve dynamics.
Rather, in this case the valve is moved to a desired position which is maintained
throughout the filling and packing stages – the cavity pressure is determined
solely by the process dynamics.
Figure Chapter 4: -7 plots the simulation inputs consisting of the valve

stem position and injection pressure. In these figures, the solid ‘noisy’ trace is
the observed process data. The dotted, smoother trace is the simulation input
which has been chosen carefully to closely match the observed data. The valve
position shown in the left hand figure is roughly 60% open. For reference, the
melt and mold temperatures were 280°C and 70°C, respectively. At a melt
temperature of 280°C, the polycarbonate’s viscosity is estimated from Figure 3-
11 to be roughly 800 Pa· Sec.
94
40 100
35 90
80
30
Injection Pressure (MPa)

Valve Position (mm) 70
25
60
20 50
15 40
Key
30
10 Simulation Input
Experimental Data 20
5
10
0 0
0 2 4 6 8 0 2 4 6 8
Time (sec) Time (sec)
Figure Chapter 4: -7: Input Profiles for Simulation: Valve Position and Injection Pressure
The dotted and solid traces plotted in Figure Chapter 4: -8 represent the
predicted and observed cavity pressure responses – the Matlab code used to
generate the pressure prediction is listed in Appendix A-3. The scale in Figure
Chapter 4: -8 is matched with the plot of injection pressure in Figure Chapter 4: -
7 for comparison purposes.
100
90 Key
Simulation Output
Experimental Data
80
70
(3)
60
50 (4)
40
30
20
(2)
10
(0)
(1)
Figure Chapter 4: -8: Simulation and experimental pressure profiles for constant valve
position
95
For clarity, a description of the process dynamics for each trace is

provided before directly comparing the predicted and observed response. In the
observed molding process (shown by the solid trace in Figure Chapter 4: -8) at 0
seconds (0), the valve opens immediately and the ram begins acceleration. The
increasing injection pressure begins to rise as plotted in Figure Chapter 4: -7.
However, the opening of the valve creates a void in the feed system which must
be filled before the melt can enter the cavity. Hence, no pressure is detected by
the cavity pressure transducer until 1.0 seconds into the filling stage (1). The
melt progresses down the cavity and the cavity pressure increases fairly linearly.
At 2.0 seconds, the cavity is filled – the flow rate drops and the cavity pressure
quickly climbs towards the injection pressure (2). At 3.5 seconds, the molding
machine drops to a lower level of packing pressure causing a slight decay in the
cavity pressures (3). The cavity pack pressure maintains 45 MPa until the
injection pressure is released at 9.0 seconds and the cavity pressure decays to
zero (4).
The simulation is an adequately close approximation of the process

dynamics. The melt enters the cavity at about 1.0 seconds followed by a fairly
uniform rise in cavity pressures (1). At 3.0 seconds the cavity is full and the
cavity pressure quickly approaches the injection pressure (2). A cavity pack
pressure of 58 MPa is maintained until the injection pressure is released and the
cavity pressures begins to decay (4). Generally, there is acceptable correlation
in the magnitude and shape of the cavity pressure. This indicates that the
assumptions and modeling approach are valid and the closed loop cavity
pressure control strategies developed in Chapter 3 should apply.
Figure Chapter 4: -8 does illustrate two shortcomings of the model. First,

the simulation has under-estimated the flow rates in the filling stage. This is
apparent by the slightly longer delay in the simulated filling of the cavity, the
lower slope of cavity pressure in the filling stage, and the deviation in predicted
96
filling time. This error is likely due to geometric modeling errors and should not
significantly impede the control system performance. A more critical shortcoming
in the model, however, is the deviation in asymptotic pack pressure at the end of
the packing stage. This error is likely due to the simplifying assumptions
regarding material or process behavior and might be a precursor of difficulties to
come in controlling the cavity pressure during the packing stage.
CAVITY PRESSURE CONTROL
Closed loop, cavity pressure control is the most effective strategy for
improving the flexibility and consistency of the molding process. By controlling
cavity pressures in a closed loop process, the effects of material, machine, and
operator variation may be removed to provide enhanced process consistency.
By using the Dynamic Feed Control’s multiple degrees of freedom, the process
dynamics may be effortlessly modified to provide revolutionary process flexibility.
The system’s ability to control cavity pressure is next validated.
Demonstration
Figure Chapter 4: -9 illustrates an actual closed loop cavity pressure

response (solid trace) for a typical input pressure profile (dotted line). For this
research, an input profile is defined by the filling stage start time, the slope of
cavity pressure in the filling stage, and the level of cavity pressure in the packing
stage. The closed loop and adaptive control algorithms will dynamically vary the
valve positions to obtain the desired profile shown in the figure.
97
40
35
30

25
Filling Stage Packing Stage
20
15
Key
10 Control Input
Experimental Data
5 Tolerance Bands
0
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 4: -9: Typical pressure profile with measurement indices
In this particular example, the valve was opened at 0.5 seconds to initiate
the flow. At 1.0 seconds, the melt enters the cavity, as indicated by the time at
which the pressure transducer registers a cavity pressure over 1.0 MPa. If the
cavity pressure entered the cavity more than 0.1 seconds away from the desired
entry time, the adaptive algorithm would readjust the valve timing in the
subsequent shot to correct the error. Once the melt has registered on the
pressure transducer, the closed loop control algorithm then maintains the cavity
pressure on the designated slope as shown in Figure Chapter 4: -9 between 1.0
and 2.5 seconds.
Just prior to the switchover between the filling and packing stages at 2.5
seconds, the control system moves the valve to a nearly closed position for
control of the packing stage. The rise in pressure at 2.5 seconds is due to the
displacement of flow ahead of the closing valve stem which generates temporary
increases in flow rates and pressure in the cavity. The quick decay at 2.6
seconds occurs when the valve stops moving and flow rates have been
98
significantly decreased due to increased flow resistance through the nearly

closed valve. The final rise in cavity pressure at 2.7 seconds occurs when the
cavity is completely full and the flow rates through the valve stabilize.
As described in Chapter 3, the timing of the valve’s switchover determines

the level of cavity pressures just after the switchover to the packing stage. If the
cavity pressure is more than 5 MPa above or below the desired pack pressure,
then the switchover occurred too early or late (respectively) and will be adjusted
in the subsequent shots by the adaptive control algorithm. Finally, the pack
pressure level is measured by the asymptotic cavity pressure prior to the end of
the packing stage. If the pack pressure does not conform to the desired value,
the valve stem position will be adjusted in subsequent shots to improve the pack
pressure levels. Examples of this adaptive convergence are provided in
subsequent sections.
Controller Tuning
The closed loop, cavity pressure control capability demonstrated in Figure

Chapter 4: -9 has never before been achieved – and not for lack of people
trying.27,34,35,36,38, 40,41,42,43,44,45,46 Previous attempts have had only partial success
since the process dynamics are determined by the mold steel in the conventional
molding process. Even though Dynamic Feed Control now enables extra
degrees of freedom with which to alter the process dynamics, the closed loop
control and adaptive control mechanisms are not trivial problems. The
successful development of these control systems are now described.
Filling Stage
The filling stage of the injection molding process is characterized by a

fairly homogeneous polymer melt, high flow rates, and increasing flow resistance.
Due to these characteristics, the control strategy developed in Chapter 3 utilized
proportional feedback control of cavity pressure coupled with weak proportional
99
valve position control. This type of control would tend to maintain the valve in a
semi-open position then further open or close the valve as needed to deliver the
desired filling stage response.

30 High Gains
Key
Control Input
20 Experimental Data
10 Low Gains
0
0 0.5 1 1.5 2 2.5 3 3.5
40 High Gains
Valve Position (mm)
Low Gains
30
20
10
0
0 0.5 1 1.5 2 2.5 3 3.5
Time (sec)
Figure Chapter 4: -10: Filling stage pressure profile with normal and high gains
Figure Chapter 4: -10 displays the cavity pressure response for two
different levels of gains for the closed loop control of cavity pressures in the filling
stage. The control system is trying to deliver the desired cavity pressure profile
indicated by the dotted line. The lower set of gains (proportional gains of 1
mm/sec/MPa) produces a cavity pressure trace as desired with a valve position
remaining close to 25 mm throughout the filling stage. However, a higher set of
gains (proportional gains of 3 mm/sec/MPa) produces significantly more valve
movement and results in an unstable cavity pressure response. At 3.0 seconds,
the control system closes the valve too quickly in an attempt to reduce the cavity
pressure. This produces a small increase in cavity pressure which drives the
control system to fully close the valve. The cavity pressure decays fairly quickly
after which the valve reopens and the filling of the cavity is resumed.
100
There is quite a wide range of gains for which the system is stable,
between 0.5 and 2.0 mm/sec/MPa. Moreover, the system stability was not
affected by changes in mold temperature, melt temperature, or hydraulic
pressures. The stability of control system did, however, depend on the input
pressure profile. For large slopes of cavity pressures, reducing the proportional
gains enhanced system stability. Interestingly, this effect was previously
predicted in Figure 3-19 on page 70.
Packing Stage
Control of cavity pressures in the filling stage was relatively easy. The
packing stage, however, is characterized by very low flow rates determined
solely by volumetric changes due to the compressibility and cooling of the
polymer melt. These low flow rates will make control of the packing stage
considerably more difficult since small displacements of the valve may
substantially effect the cavity pressures in the packing phase. As a first attempt,
the same control law from the filling stage was utilized, albeit with much smaller
proportional gains to minimize the valve displacement. The resulting cavity
pressure response is shown in Figure Chapter 4: -11.
Key
Control Input
Experimental Data
40
30
20
10
0
5 5.5 6 6.5 7 7.5 8
15
101
Figure Chapter 4: -11: Packing stage pressure profile with negative feedback
In this figure, the desired packing pressure is 25 MPa, as indicated by the

dotted line. The control system has transitioned into the packing stage at 5.3
seconds. At this point, the cavity pressure is below the desired level so the
control system opens the valve at 5.4 seconds. This results in an immediate
drop in cavity pressure due to negative volume displacement caused by the valve
stem’s movement. The cavity pressure then begins to increase. By 5.5 seconds,
the cavity pressure is at the desired level and the valve position has leveled off.
However, the cavity pressure continues to rise so the control system slowly
closes the valve. This act of slowly closing the valve increases the cavity
pressure due to the positive displacement produced by the valve movement.
The increase in cavity pressure drives the control system to fully shut the valve.
The cavity pressure does not decay until the melt in the cavity contracts due to
thermal shrinkage.
Given the undesirable interaction which develops between the cavity

pressure and the valve displacement, a natural solution might be to use the valve
as a piston utilizing positive feedback. If the cavity pressure is below the desired
level, then shut the valve to force more melt into the cavity and increase cavity
pressure. This approach is shown in Figure Chapter 4: -12.
102
Key

40 Control Input
Experimental Data
30
20
10
0
5 5.5 6 6.5 7 7.5 8
15
Valve Position (mm)
10
0
5 5.5 6 6.5 7 7.5 8
Time (sec)
Figure Chapter 4: -12: Packing stage pressure profile with negative feedback
Unfortunately, this is not an effective means for control because of the

limitations on valve travel. As shown in the figure, the valve has transitioned into
the packing stage at 5.25 seconds. The control system recognizes that the
cavity pressure is below the desired level and, as such, closes the valve to force
more material into the cavity. The cavity pressure responds favorably but then
begins to slowly decay as the material in the cavity solidifies. The control system
tries to maintain the cavity pressure by further closing the valve but the attempt is
ineffective. By 6.2 seconds the valve is fully closed and the cavity pressure is
free to decay.
Due to the undesired linkage between the valve displacement and cavity
pressures in the packing stage, the control system was forced to utilize strong
position control with weak negative feedback of cavity pressures. The resulting
cavity pressure response is shown in Figure Chapter 4: -13. In this scenario, a
valve position is estimated for the packing stage. The valve then remains at a
nearly constant position during the packing stage. If the cavity pressure deviates
103
significantly from the desired level, then the base valve position is changed in
subsequent shots to obtain acceptable packing pressures. This strategy,
unfortunately, is incapable of eliminating disturbances within a single molding
cycle. However, it does provide effective removal of systematic disturbances
such as material, machine, and operator variation. Several other feed system
design concepts have been developed in Appendix C to provide better packing
stage control of cavity pressures.
Key
Control Input
40
Experimental Data
30
20
10
0
5 5.5 6 6.5 7 7.5 8
15
Valve Position (mm)
10
0
5 5.5 6 6.5 7 7.5 8
Time (sec)
Figure Chapter 4: -13: Packing stage pressure profiles for constant valve position
Adaptation
An adaptation mechanism is necessary to govern the transitions between

the control of the filling and packing stages which have been described.
Altogether, there are three variables which must be determined: valve initiation
time for the filling stage, valve transition time between the filling and packing
stages, and valve position for the packing stage. Each of these parameters is
estimated before the molding process begins and is updated with new process
information as described in Chapter 3. Figure Chapter 4: -14 plots the cavity
104
pressure traces for four sequential shots, starting at the beginning of a molding
trial.
10000
9000 (1)
8000
(2)
7000
Cavity Pressure (psi)
6000
5000 (4)
4000
(3)
3000
Key
2000 Control Input
Experimental Data
1000
0
0 1 2 3 4 5 6 7
Time (sec)
Figure Chapter 4: -14: Typical adaptive cavity pressure convergence in four shots
As shown in Figure Chapter 4: -14, the initial estimates provide a late

transition to the packing stage and a large overshoot in the cavity packing
pressure (1). The adaptive algorithm recognizes this pressure spike and reduces
the transition time to obtain the response shown in trace two (2). The cavity
pressure again overshoots resulting in another reduction in transition time to
obtain trace three (3). The transition time is excellent. However, the control
system recognizes that the pack pressure decays below the desired level of pack
pressure. As such, the algorithm (shown in Figure 3-25 on page 77) increases
the valve position in the packing stage to obtain the pressure response shown in
trace four (4). This response is within the defined tolerance limits of the control
system. The part is designated as ‘good’ and the process parameters are used
for the subsequent shots. If a process change were to occur during the
manufacture of a batch of parts – a change in process conditions or material
105
properties, for instance – the adaptive system would take corrective actions to
maintain the desired process dynamics.
The performance of the adaptive system might best be measured by the

number of shots required for the molding process to converge to an acceptable
response. Forty molding trials were performed for different part geometries, wall
thicknesses, process conditions, and input profiles. For each molding trial, the
number of shots required to obtain good parts was recorded. Figure Chapter 4: -
15 displays the frequency distribution for all the molding trials.
35
30
25
Frequency (%)
20
15
10
0
0 2 4 6 8 10 12 14 16
Number of Shots To Acceptable Profile
Figure Chapter 4: -15: Number of shots required to manufacture good parts
In every single case, the adaptive algorithm did converge to an acceptable

process response. The number of shots which were required depended largely
on the accuracy of the initial estimate. In two cases, the initial estimates of
process parameters were very good and resulted in production of good parts on
the second shot. In two other cases, the initial estimates were very poor,
requiring more than a dozen shots to obtain good parts. On average, the
adaptive algorithm converged in about eight or nine shots after startup. This
106
performance is satisfactory considering the large number of parts which are

commonly produced in a batch of molded parts.
The adaptive algorithm is itself a form of feedback control. The speed of

convergence may be increased significantly by increasing the adaptation gains,
i.e. the amount with which each process parameter is adjusted between shots.
However, this strategy was found to occasionally lead to severe oscillations in
the process parameters as well as an unacceptable process response.
Accordingly, a conservative set of adaptive gains was used to ensure consistent
convergence across a variety of mold geometries, process conditions, material
properties, and input profiles.
SUMMARY
This chapter described the implementation of the process concept as well

as the validation of the control strategy. The implementation does not require
significant modifications to conventional feed systems or machine operating
procedure. Moreover, the implementation proved quite robust throughout the
experimental validation – more than 6,000 parts were molded in over 100
molding trials without a single failure in the prototype hardware!
For the first time, simultaneous closed loop control of multiple cavity
pressures was achieved in both the filling and packing stages by exploiting the
degrees of freedom provided by Dynamic Feed Control. The control system
successfully utilized an adaptive gain scheduling procedure for transition
between the filling and packing stages. Both the closed loop and adaptive
control mechanisms were conservatively designed to provide an acceptable
response across a wide variety of process dynamics.
Now that the control strategy has been experimentally validated, the
process capability may be investigated as described in the next chapter.
107
108
91
Temperature Control of Kona VG Series Manifold, Kona Corporation Manifold Design Manual, Kona
Corporation, Gloucester, MA, 1994.
92
Time Response of the PT-XL Amplified Pressure Transducer, Dynisco Instruments Technical Literature,
Dynisco Instruments, Sharon, Ma, 1988.
109
TABLE OF CONTENTS
CHAPTER 5: EXPERIMENTAL VALIDATION OF PROCESS CAPABILITY.............................107

PROCESS FLEXIBILITY..........................................................................................................................107
Filling Stage Flexibility.........................................................................................................................108
Packing Stage Flexibility.......................................................................................................................113
PROCESS CONSISTENCY ......................................................................................................................118
Cavity Pressure Response .....................................................................................................................121
Filling Stage Consistency ......................................................................................................................122
Packing Stage Consistency....................................................................................................................125
PROCESS LIMITATIONS ........................................................................................................................129
Limitations from Process Physics .........................................................................................................129
Filling Stage ...................................................................................................................................................... 129
Packing Stage .................................................................................................................................................... 130
Temperature Effects .......................................................................................................................................... 131
Limitations from System Design............................................................................................................132
Response Time .................................................................................................................................................. 132
Hydraulic Pilot .................................................................................................................................................. 134
Valve Interaction ............................................................................................................................................... 135
SUMMARY ...............................................................................................................................................136
TABLE OF FIGURES
FIGURE 5-1: MULTI-CAVITY MOLD WITH UNBALANCED FILL........................................................................109
FIGURE 5-2: PRESSURE PROFILES FOR SMALL AND LARGE CAVITIES WITH CONVENTIONAL MOLDING ..........110
FIGURE 5-3: PRESSURE PROFILES FOR SMALL AND LARGE CAVITIES WITH DYNAMIC FEED ..........................111
FIGURE 5-4: EFFECT OF CAVITY PRESSURE SLOPE ON FLOW RATE THROUGH VALVE ....................................112
FIGURE 5-5: MULTI-GATED PART .................................................................................................................115
FIGURE 5-6: PRESSURE PROFILES FOR 3MM MULTI-GATED PART ..................................................................115
FIGURE 5-7: EFFECT OF CAVITY PACK PRESSURE ON LINEAR SHRINKAGE .....................................................117
FIGURE 5-8: RANGE OF CAVITY PRESSURE RESPONSES IN CONVENTIONAL MOLDING ...................................121
FIGURE 5-9: RANGE OF CAVITY PRESSURE RESPONSES WITH DYNAMIC FEED ..............................................122
FIGURE 5-10: EFFECT OF INPUT NOISE ON FILLING STAGE WITH CONVENTIONAL PROCESS ...........................123
FIGURE 5-11: EFFECT OF INPUT NOISE ON FILLING STAGE WITH DYNAMIC FEED ..........................................124
FIGURE 5-12: EFFECT OF INPUT NOISE ON PACKING STAGE ...........................................................................125
FIGURE 5-13: EFFECT OF INPUT VARIATION ON PART DIMENSIONS ...............................................................127
FIGURE 5-14: EFFECT OF SMALL DEVIATIONS IN TRANSITION TIME ON CAVITY PRESSURE RESPONSE ...........134
FIGURE 5-15: EFFECT OF VALVE INTERFERENCE ON CAVITY PRESSURE ........................................................136
1
TABLE OF TABLES
TABLE 5-1: PART DEFECTS CAUSED BY FLOW RATES IN THE FILLING STAGE .................................................109
TABLE 5-2: PART DEFECTS DETERMINED BY PRESSURES IN THE PACKING STAGE .........................................114
TABLE 5-3: SOURCES OF VARIATION IN INJECTION MOLDING .......................................................................118
TABLE 5-4: EXPERIMENTAL DESIGN FOR PROCESS CONSISTENCY .................................................................120
TABLE 5-5: CONSISTENCY RESULTS FOR CONVENTIONAL PROCESS ..............................................................128
TABLE 5-6: RANGE OF CAVITY PRESSURE SLOPES FOR 2 MM AND 3 MM WALL THICKNESS ...........................129
TABLE 5-7: ACHIEVABLE PACK PRESSURE DIFFERENTIAL ...........................................................................130
TABLE 5-8: RESPONSE TIMES OF SEVERAL CONTROL ELEMENTS ..................................................................133
START OF CHAPTER:107
2
Chapter 5: Experimental Validation of Process Capability
Chapter 5: Experimental Validation of
Process Capability
Dynamic Feed’s ability to control multiple cavity pressures was validated in

Chapter 4. This chapter demonstrates the benefits of such an ability, comparing
Dynamic Feed Control’s process capability to the conventional injection molding
process. First, the process flexibility enabled by Dynamic Feed Control is
demonstrated in the filling and packing stages. Then, the process consistency is
verified by experimental investigation of the effects of input noise upon the
consistency of the molding process and critical properties of the molded product.
Finally, the process limitations are identified for consideration in future research.
Once the capability of the process is known, a methodology for using the
process in complex part design will be presented in Chapter 6.
PROCESS FLEXIBILITY
In conventional injection molding, the process dynamics and part

properties are largely determined by the geometry of the mold and feed system.
Changes in machine settings – melt temperatures, ram velocity profile, and
injection pressure profile, for example – are transmitted through the runner
system to the entire part during manufacture. This limits the process’ ability to
selectively improve the part properties. Often, re-tooling the mold steel is
required to achieve the process dynamics which are necessary to obtain the
desired part properties.
107
Process flexibility is achieved by two fundamental mechanisms of the

Dynamic Feed process. First, the multiple valves enable the process
dynamics in each area of the cavity to be controlled independently.
Shrinkage in one local area of the part, for instance, can be altered without
affecting properties in other areas of the part. Second, closed loop cavity
pressure control decouples the process dynamics of the filling and packing
stages. In conventional molding, for instance, it is not possible to tool a mold to
obtain fast flow rates in one area of the mold followed by low packing pressures
while maintaining higher pack pressures elsewhere in the cavity – the large
runner diameters which prove useful in the filling stage are detrimental in the
packing stage. The removal of this limitation with Dynamic Feed greatly
increases the process flexibility.
Filling Stage Flexibility
Hesitation, jetting, flash, unbalanced filling, knit-line location, orientation,

and race-tracking are defects which are determined predominantly by the flow
rates in the filling stage. Table Chapter 5: -1 provides a description and cause of
each of these part defects. Filling stage flexibility is desirable to control the flow
rates in different areas of the cavity to achieve the desired process dynamics and
part properties without re-tooling mold steel.
108
Table Chapter 5: -1: Part defects caused by flow rates in the filling stage
Defect Description Cause Dynamic Feed Solution
Hesitation Discolored Low flow rates cause cyclic Modulate flow rate to provide
bands on cooling, then advancement of constant velocity of melt front
surface of part the melt during filling during filling
Jetting Discolored High flow rates in a thin to Reduce flow rate just prior to
swirls near thick region cause rupture melt front reaching area of
gates and thin and jetting of the melt front jetting
sections
Flash Film at edge of High flow rates force molten Reduce flow rates in areas
part plastic into very thin sections where flash is prevalent
Unbalanced Part warpage, Uneven flow rates cause one Route greater percentage of
fill burn marks, or area of the mold to fill early flow to areas where short
short shots while other area remains shots are occurring and
unfilled reduce pressure in over-
packed areas
Knit-lines Thin, visible line Occurs where cool melt fronts Meter flow to locate knit-lines
along surface of meet in acceptable areas or
part sequence gate opening
Orientation Anisotropic part Stress tensors dependent Route direction of flow to
properties, upon melt shearing and achieve desired orientation
warpage elongation during flow and vary flow rates to control
magnitude.
Race- Burn marks in Flow races around the thick Reduce melt flow in area
tracking center of part, lip of a thinner center section, leading to race-tracking and
circular knit- entrapping air as center increase flow rates in center
lines section fills of part.
For instance, Figure Chapter 5: -1 illustrates a multi-cavity mold which

would result in an unbalanced fill.
Figure Chapter 5: -1: Multi-cavity mold with unbalanced fill
If conventional injection molding were utilized with this design, the proper
manufacture of the small part at right would result in a short shot of the larger
109
cavity, as indicated by the dashed line in the figure. To manufacture the large
cavity, a molder would resort to higher filling speeds and pressures which results
in the immediate filling and over-packing of the smaller cavity as shown by the
cavity pressure history in Figure Chapter 5: -2. The conventional molding
process resulted in a filling of the small cavity within the first quarter second of
filling, followed by prolonged over-packing with cavity pressures approaching 80
MPa (12,000 psi) while the larger cavity continued to fill.
A tool designer could iteratively re-design the feed system to achieve a

uniform filling for the two cavities, but would be unable to compensate for packing
pressures or other material properties. As such, this type of mold design is rarely
utilized in industry.
80
70
60
50
40
30
20
10
0
0 1 2 3 4 5
Time (sec)
Figure Chapter 5: -2: Pressure profiles for small and large cavities with conventional
molding
Dynamic Feed Control was employed to regulate the flow rates in the
filling stage to avoid the unbalanced filling of this multi-cavity mold. Figure
110
Chapter 5: -3 is a plot of the cavity pressure histories for the small and large
cavity. A steep slope of cavity pressure was used to induce increased flow rates
for the large cavity and produce a full part. A time delay and lesser slope of
cavity pressure was used to avoid over-packing of the smaller cavity. Moreover,
the pack pressures of each part are specified independently, with benefits which
will be discussed in the next section. Such manufacture of these parts has never
before been accomplished.
90
80
70
60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8
Time (sec)
Figure Chapter 5: -3: Pressure profiles for small and large cavities with Dynamic Feed
To assist the product designer and process engineer in utilizing this

process flexibility, it is necessary to establish the relationship between the slope
of cavity pressure and the melt flow rate. To determine this relationship, parts
were molded with Dynamic Feed at many different slopes of cavity pressure for
the single (rectangular) cavity mold insert shown in Figure 4-4 on page 90. For
each part, the time was recorded when the melt first registered on the pressure
transducer near the entrance of the cavity, t1. The time was then recorded when
the melt registered on a pressure transducer farther in the cavity, t2. Knowing
111
these times and the cavity geometry, the volumetric flow rate was determined for
each cavity pressure slope:
L⋅w⋅h
Q= (5-1)
t 2 − t1
Figure Chapter 5: -4 graphs the relationship between cavity pressure

slope and flow rate for eighty different molding trials and two different wall
thicknesses. For a given wall thickness, a higher slope will generate a
proportionally higher flow rate during filling. This is consistent with the analytical
relation for Newtonian flow:
h3 ∂ p
m& = 12 ⋅ w ⋅ ρ ⋅ (5-2)
η ∂t
60
3 mm
50
Flow Rate (cc/sec)
40
30
2 mm
20
10
0
0 10 20 30 40
Figure Chapter 5: -4: Effect of cavity pressure slope on flow rate through valve
For a given pressure slope, the mass flow rate will vary significantly with
wall thickness due to the varying flow resistance. According to equation (5-2), a
112
2 mm wall thickness imposes roughly 3.3 times ( ( ) ) the flow resistance of a 3

3
2
3
mm wall thickness. This is again consistent with the experimental findings: the
slopes of the 3 mm and 2 mm linear fits are 3.5 cc/MPa and 1.5 cc/MPa,
respectively – a factor of 2.3. There is some deviation between theoretical and
experimental results attributable to the non-Newtonian viscosity of the polymer
melt. At thinner wall thicknesses, shear rates are higher which produces a lower
viscosity entering the power law region (shown in Figure 3-8). This lower
viscosity produces a lower slope of cavity pressure than theoretically predicted
assuming Newtonian behavior.
According to equation (5-2), the curves should pass through the origin, i.e.
a zero flow rate occurs at a pressure slope of zero. In actual molding there is a
minimum pressure slope, greater than zero, required to propel the melt front to
the end of flow before cooling causes the melt to solidify, i.e. there is some
minimum flow rate required so that heat transfer due to convection of the hot melt
overcomes heat transfer due to conduction to the cooling lines. According to this
data, those minimum pressure slopes are 8 MPa/sec and 18 MPa/sec for the 3
mm and 2 mm wall thicknesses, respectively.
Packing Stage Flexibility
Sink, flash, shrinkage, warpage, and residual stress are other part defects
that are affected by cavity pressures in the packing stage. Table Chapter 5: -2
describes these part defects and their origins. By controlling the cavity pressure
in each area of the cavity independently, Dynamic Feed Control enables the
flexibility to selectively alter part properties without re-tooling mold steel.
113
Table Chapter 5: -2: Part defects determined by pressures in the packing stage
Defect Description Cause Dynamic Feed
Solution
Sink Discoloration Higher volumetric Increase pressure in
opposite ribs shrinkage causes inflection thicker regions
of surface opposite ribs
Flash Film at edge High pressures cause mold Reduce pressure in
of part deflection and force molten areas prone to flash
plastic into opened crevices
Shrinkage Deviation of Volumetric shrinkage Increase pressure
dimensions causes molded part where reduced
from tool steel dimensions to shrink shrinkage is desirable
Warpage Out of plane Non-uniform volumetric Specify pressure
part distortion shrinkages causes non- distribution to minimize
linear, out of plane non-uniform shrinkage
deformations
Residual Dimensional Solidification locks in stress Profile pack pressure
Stress creep, fields caused by pressure, decay to reduce
performance orientation, and cooling residual stresses while
degradation cooling
The last section illustrated the process flexibility which enabled the proper
filling of an unbalanced multi-cavity mold. Referring back to the example of
Figure Chapter 5: -1, the smaller cavity was filled instantly and over-packed with
pressures exceeding 80 MPa (12,000 psi) with conventional injection molding.
By utilizing Dynamic Feed, however, the pressure history in Figure Chapter 5: -3
was attained. Not only was the smaller cavity filled more slowly, but packing
pressures were delivered at a specified level of 28 MPa, 25 MPa below the
packing pressures of the larger cavity. This separate control of cavity filling
pressures and packing pressures is not possible in conventional injection
molding.
In a multi-gated part, shown in Figure Chapter 5: -5, the goal of the

Dynamic Feed control system is to control the cavity pressure distribution
throughout the part during the packing stage. This can be achieved by utilizing
Dynamic Feed to specify varying levels of pressures at P1 and P2 simultaneously.
114
Gates Pressure Transducers
P3
P1 P4 P2
Figure Chapter 5: -5: Multi-gated part
Figure Chapter 5: -6 plots the cavity pressure response for the 220 x 50 x 3 mm
rectangular plaque drawn in Figure Chapter 5: -5. The pressure profiles show
that packing pressures of 25 MPa and 32 MPa were imposed at the ends of the
part – a 7 MPa pressure differential.* The cavity pressures were coupled,
constraining the maximum differential between gates. Subsequent testing
indicated that the maximum pressure differential between P1 and P2 was
approximately 13 MPa. For example, Dynamic Feed was incapable of providing
a cavity pressure of 20 MPa at P1 and a cavity pressure of 50 MPa at P2.
35 2
30
25
1
20
15
1
10
5 2
Figure Chapter 5: -6: Pressure profiles for 3mm multi-gated part
*
As a side note, Figure Chapter 5: -6 also re-iterates the decoupling of the filling and packing
stages. In this figure, trace two (2), the area with lower flow rates, was also the area with higher
specified packing pressures.
115
Once the polymer fills the mold, the cavity pressures are coupled since the
melt may flow from an area of high pressure to influence areas of lower pressure.
This will limit the maximum pressure differential which can be maintained across
the part during the packing stage. In fact, the magnitude of these cavity pressure
gradients is dependent upon the melt viscosity, solidification dynamics, wall
thickness, and distance between gates. For the same part geometry shown in
Figure Chapter 5: -5 with 2 mm wall thickness, however, the pressure differential
was found to increase to 30 MPa. The ramifications of this limitation will be
discussed in subsequent sections.
The experimental validation also showed that the level of packing

pressures in the cavity may be widely specified, from 0 MPa to 100 MPa (14,500
psi). Higher levels of packing pressures are not commonly used in industry and
were not examined in the experimental portion of this research.
Dynamic Feed Control may be used to selectively alter specific part

properties. For instance, cavity pressures in one area of the part may be
increased to reduce the amount of sink. Alternatively, the cavity pressures in
another area of the part may be reduced to eliminate flash along that adjacent
edge of the part. Or, the packing pressure distribution across the part may be
manipulated to bring the part within tolerances or minimize warpage and residual
stress. Figure Chapter 5: -7 is a graph of the relationship between cavity
pressure and linear shrinkage for the 3 mm, multi-gated rectangular plaque.
116
1.2
1.1
0.9
Shrinkage (%)
0.8
0.7
Assumed level of
shrinkage used in
0.6 mold design
0.5
0.4
0.3
0 10 20 30 40 50 60
Figure Chapter 5: -7: Effect of cavity pack pressure on linear shrinkage
The gray box in Figure Chapter 5: -7 indicates the assumed range of

linear shrinkage, data supplied by the material supplier for use in estimating tool
steel dimensions.93 It is not possible to precisely predict the exact mold
shrinkage so tool designers have historically used conservative shrinkage
estimates and relied upon multiple tooling iterations to achieve desired
tolerances. Difficulties do not usually arise when the whole part shrinks uniformly
– the level of pack pressures may be increased or decreased to bring the part
within tolerance. The primary difficulty arises when unforeseen process
dynamics impose non-uniform cavity pressures, uneven cooling, or molecular
orientation which cause non-uniform shrinkage. In these instances, no level of
process conditions will result in adequate part properties, so tool changes are
utilized.
With Dynamic Feed, an estimate of 0.6% shrinkage may be used as an

initial estimate for tooling. The process flexibility enables the molded part
dimensions to meet design tolerances by varying the cavity pressures in the
117
molding process without re-tooling. The entire cavity pressure distribution can be
easily increased or decreased, or the shape of the pressure distribution can be
varied across the part. As previously indicated, melt flow in the packing stage
limits the pack pressure differential to 13 MPa/200 mm for a 3mm wall thickness.
This pressure range corresponds to a ± 0.15% dimensional freedom across this
particular part – enough to compensate for unexpected material and process
variation without re-tooling. Similar methods may be utilized to adjust for sink,
flash, warpage, and residual stress.
PROCESS CONSISTENCY
Product inconsistencies among a batch of molded parts is most often

assigned to lot to lot variation in material properties. However, material variation
is only one source of variation – many others exist which are not usually
considered as listed in Table Chapter 5: -3.
Table Chapter 5: -3: Sources of variation in injection molding

Type of variation Source of variation
Material properties Viscosity, density, etc.
Machine to machine repeatability Machine dynamics, tuning, wear
Model to model repeatability Controls tuning, barrel and hydraulic wear
Shot-to shot repeatability Shot size, velocity, switchover, pressure,
etc.
Process engineer repeatability Selection of set-points, set-up, etc.
Operator repeatability Cycle time, parts handling, etc.
Environment repeatability Temperature, humidity, etc.
White noise Small variations in all the above
To evaluate the process consistency of both the conventional and

Dynamic Feed process, a design of experiments was utilized to simulate the
commercial production of a typical part. In a typical production run of 50,000 to
100,000 parts per month, three to five identical molds would be tooled and sent
to different molders where the parts would be molded. Each molder is given an
identical sheet of process parameters with which to mold the products, however,
118
each molder is free to decide which of their machines is most suitable for molding
the product.
Since molding machines have different screw and barrel designs – not to
mention different clamp tonnage, shot size, and injection pressure capacities –
the process behavior will vary significantly even though the same ‘process
conditions’ are used on every machine. Moreover, molders are given some
latitude to optimize the process parameters of their machine for use with the
given mold. For instance, mold open time, cushion size, barrel temperature
profile, and switchover points are variables that the molder must determine
independently. (Or, if they used ‘standard’ settings across molders and molding
machines, the machine-specific behavior might result in even more severe
process inconsistencies.)
Table Chapter 5: -4 lists the half-factorial design of experiments utilized for

this investigation (an L8 with four variables of two levels).94 The levels do not
represent the boundaries of the process window for this design, rather they are a
very small subset of that process window. The levels of each of these process
conditions was chosen to emulate the range of noise which would be
encountered in the scenario described above. For instance, a ± 5°C fluctuation in
melt temperature represents the variation in actual melt temperatures across
different molding machines and molders. The ± 8°C range of mold temperatures
might reflect variation in water flow rates through the tool which are not specified.
Similarly, the levels of injection speed and hold pressure shown in the table are
indicative of the machine to machine variations in barrel, hydraulic, and controller
systems.
119
Table Chapter 5: -4: Experimental design for process consistency

Run Melt Mold Injection Pack
Number Temperature Temperature Speed Pressure
1 277 °C 57 °C 35 cc/sec 53 MPa
2 277 57 41 62
3 277 74 35 62
4 277 74 41 53
5 288 57 35 62
6 288 57 41 53
7 288 74 35 53
8 288 74 41 62
For each molding trial involving a temperature change, the set-points were
entered into the molding machine controller – all machine parameters not listed
in Table Chapter 5: -4 were left unaltered. The machine was left idle for thirty
minutes during which time the melt and/or mold temperatures would equilibrate.
Afterwards, twenty shots were molded with a consistent cycle time of 29 to 31
seconds. All these parts were discarded. Then six parts were molded utilizing
the conventional process followed by another six parts utilizing Dynamic Feed.
The parts were measured one week later in a 21°C environment using a dial
caliper with an accuracy of 0.006 mm (0.0003 in). All process data and part
measurements were analyzed using RS/Explore, a statistical analysis package
available under license from Bolt Bareneck and Newmann, Inc. (New York,
NY).95
120
Cavity Pressure Response
The cavity pressure response of the final run of each molding trial is
graphed in Figure Chapter 5: -8 for the conventional injection molding process.
Given the relatively small range of input noise, the magnitude of variation is quite
surprising. There is variation in the slope of cavity pressure between 1 and 2
seconds, the nature of the transition between 2 and 4 seconds, as well as the
final level of pack pressures near 8 seconds.
70
60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8
Time (sec)
Figure Chapter 5: -8: Range of cavity pressure responses in conventional molding
Figure Chapter 5: -9 displays the cavity pressure response from the last
run of each molding trial utilizing Dynamic Feed, as well as the input pressure
profile indicated by the dotted line. It is clear that the Dynamic Feed delivers
more consistent cavity pressures in the presence of variation with the following
characteristics:
• the cavity pressure slopes are very uniform during the filling stage,
121
• the transition between the filling and packing stages varies slightly due to
valve timing imprecision, and
• the pack pressures are very flat and consistent.
70
60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8
Time (sec)
Figure Chapter 5: -9: Range of cavity pressure responses with Dynamic Feed
Filling Stage Consistency
The average slope of cavity pressure was estimated for each run in each
molding trial. This was evaluated by first passing the raw cavity pressure data
through a second order low-pass digital Butterworth filter to reduce the effects of
process noise. Then, the start of filling was identified as the point at which the
cavity pressure exceeded 0.5 MPa, normally around 1.0 seconds. Finally, a
linear regression was performed over the next 1.0 seconds of data to arrive at a
mean slope for cavity pressure.
The cavity pressure data for each run of the half-factorial design (Table
Chapter 5: -4) is presented in Appendices B-1 and B-2. A regression analysis
was then performed over the entire set of data. Figure Chapter 5: -10 lists the
122
estimated effect of each noise parameter on the slope of cavity pressure. The
dark vertical line in the center of each bar represents the estimate while the
smaller vertical bars represent the 95% confidence interval. Any confidence
interval crossing zero indicates that noise parameter has little impact on the
system response. For instance, increasing the melt temperature is estimated to
increase the slope of cavity pressure about 2.5 MPa/sec while increasing the
mold temperature is estimated to have little effect.
Increase Melt
Temperature
Input Noise
Increase Mold
Temperature
Increase Ram
Speed
Increase Pack
Pressure
-5 0 5 10 15 20
Estimated Effect on Cavity Pressure Slope (%)
Figure Chapter 5: -10: Effect of input noise on filling stage with conventional process
The results shown in Figure Chapter 5: -10 seem reasonable. Increasing

the speed of the ram would increase the volumetric flow rate into the cavity
which, in turn, should increase the slope of cavity pressure. Moreover,
increasing the melt temperature reduces the melt viscosity which in turn reduces
the pressure drop through the barrel, nozzle, and feed system and will result
again in higher cavity pressures. Increasing the pack pressure should not effect
the filling stage dynamics (unless the machine’s controller responds to packing
pressure inputs in the filling stage control algorithm.)
Figure Chapter 5: -11 graphs the effect of the same noise parameters with
Dynamic Feed.
123
Increase Melt
Temperature
Input Noise
Increase Mold
Temperature
Increase Ram
Speed
Increase Pack
Pressure
-5 0 5 10 15 20
Estimated Effect on Cavity Pressure Slope (%)
Figure Chapter 5: -11: Effect of input noise on filling stage with Dynamic Feed
As anticipated, the effects of the noise parameters have been significantly

reduced by Dynamic Feed since it utilizes closed loop cavity pressure control.
Interestingly, the input noise affects the Dynamic Feed process trends in the
same manner as the conventional process, albeit to a lesser degree. Increasing
the melt temperature and ram speed increase the slope of cavity pressure,
though not to the same extent as in the conventional process. In this regression,
mold temperature was found to have a slight, though distinct, effect on slope
which is logical – increasing the mold temperature reduces the amount of heat
transfer during filling, thereby keeping the melt fluid and the cavity pressures
lower.
For conventional molding, these results and equation (5-2) indicate that
the flow rates in the filling stage may vary 15% during an application’s production
run. Flow rate has a dominant effect upon the level of defects such as flash, knit-
lines, and over-packing as well as some part properties related to orientation
(see Table Chapter 5: -1). As such, applications which are difficult to mold may
experience lower process yields or systematic defects when subject to typical
process variation.
124
Dynamic Feed has a significantly better process capability. First, the

flow rates in the filling stage can be immediately and selectively controlled
to move the molding process away from potential defects. In the
conventional molding process, this previously required design and tooling
changes. Second, this experimental analysis has shown that Dynamic Feed
can control the slope of cavity pressure about four times better than the
conventional molding process. With this capability, Dynamic Feed enables
the molding of more robust and consistent plastic parts.
Packing Stage Consistency
The packing pressure was identified as the pressure at 7.0 seconds, one
second before the end of the packing stage. The half-factorial design table
(Table Chapter 5: -4) was again augmented with the cavity packing pressure of
each run in each molding trial, as listed in Appendices B-1 and B-2. A
regression analysis was then performed over the entire set of data using
RS/Explore to determine the net effect of each noise parameter on the cavity
pack pressure. The results for both the conventional molding and the Dynamic
Feed process are shown in Figure Chapter 5: -12.
Increase Melt
Temperature
Key
Input Noise
Increase Mold
Conventional
Temperature Dynamic Feed
Increase Ram
Speed
Increase Pack
Pressure
-10 -5 0 5 10 15 20
Estimated Effect on Cavity Pack Pressure (%)
Figure Chapter 5: -12: Effect of input noise on packing stage
125
In the figure, the conventional molding process is indicated by the dashed

horizontal lines while the Dynamic Feed process is shown with the solid lines.
The trends are as expected. Increasing the melt temperature reduces the melt
viscosity and pressure drop through the feed system which causes an increase in
pack pressure. Mold temperature has no effect in the Dynamic Feed process,
though there is a significant negative correlation in the conventional process.
Increasing the speed has little effect with Dynamic Feed but a significant positive
effect on the packing pressure in the conventional process, likely due to a larger
pressure spike at 3.0 seconds which endures throughout the packing phase (see
Figure Chapter 5: -8). Finally, increasing the pack pressure produced increases
in the cavity pack pressure for both the Dynamic Feed and conventional molding
process, as expected. For every noise parameter, Dynamic Feed significantly
reduced the amount of process variation as could be expected from a closed
loop process. Moreover, the process response could be further improved by
tightening the tolerance limits which were set at 4 MPa (600 psi) as shown in
Figure 4-9.
Consistent control of the packing stage is necessary to produce consistent

parts from molder to molder as well as shot to shot. Any variation in the process
dynamics will be transmitted directly to the part properties, resulting in excess
sink or flash, inappropriate shrinkage or warpage, or end-use performance
degradation. To investigate the effect of process variation on final molded part
properties, the width of each molded part was measured at the location of the
pressure transducer. Figure Chapter 5: -13 graphs the regression analysis
results showing the effect of the noise parameters on linear dimensions. As
before, the dashed and solid lines indicate the conventional and Dynamic Feed
processes, respectively.
126
Increase Melt
Temperature
Key
Input Noise
Increase Mold
Temperature Conventional
Dynamic Feed
Increase Ram
Speed
Increase Pack
Pressure
-0.50 -0.25 0 0.25 0.50 0.75
Estimated Effect on Part Dimensions (% Length)
Figure Chapter 5: -13: Effect of input variation on part dimensions
Typically, ± 0.2% is a standard commercial-grade tolerance.96 For

example, if a dimension was specified as 10 cm, the tolerances for that
dimension would normally be specified as 10.00 ± 0.020 cm. For such a
tolerance, any one of the noise sources in this experimental validation could
significantly reduce the production yield in the conventional injection molding
process since the corresponding effects on part dimension are greater than
0.2%. Dynamic Feed should produce more consistent parts with fewer defects
by reducing the effects of variation.
As a final check on process capability, the dimensional results listed in

Appendices B-1 and B-2 were analyzed for consistency. Table Chapter 5: -5
summarizes the consistency results from measurement of the molded parts.
Note that the standard deviation from the average (expressed as a percent of
length) for Dynamic Feed is roughly one-third of the conventional process. If
industry’s tolerance specification is defined at 0.2%, then the process capability
of the conventional and Dynamic Feed process can be calculated with the results
expressed in the table.*,97
*
A Cp of 1.0 represents a centered process with three standard deviations on either side,
corresponding to a production yield of 99.73%. While the molding community generally accepts
yields of 95%, higher Cps are desirable. A Cp of 0.5 represents a yield of 86% and is not
acceptable.
127
Table Chapter 5: -5: Consistency results for conventional process

Performance Conventiona Dynamic
Measure l Molding Feed
Standard deviation from average, % 0.064 0.020
Maximum over average, % 0.106 0.026
Minimum under average, % 0.148 -0.024
Process capability, Cp , for industry
standard 0.2% tolerance
0.52 1.67
specification
128
PROCESS LIMITATIONS
In this section, several sources of limitations with the Dynamic Feed

process are reviewed. These arise from two sources: first, the capabilities of
Dynamic Feed are still governed by the physics of the injection molding
process. As such, there are fundamental boundaries for the extent of flexibility
and consistency of the Dynamic Feed process. Second, the addition of flow
metering elements in the feed system and supporting hydraulic sub-
systems are new sources of process noise and instability – these may
negatively impact the process capability and point to design modifications in
future work.
Limitations from Process Physics
Filling Stage
During the filling stage, the slope of cavity pressure governs the melt front
advancement and subsequent part properties listed in Table Chapter 5: -1. As
such, the design or process engineer would like to arbitrarily select the cavity
pressure profile and flow rate at each gate. There are limitations to the minimum
and maximum slopes which may be profiled, however, as shown in Table
Chapter 5: -6. For a specific wall thickness and melt resistance, the cavity
pressure is approximately proportional to the volumetric flow rate. Thus, the
maximum slope of cavity pressure is limited by the flow rate capacity of the
molding machine. At lower flow rates, cooling of the melt may result in significant
increases in viscosity. To maintain a minimum pressure slope, the control
system further reduces flow rates, which results in further cooling, and eventual
solidification, of the melt which results in a short shot.
Table Chapter 5: -6: Range of cavity pressure slopes for 2 mm and 3 mm wall thickness
2 mm Wall 3 mm Wall
Maximum Pressure 25 MPa 46 MPa
129
Slope
Minimum Pressure 7.5 MPa 18 MPa
Slope
Knowledge of these limitations is critical to the design engineer when considering

the manufacturability of a candidate design. For instance, a design may require
a knit-line to be placed outside of a visible surface. The flow rates through each
gate may be chosen to fulfill this requirement, subject to the limitations listed in
Table Chapter 5: -6 and Figure Chapter 5: -4.
Packing Stage
The goal of Dynamic Feed is not only to control the level of packing
pressure at a discrete point in the cavity but also to control the distribution of
packing pressures across the cavity as well. The magnitude of cavity packing
pressures is limited only by the pressure capacity of the molding machine.
However, the gradients of the cavity pressure in the packing stage are
constrained since molten plastic will flow from an area of high pressure to one of
lower pressure through the part. Such flow during the packing stage is minimal,
driven only by the compression and solidification of the melt. As such, the
magnitude of the cavity pressure gradients is dependent upon the melt viscosity,
solidification dynamics, wall thickness, and distance between gates. Table
Chapter 5: -7 lists the maximum pack pressure differential which can be
maintained for the 220 x 50 rectangular plaque at 2 mm and 3 mm wall
thicknesses.
Table Chapter 5: -7: Achievable Pack Pressure Differential

2 mm Wall 3 mm Wall
Pack Pressure 29.5 MPa 12.9 MPa
Differential
Confidence Interval 1.6 MPa 2.3 MPa
This data is critical to the design or process engineer trying to manipulate the
pack pressure distribution to obtain the desired part properties, as an example in
130
the next Chapter will demonstrate. For a specific application, the wall thickness,
distance between gates, and material properties will determine the maximum
pack pressure differential which can be maintained – commercial flow analyses
can provide good estimates of the attainable cavity pack pressure distribution.98
Temperature Effects
Part properties are determined by both the pressure and temperature

history in the cavity during the molding process. Cavity pressure control was
enabled by the addition of flow metering elements with a response time of 0.050
sec – on the order of the process dynamics being controlled. However,
temperature control is inherently a much slower process – process response is
on the order of seconds or minutes – which greatly limits the possibility of
controlling temperatures to substantially impact the part properties during the
molding cycle; the temperature history of the part is determined solely by the
process physics once the melt and mold temperatures have been set. As such,
structural properties (part dimensions, warpage, knit-line strength) and cosmetic
properties (gloss, sink, blush) may vary significantly with melt and mold
temperatures.
It may be possible to compensate for some effects of temperature

deviation by altering the cavity pressure distribution. If models were developed
relating the local temperature and pressure history to part properties, then the
effect of measured deviations in melt temperatures could be estimated which, in
turn, could be compensated by altering the cavity pressure distribution. For
instance, a 10°C change in mold temperature might increase the mold shrinkage
by 0.1% which could be offset by an increase of cavity pressure of 12 MPa for
the reduction in mold shrinkage by 0.1%.
Unfortunately, there are some part properties (surface gloss, for instance)
which are mostly determined by mold temperature for which Dynamic Feed, by
131
itself, would be ineffective in controlling. Kim et. al. utilized small heating/cooling
plates within the mold steel to selectively control the magnitude and distribution
of mold temperatures during the molding process.99 Together, control of the
temperature and pressure dynamics in the mold cavity would provide attainment
of nearly any molded part property.
Limitations from System Design
As a prototype manufacturing process, the present system has functioned

well, enabling verification of the control strategy and process concept. This
system has greatly enhanced the capability and flexibility of the injection molding
process as the results in this chapter have shown. The Dynamic Feed control
elements, however, were installed on a molding machine designed for
conventional use. This resulted in some interesting process consistency issues
due to the timing and hydraulic system design specifics. Moreover, there was
significant interaction between the multiple valves of hydraulic feed, interactions
which were assumed insignificant throughout development of the control
strategy.
Due to these effects, the process consistency of the adaptive algorithm

was reduced. The closed loop control system was utilized to actively monitor for
process inconsistencies and indicate when parts should be discarded. As such,
the fraction of the 2000 parts which were molded with Dynamic Feed during the
experimental validation and met the specifications indicated in Figure 4-9 was
approximately 70%. The following discussion highlights the sources of
inconsistency and points to areas for future system design.
Response Time
At the beginning of the molding cycle, the injection molding machine

sends a signal to the control system indicating the start of injection. The control
system opens the valve stems, the molding machine accelerates the ram, and
132
the melt begins to flow into the cavity. Near the end of the filling stage, the valve
stems are moved to a semi-closed position for control of the packing stage. This
work has shown that the timing of switchover is critical and must be adaptively
fine-tuned to within 0.050 sec to obtain a smooth transition.
Table Chapter 5: -8: Response times of several control elements

Control element Source of delay Response time
Hydraulic spool position Actuation of spool in valve 0.020 sec
Pressure transducers Transmission of pressure through 0.020 sec
long mercury capillary
Hydraulic signal amplifiers Amplification of 24 DC to 2 V, 10 0.010 sec
amp
Control system Selected in system development 0.010 sec
Table Chapter 5: -8 lists the response times of several process control

elements. While each of these devices has a response time less than 0.020 sec,
the system response time (pressure transducer Æ control system Æ signal
amplifier Æ spool position Æ valve stem actuation) is on the order of 0.050 sec.
This indicates that small deviations in process consistency, such as start of
injection or ram acceleration or valve stem positioning, may result in significant
deviations in the transition between the filling and packing stages. While not an
extensive problem, the system response times did result in a reduction in the
consistency of cavity pressure profiles. Figure Chapter 5: -14 plots the cavity
response for transition times ranging from -0.100 seconds to +0.100 seconds
away from the nominal (centered) transition time. An improved design would
reduce the system response time.
133
30
25

20
15
10
0
0 1 2 3 4 5 6 7 8
Time (sec)
Figure Chapter 5: -14: Effect of small deviations in transition time on cavity pressure
response
Hydraulic Pilot
Each of the valve stems was controlled by a proportional servo-valve

mounted on the hydraulic manifold of the molding machine. Surprisingly, the
pressure to the hydraulic manifold for the actuation of the valve stems is also the
supply pilot pressure to the molding machine’s hydraulic pump.100 The hydraulic
manifold is normally used for valve shut-offs, core pulls, or hydraulic slides –
activities which do not dynamically require fluid flow during the filling and packing
stages of injection molding.
With Dynamic Feed, however, the frequent opening and closing of the
valve stems requires continuous hydraulic flow throughout the filling and packing
stages. More specifically, the transition between the filling and packing stages
requires a significant amount of flow to quickly move the valve stem from a
mostly open to a mostly shut position, just at the time when the molding machine
is also transitioning from the filling to the packing stage. This results in an
instability as the valve stem actuation reduces the molding machine’s pilot
134
pressure which causes a brief, though noticeable, decay in the injection

cylinder’s hydraulic pressure. Dynamic Feed Control’s adaptive gain scheduling
compensates for this effect automatically, though likely requires more iterations
(shown in Figure 4-15) to converge to an acceptable response. Moreover, small
instabilities around the transition time, as discussed in the previous section, can
lead to unacceptable process dynamics. In future designs, a hydraulic
accumulator between the valve stems and the injection unit would eliminate this
problem.
Valve Interaction
The control system strategy presented in Chapter 3 was developed for

one branch of a feed system, assuming negligible interaction between multiple
valves. As considered in model development, a volume of melt is displaced by
the movement of the valve stem when a valve stem is moved from an open to a
closed position. Some of this volume is forced into the cavity ahead of the valve
stem and results in a temporary increase in cavity pressure – this effect was
considered in control system development. However, the remaining portion of
the displaced volume of material is forced back into the feed system and results
in an increase in injection pressure which subsequently effects the pressure
dynamics at the other valves – this effect was not considered in control system
development.
Figure Chapter 5: -15 illustrates this undesired valve interaction. The

dark line is the normal cavity pressure response when a valve stem is moved
from 70% open to 30% open at 1.5 seconds and the other valve stem remains
stationary. There is an immediate slight increase in cavity pressure when some
of the melt in front of the valve stem is forced into the cavity. This is followed by
a slight decay in cavity pressure as the melt dynamics adjust to the increased
flow resistance. By 1.8 seconds, the dynamics are stable and the cavity
pressure begins to rise.
135
25
20

15
10
Key
5 Other valve stationary
Other valve opening
0
1.4 1.6 1.8 2 2.2
Time (sec)
Figure Chapter 5: -15: Effect of valve interference on cavity pressure
The dotted curve indicates the response when that same valve is moved
from 70% open to 30% open at 1.5 seconds and the second valve is opened
from 0% open to 100% open at the same time. In this case, the behavior at the
first valve is the same until 1.6 seconds when the second valve’s interaction
becomes noticeable. A vacuum has now been created in front of the second
valve. Moreover, the flow resistance through the second valve is much less than
that of the first valve. Both of these effects result in a lengthy decay and delay in
control of the cavity pressure at the first valve. Fortunately, the magnitude of this
interaction is not as significant during the molding process since the valve stems
are not normally manipulated to this extent. Future system designs could reduce
the valve interaction by minimizing the volume displaced by the valve stem.
SUMMARY
Experimental investigations have validated the process flexibility and

consistency of the Dynamic Feed concept. Results indicate that the flow rates
136
and packing pressures may be arbitrarily specified for a variety of part

geometries and wall thicknesses. The ability to control flow rate was
demonstrated to make possible the high quality production of a very unbalanced
multi-gated mold, previously before impossible with conventional injection
molding. Moreover, the flexibility to manipulate cavity packing pressures was
shown to bring a molded part within specified tolerances without re-cutting mold
steel.
An experiment was designed to investigate the effects of typical noise

during the commercial production of a molded part on a variety of machines at
different molders. The results indicated that the effects of these process
variations were reduced by 75% when compared to the conventional injection
molding process. Molded part dimensions were likewise affected. For parts
specified to conform to a conventional commercial grade tolerance of 0.2%, the
conventional molding process with noise was found to have a process capability
(Cp) of 0.52. Utilizing the Dynamic Feed process resulted in a Cp of 1.67.
Finally, this chapter identified various process and design limitations of

Dynamic Feed. Many of these limitations are due to process physics which are
inherent in injection molding and Dynamic Feed. Other limitations involving the
system design – including timing, hydraulic design, and valve design – were
found to reduce the process consistency and should be considered in future
system design.
137
93
Mold Shrinkage Estimates, GE Plastics Lexan™ Resin Design Guide, GE Plastics, Pittsfield, MA, p. 31,
(1987).
94
Box, G.E.P., Hunter, W.G., Hunter, J.S., Statistics for Experimenters, John Wiley & Sons, New York,
1978.
95
RS/Explore, Release 3.01, Sub-module of RS/1, Release 5.0.1, Used under license from Bolt Bareneck
and Newmann, Inc., 1994.
96
Beris, M. C., Standards and Practices of Plastics Custom Molders, SPI Plastics Engineers
Handbook of the Society of the Plastics Industry, 5th Edition, Van Nostrand Reinhold Co., New
York, 1991.
97
Farnum, N.R., Modern Statistical Quality Control and Improvement, Duxbury Printers, Belmont, California,
1994.
98
Wang, V.W., Hieber, C. A., Wang, K. K., Dynamic Simulation and Graphics for the Injection Molding of
Three-Dimensional Thin Parts, Journal of Polymer Engineering, v. 7, n. 1, pp. 21-45 (1986).
99
Kim, B., Low Thermal Inertia Injection Molding, Doctoral Dissertation submitted to the Massachusetts
Institute of Technology Department of Mechanical Engineering, 1983.
100
Personal conversation with Gary Stanton, Field Service Representative, Mitsubishi Heavy Industries,
Learson, Texas, Pleasanton, California (February 11, 1995).
138
TABLE OF CONTENTS
CHAPTER 6: DESIGN METHODOLOGY...........................................................................................137

PRODUCT DESIGN PROCESSES ...........................................................................................................138
Iterative Design .....................................................................................................................................138
Optimization ..........................................................................................................................................140
PROPOSED METHODOLOGY................................................................................................................143
Specifications ........................................................................................................................................145
Part Design ...........................................................................................................................................146
Mold Design ..........................................................................................................................................146
Evaluation .............................................................................................................................................148
Process Relations............................................................................................................................................... 150
Design Optimization.......................................................................................................................................... 155
Robustness.............................................................................................................................................156
A CASE STUDY ........................................................................................................................................157
Design....................................................................................................................................................158
Objectives and Constraints ................................................................................................................................ 161
Design Evaluation .................................................................................................................................162
One Center Gate, Constant Pressure.................................................................................................................. 162
One Center Gate, Optimal Pressure................................................................................................................... 164
Three Gates, Optimal Pressure .......................................................................................................................... 165
Two Gates, Optimal Pressure ............................................................................................................................ 166
PROCESS CONTROL FOR MULTIOBJECTIVE OPTIMIZATION.......................................................167
SUMMARY ...............................................................................................................................................170
TABLE OF FIGURES
FIGURE 6-1: COUPLING IN THE DEVELOPMENT OF INJECTION MOLDED PARTS ..............................................138
FIGURE 6-2: CLASSIC DESIGN TO PRODUCTION DEVELOPMENT PROCESS ......................................................140
FIGURE 6-3: HIERARCHY OF DESIGN RELATIONSHIPS ...................................................................................142
FIGURE 6-4: PROPOSED PRODUCT DESIGN METHODOLOGY ...........................................................................144
FIGURE 6-5: DESIGN EVALUATION ...............................................................................................................149
FIGURE 6-6: TYPICAL MOLDED PART AND SPECIFIED DIMENSIONS ...............................................................158
FIGURE 6-7: STOCHASTIC SHRINKAGE BEHAVIOR OF POLYCARBONATE .......................................................160
FIGURE 6-8: CENTER-GATED BOX ................................................................................................................163
FIGURE 6-9: DYNAMIC FEED CONTROL OF THREE GATES .............................................................................165
FIGURE 6-10: TWO GATES, OPTIMAL PRESSURES ..........................................................................................167
FIGURE 6-11: PROCESS CONTROL FOR MULTI-OBJECTIVE OPTIMIZATION .....................................................169
1
TABLE OF TABLES
TABLE 6-1: TYPICAL PRODUCT SPECIFICATIONS FOR A MOLDED PART .........................................................146
TABLE 6-2: TYPES OF SPECIFICATIONS .........................................................................................................147
TABLE 6-3: SPECIFIED DIMENSIONS FOR MOLDED PART ...............................................................................158
TABLE 6-4: PERFORMANCE OF ONE CENTER GATE AT CONSTANT PRESSURE.................................................163
TABLE 6-5: PERFORMANCE OF ONE CENTER GATE AT OPTIMAL PRESSURE ...................................................164
TABLE 6-6: PERFORMANCE OF THREE GATES AT OPTIMAL PRESSURES .........................................................166
TABLE 6-7: PERFORMANCE OF TWO GATES AT OPTIMAL PRESSURES ............................................................167
TABLE 6-8: PERFORMANCE OF THREE GATES AT NON-OPTIMAL PRESSURES .................................................170
2
Chapter 6: Design Methodology
As described in Chapter 1, the plastics industry is increasingly hard

pressed to improve performance, reduce cost, and shorten delivery time. Despite
numerous technological advances, commercial applications continue to
experience significant difficulties in plastic part production, delaying product time
to market while incurring excessive costs. Much of the difficulty arises from
thermoplastics’ complex material properties and unknown process variation –
these effects are not considered in the design and product development stage,
only later to arise and require product and tool changes to produce parts of
acceptable quality.
By directly controlling cavity pressures in a closed loop process, Dynamic

Feed Control removes the dependence of a part's fabrication from its geometric
form. Changes in design features, such as wall thickness or rib height need no
longer determine the pressure dynamics. The cavity pressure distribution
becomes largely geometry independent so that fabrication is an independent,
controlled variable. As presented in Chapter 5, Dynamic Feed Control provides
the process capability to vary the magnitude and distribution of cavity pressures
in the filling and packing stages. If the relationship between part fabrication and
part performance is at least partially known, then Dynamic Feed Control can be
used in production to directly improve the part’s performance without re-tooling.
This chapter describes design and production methodologies which

explicitly utilizes the extra degrees of freedom to ensure product quality. The
concept of design robustness is clearly demonstrated when stochastic variation
due to material or process variation in the production stage is modeled. This
methodology will then be utilized to investigate the robustness of different gating
scenarios for a four-sided housing with three critical dimensions.
137
PRODUCT DESIGN PROCESSES

Modeling of the design process has been a perennial topic for discussion
and research.101 However, one common view of the design process is that most
design is re-design, i.e. the generation of new solutions to problems which have
previously existed.102 As such, design is inherently an iterative process, both
during a specific product’s development cycle as well as between successive
product generations. Injection molded part design is a classic example of the
iterative process of detail design.
Iterative Design
“Once an initial design has been created, its performance is
tested or predicted by analysis. (Correct analysis is a critical support
needed for good design.) With the trial design's performance known, the
acceptability of the design can be judged. If it is acceptable, the task is
complete. If not, the design must be redesigned, the analysis for
expected performance repeated, and so on iteratively until a design is
accepted.” 101
A molded part is ordinarily one component in an assembled product,

designed and manufactured to deliver its share of functionality to the integrated
product. As such, the molded part must be designed with a set of functional
requirements and specifications, some implicitly understood, others explicitly
stated. Fulfilling these part requirements is difficult since there is significant
interaction between the part’s geometric form, molding dynamics, and end-use
performance. These relationships are shown in Figure Chapter 6: -1.103
Geometric
Form
Molding End-use
Dynamics Performance
Figure Chapter 6: -1: Coupling in the development of injection molded parts
138
For instance, a part might exhibit warpage due to varying wall thicknesses across
the part which have caused non-uniform shrinkage. To reduce the warpage, a
tool designer might re-cut the steel to utilize a single, nominal wall thickness.
However, this change in wall thickness could significantly vary the melt
orientation and/or cavity pressure distribution which might also result in warpage.
Additional design and tooling iterations would be required to achieve an
acceptable design.
These difficulties are compounded in technical applications with multiple

requirements, subject to process dynamics and limitations which are unknown to
the part designer. To overcome these difficulties, improved analysis techniques
have been developed to better predict part performance for candidate designs as
presented in Chapter 1. In theory, more accurate analysis techniques could
eliminate the need for costly mold tooling and evaluation iterations. In reality,
even the most advanced analyses remain incapable of providing accurate
estimates of performance for candidate designs given the effects of uncertain
material properties and stochastic process variation. As such, the molded part
development process is forced to utilize iterative molding evaluations in which
steel must be cut with no guarantee that the mold alterations will deliver the
desired product performance. This process, first shown in Chapter 1, has been
reproduced in Figure Chapter 6: -2.
139
Specifications
Part Design
Prototype Mold Design
Rework
Evaluation
Accept
Production
Figure Chapter 6: -2: Classic design to production development process
Optimization
Within the design process, optimization techniques may be used to more

quickly converge to an acceptable design, or to synthesize a more robust design
than had previously been considered. As such, the terms ‘optimal engineering
design’ and ‘design optimization’ refers to one stage within the iterative design
process, rather than a design process unto itself.
Optimization techniques have been applied in molded part design primarily

to enhance product performance or reduce unit cost. Examples include shape
optimization, wall thickness minimization, and cycle time reduction.104,105 All
these applications of optimization have utilized well defined relationships
between the independent design variables and the performance measure being
optimized.
In practice, all design decisions must be made in the face of uncertainty.

For the injection molding process, there is:
• inherent variation in the molding process itself,
• inconsistencies between assumed and actual process conditions,
• batch-to-batch fluctuations in material properties,
140
• unmeasured errors in processing and tooling, and

• unknown relationships between process dynamics and part performance.
If ignored, these effects can result in unsatisfactory part performance, low

production yields, and increased product cost. Taguchi’s methods for robust
parameter design have become commonplace in modern engineering design and
manufacturing practice.106 However, the effect of stochastic process variation
and uncertain material properties and process dynamics in a part’s manufacture
have not been explicitly modeled within his definition of robustness.
Optimization techniques can be utilized to obtain robust designs that are

more tolerant to unknown material behavior and stochastic process variation. As
with all optimization problems, the approach and formulation are critical
components in developing a useful model. Recent work has focused on
development of optimal design methods utilizing assumed probability distribution
functions for part tolerances and other properties.107,108,109,110 However, these
works have not considered the underlying sources of variation in the
manufacturing process. In this chapter, stochastic variation is explicitly
modeled and the effects conveyed through the molding process to the final part
properties. Once the model has been developed, the robustness of different
candidate designs and process strategies may be evaluated.
Figure Chapter 6: -3 shows the hierarchy of relationships usually defined

in design optimization problems.111 The designer starts by specifying adjustable,
independent design variables at the lowest level of decision making, for example:
what wall thickness should be used to minimize the molded part cost? With
these decisions the designer hopes to generate certain measures of design
characteristics such as cooling time. The level of these design variables must be
chosen to not violate design or process constraints, such as maximum flow
resistance. The relationship between the design characteristics, yi, and
independent design variables, xi, represent the form-fabrication relationships
from Figure Chapter 6: -1 which are not always known.
141
Adjustable Design Variables: xi
Design Characteristics: yi
Value Functions: wi
Objective Function: G
Performance: Π
Figure Chapter 6: -3: Hierarchy of design relationships
In a multi-objective optimization problem, the relative impact of the design

characteristics are calculated through the value weighting functions, wi.
Decreasing a part’s wall thickness, for example, may result in two competing
effects. As the thickness decreases, the cooling time also decreases which
reduces unit cost. However, reducing the wall thickness also increases the flow
resistance and difficulty in molding good parts, which should be penalized in its
value function. The indicial notation on xi, yi, and wi indicates that the design has
multiple adjustable design variables, design characteristics, and value functions
which must be determined by the design engineer.
These multiple values are resolved in the objective function, G. The

objective function might represent, for instance, the compromise between cycle
time, part weight, and cavity pressure which determines part cost for the
candidate design. Once the problem is formulated, the level of the design
variables are varied in an attempt to maximize or minimize the objective function,
G. This procedure is usually fully automatic, performed through non-linear
programming techniques subject to a set of constraints upon xi and yi. This
results in a set of ‘optimal’ design parameters...optimal for the specific problem
formulation. For molded part design, an optimal thickness would enable minimal
cooling time and material utilization without incurring excessively high cavity
pressures or compromising the part’s structural properties.
142
The optimal set of design parameters, xi, can then be used to assess the
design’s performance, Π. Often, the performance index might be equal or
proportional to the objective function. If the objective function for a molded part
design was developed as a cost model, for instance, then the same model could
be used to directly assess the design’s performance. However, if the
performance concern is actually part stiffness, then the relationships between the
design variables and end-use performance are incorrect. The objective function
would have been developed inappropriately, resulting in a poor choice and
estimation of design variables.
With optimization techniques, the designer’s experience plays a crucial

role in the evaluation and acceptance of a candidate design. If the ‘optimal’
design is not acceptable, the designer must re-formulate the optimization
problem, adjust the relationships between design variables and part
performance, and guide the design to a more satisfactory optimal design. In this
way, optimization becomes one stage of the iterative design process.
In commercial development of injection molded parts, the design problem

is often over constrained – there are more requirements than adjustable design
parameters. A methodology for evaluating candidate designs and processing
strategies will next be proposed to enable Dynamic Feed Control to
simultaneously satisfy multiple product requirements.
PROPOSED METHODOLOGY
"There are two primary approaches to obtaining an optimum design –
through an iterative design process or by solving an optimization
problem. In the first approach, the design is improved through repeated
modification and the values of the design are changed sequentially. The
decision as to what to change and how to change is crucial. This process
relies heavily on the designer. The second approach represents a formal
optimization problem in which all the design variables are determined
143
simultaneously so as to satisfy a set of constraints and optimize a set of

objectives."112
The proposed methodology, presented in Figure Chapter 6: -4, utilizes

both approaches – formal optimization within conventional, iterative redesign – to
obtain a robust product and tool design. The methodology begins with the
product specifications and progresses through part design and mold design as in
the conventional design process. Once a candidate design is synthesized, the
robustness of the design is evaluated through Monte Carlo simulation in which
the simulated manufacture of the part is subject to stochastic variation of material
properties, process conditions, and random noise. For each iteration in the
simulation, an optimization is utilized to find the cavity pressure distribution which
maximizes the design’s production yield, subject to weighting and constraints
defined by the designer. After many iterations, the production yield and the
robustness of the design can be assessed. If the design is not acceptable,
corrective actions may be explored until a candidate design is accepted.
Specifications
Part Design
Mold Design
Process Relations
Analysis
(See Figure 6-5)
Stochastic Variation
Performance and
Variance Estimates
Evaluation
Reject
Accept
Production
Figure Chapter 6: -4: Proposed product design methodology
144
The objective of this design methodology is to create robust part and mold
designs whose molded part properties are within desired specifications, even in
the presence of uncertain material properties and stochastic process variations.
In this work, design robustness is defined as the predicted yield of
manufactured parts that satisfy all design specifications:113
USL − LSL
ℜ= , where:
6 σ 2x − ( µ x − τ )
2
USL ≡ upper specification limit of product requirement

LSL ≡ lower specification limit of product requirement (6-1)
τ ≡ Target, or mean, of product requirement
µ x ≡ Mean of molded part properties
σ x ≡ Standard deviation of part properties
Thus, design robustness is an aggregate performance measure which

includes the consequences of product and tolerance design, process capability,
and stochastic variation. There are several beneficial properties of this definition
for robustness:
• directly extensible to multiple objectives
• allows for direct inclusion of different kinds of specifications
• consistent with Taguchi’s concept of tolerance design since it promotes
central tendencies with small deviations in part properties, rather than a
goal post mentality114
• consistent with Nam Suh’s axiom – minimize information content71 – since
the production yield will tend to decline geometrically as the number of
requirements rise.
Specifications
Specifications represent a broad set of product attributes which the

candidate design’s molded parts are required to satisfy. Table 2 lists some
specifications for a typical molded part. It is important to note that each
requirement in the set of specifications may have a varying level of importance to
145
the product, a different mode of measurement, and a distinct scale of

measurement.*,115 For instance, a tolerance specification can be accurately
measured and its effect on product performance well defined. An operator’s
visual check of aesthetic defects, however, is more subjective and, while it may
not affect the product’s functional performance, it may affect the product’s
perceived quality and overall value.
Table Chapter 6: -1: Typical product specifications for a molded part

Dimension 1 not to exceed 10.1 cm
Dimension 2 must fall within 18 ± 0.1 cm
Surface A must be flat
Outside surface of B must appear free of
visual defects
Part Design
Part design encompasses the conceptual design and embodiment stages

typical in engineering design. Through example, analysis, or intuition the
designer creates several candidate designs. The compatibility of these design
alternatives with the design objectives and specifications are then checked and
the best candidate selected and further developed. By the end of the part design
stage, the geometric form and features have been specified.
Mold Design
Mold design involves the conversion of the part’s geometric form to an

image in steel from which the part is molded. Mold design includes completion of
the detailed part design as well as tool design issues which include, in part,
layout of ejector pins, specification of draft to allow ejection of the molded part,
design and sequencing of cores and slides, and structural tool design to
withstand large, cyclic forces. These activities are not the focus of this design
*
In formal measurement theory,115 there are four types of measurement scales: nominal, ordinal, interval,
and ratio. Each increasing type infers a greater degree of precision and affords more admissible
transformations. Length is a ratio scale, permitting full mathematical operators whereas aesthetic defects
might be measured with an ordinal scale, only allowing comparison of ‘better’ or ‘worse’ than.
146
methodology, since they are well understood and do not commonly induce
lengthy delays or excessive costs during product development. Some aspects of
mold design which are crucial in product development include:
• Selection of material and grade of thermoplastic resin
• Layout and sizing of feed system
• Selection and adjustment of wall thickness
• Compensation for part shrinkage and warpage
These mold design details are crucial to product development since they
largely determine the quality of molded parts. A mold design can be considered
robust if defects do not arise during production or, if defects do arise, the mold
can be easily modified to obtain acceptable quality levels. There are an
unlimited number and type of specifications which have been used to define the
quality of molded parts. While some types of specifications (such as gloss) are
not a function of the pressure distribution and thus outside the scope of this
methodology, there are many part attributes which may be controlled as listed in
Table Chapter 6: -2.
Table Chapter 6: -2: Types of specifications

Type of Specification Acceptance Criteria
Knit-Line Position LSL≤xknit-line≤USL
Orientation LSL≤xorientation≤USL
Jetting, Hesitation xjetting≤USL
Knit-Line Strength LSL≤xstrength
Shrinkage LSL≤xshrinkage≤USL
Sink xsink≤USL
Flash xflash≤USL
Residual Stress xresidual stress≤USL
Warpage xwarpage≤USL
Each specification may have a unique acceptance criteria. For instance,

knit-line strength may be measured with a tensile test to ensure that the molded
part’s strength is above a lower specification limit. Each type of specification, xi,
and their corresponding upper and lower specification limits are shown in bold to
147
indicate an array of requirements within each type of specification. Normally, the

specification limits are chosen for a pre-specified range of variation, i.e. 6σ for
standard Cp116 or 12σ for Motorola’s quality targets.117 Once the upper and
lower specification limits have been selected, the robustness of the design be
evaluated considering the multiple requirements as:113
− 1 −1 ⎛ 1 1 n ⎞
ℜ= Φ ⎜ − Π ( 1 − 2 Φ ( −3ℜ i ) ) ⎟ , where:
3 ⎝ 2 2 i =1 ⎠
(6-2)
ℜ i ≡ Robustnessof i - th performance parameter, eq. (6 - 1)
Φ ≡ Normal probability density function
n ≡ Number of performance parameters
Φ −1 ≡ Inverse normal probability desity function
Evaluation
With the candidate design completed, the design evaluation still requires
two external inputs. First, process relationships are needed to relate the cavity
pressure distribution to molded part properties which determine the production
yield and drive the objective function. These process relations may be analytical,
empirical, or from more complex process simulations.
Stochastic variation is a second external input to the design evaluation.

This input is necessary to evaluate the robustness of the product design and
process capability in the presence of unknown material properties, random
process variation, and other factors. Some of the real-life sources of variation
which could be represented by the stochastic model include:
• variation in material properties, such as batch-to-batch variation
• effect of unmodeled or unknown complex material properties
• systematic errors in melt temperature and other process conditions
• random, time-varying process noise
• inaccurate steel dimensions
The stochastic variation is input to the evaluation through the process relations,
now assumed to have probabilistic coefficients; Monte Carlo methods are used
148
to generate instances of random variables with the specified distribution types

and characteristics.118
For each instance of varying coefficients, the design evaluation seeks to

maximize total production yield by optimizing the cavity pressure distribution
within the defined constraints, as shown in Figure Chapter 6: -5.
Stochastic Variation Estimate

stochastic behavior
Optimize
cavity pressure
Calculate part
Process Relations
properties
Estimate
batch yield
No
Done?
Yes
Calculate
robustness
Figure Chapter 6: -5: Design evaluation
For each iteration in the simulation, the optimization will converge to a set of
product attributes and a predicted ‘batch’ yield. Many iterations can be
performed to assess the effect which stochastic variation has upon the pressure
distributions and the final part characteristics, after which the design robustness
is evaluated. An example will utilize this technique for evaluating gating design
and process strategies later in this chapter.
149
Process Relations
Process relations are necessary to link the cavity pressures to the molded
part attributes which determine the product robustness and production yield.
Unfortunately, the molded part properties depend upon several factors:
• cavity pressure/temperature distribution and history,
• mold geometry and gating locations,
• material properties,
• process conditions, and
• stochastic variation.
The number of factors and complex interaction between factors make it difficult to
predict the molded part properties. Rather than try to accurately predict the
relation exactly, first-order empirical models may be utilized to estimate the main
effect between the cavity pressure distributions and molded part properties, such
as knit-line location, sink, or shrinkage:
~ ~ ~
x i = Wi ⋅ P j + V i ⋅ T j + U i , where:
(6-3)
x i ≡ i − th part property
P j ≡ Cavity pressure distribution
T j ≡ Cavity temperature distribution
~
Wi ≡ Pressure → property correlation matrix
~
V i ≡ Temperature → property correlation matrix
~
U i ≡ Property base levels
The indicial notation is crucial to understanding the model. Each part

specification (indicated by the subscript j) has a set of correlation matrices which
relate the cavity pressure and temperature distributions to the respective molded
part’s property. The accuracy of this model (represented by Wi,j, Vi,j, and Ui) is
not particularly crucial since the model coefficients will be altered probabilisticly to
relate the effects of variation due to mold geometry, material properties, process
conditions, or random noise. Additionally, inaccuracies in the process relations
modeled in the design stage can be compensated in the actual process by
150
varying the cavity pressures and temperatures to obtain the desired part
properties.
Knit-Line Location
Knit-lines are visual and structural defects occurring whenever a melt is

split (as it travels around a cut-out or through a multi-branched feed system, for
example) and then later recombines. A part’s specifications might require an
area to be free of knit-lines. With conventional molding, the designer would
estimate the melt front advancement and iteratively adjust the gating location or
runner sizing to obtain an acceptable flow pattern. With Dynamic Feed, the flow
rate through each gate can be profiled by dynamically controlling the cavity
pressure in the filling stage which subsequently determines the location of the
knit--lines.*For a given material and mold geometry, a set of empirical
~
constants, Wknit − lines , j , can be defined from Figure 5-3 which relates the knit-line
locations, xknit-line,j, to the cavity pressures in the filling stage, Pi. The indicial
notation implies that there might be multiple requirements within each type of
performance parameter, i.e. that knit-lines are to be controlled at several
locations.
Orientation
Orientation of polymer molecules and fillers determines the local structural

properties and shrinkage characteristics in molded parts. As such, orientation
may be critical in tight tolerance applications or structural applications concerned
with tensile strength, impact strength, creep or fatigue. Orientation is driven
primarily by the stress tensor generated in shearing and elongation of the melt
during the filling stages. As such, orientation is dependent upon the material
properties, pressure dynamics, and cooling history. Prediction of orientation is
currently an active field of research.119 Dynamic Feed provides the ability to
*
Alternatively, Dynamic Feed can be used to obtain a minimum number of knit-lines by sequentially opening
each gate as the melt is detected passing by each gate’s pressure transducer. This technique is currently
utilized in the industry and known as valve gate timing or sequential gating.182,183
151
control the flow rate and pressure history throughout the cavity to subsequently
adjust the shear rate and stress tensors which effect orientation.
Jetting and Other Flow Rate Dependent Defects
Jetting occurs when the melt flows into a cavity or out of a rapidly
converging mold section at excessively high flow rates, resulting in melt
instabilities and non-laminar flow.120 Since the melt’s characteristic Reynolds
numbers are very small, ϑ(10-3), jetting and related defects (shark skinning, etc.)
comes as a surprise to molders. Although normally due to poor gate or mold
design, Dynamic Feed can be used to eliminate jetting by decreasing the melt
flow rate at the critical time when jetting normally occurs. It should be noted that
similar constraints may be utilized to eliminate other defects caused by
inappropriate flow rates such as splay, blush, degradation, race tracking, and
hesitation.121,122,123
Knit-Line Strength
Published investigations have found the tensile strength across the knit-
line to be roughly 70% of the nominal part strength for glass fiber-filled
materials.124 This reduction in strength is attributed to three effects:
• a molded in v-notch on the outer surfaces at the knit-line
• incomplete melt diffusion across the melt boundaries
• absence of fibers crossing the knit-line in the flow direction
Several studies have indicated that the knit-line strength is related to the
melt temperature and cavity pressure in the area around the knit-line.125 As
such, Dynamic Feed can be utilized to provide optimal pressures to maximize the
knit-line strength. Alternatively, Dynamic Feed has been used to induce flow
after the filling stage by oscillating the cavity pressure distribution. This forces
the polymer melt and glass fibers to flow across the knit-line, permitting greater
molecular diffusion and creating a larger, parabolic inter-facial area. This
126
technique has resulted in knit-line strengths 110% of the nominal part.
152
Shrinkage
Part shrinkage results from an equilibrium between thermal contraction as

the part cools and expansive stresses induced by cavity pressures in the packing
stage. As such, the shrinkage for a specific dimension on the part is a function of
the polymer temperature and cavity pressure history in that dimension’s localized
area. Many studies have found that shrinkage is largely determined by cavity
pressure, followed distantly by melt temperature and mold temperature.127 This
implies that part shrinkage is a local, not global, property. As such, multiple part
shrinkages could be specified if the cavity pressure in the related areas is
controlled. This would allow Dynamic Feed to control multiple dimensions
simultaneously with proper part and gating design. An example will later clarify
this capability in regards to robust part design.
Sink and Internal Voids
Often appearing as a discoloration behind a rib or on a thick section, sink

is in fact an depression on the surface which scatters light and reduces a part’s
aesthetics. Sink occurs due to large volumetric contractions near the center of a
wall section which pulls the part’s skin away from the surface of the mold. While
not affecting a part’s structural or dimensional performance, it is a common
cause for part rejection.
Similar to sink, voids appear when large volumetric contractions in the

center of a thick section create an internal vacuum in the center of the part. Voids
may cause structural failure due to stress concentrations during end-use.
Surprisingly, internal voids are rarely cited as a defect, likely because they can
be detected only through rigorous destructive testing. Sink and internal voids
can be minimized by increasing the local cavity pressure to reduce the volumetric
shrinkage in areas of thick sections.
153
Flash
While sink and voids occur due to low cavity pressures, higher cavity
pressures may result in flash. Flash occurs when cavity pressures cause local
mold deflection which allows some material to flow into the parting plane of the
mold. Very high cavity pressure may overcome the clamp tonnage and force the
mold halves to separate significantly, allowing significant melt flow around the
entire perimeter of the part.
Flash is a common defect. Even though the flash may be manually

trimmed if it interferes with a parts function or assembly, it may cause premature
wear of the mold and excessively high part costs due to manual labor and low
production yields. Flash may be avoided by reducing the cavity pressure in
areas where flash is occurring. This requirement, together with the pressure
requirement for sink, form two conflicting constraints for cavity pressure.
Residual Stress
Prediction of residual stress in molded parts is another active area of

research.128 While not considered a measurable defect, residual stress is of
primary concern to molders due to implications in long term structural and
dimensional performance of molded parts. The residual stress is dependent
upon the cavity pressure distribution, temperature history, and material properties
during processing. The resulting stress field can be exceedingly complex: a
part’s skin normally exhibits tensile residual stresses while the core exhibits
compressive stresses – the magnitude of both vary significantly across the area
of the part. Dynamic Feed provides the ability to profile the cavity pressure
distribution throughout the packing phase. In essence, the cavity pressure can
be varied as the frozen layer propagates from the mold walls to control the
residual stress distribution across, and through, the part. Thus, the residual
stresses can be optimized to provide excellent end-use performance.
154
Warpage
Warpage is one of molder’s greatest concerns since it is a non-linear

phenomenon with complex dependencies upon cavity pressure, cavity
temperature, material properties, and part geometry. Warpage sometimes results
from asymmetric cooling across the part’s cross section, when the mold halves
are run at different mold temperatures or the cooling line design provides
inadequate heat transfer. This results in an asymmetric stress distribution across
the thickness which forms a permanent bending moment on the part.129 This
methodology does not consider this form of warpage since it may be corrected
through alteration of mold temperature controllers.
Warpage is also commonly caused by differential shrinkage across the

part. Differential shrinkage may be caused by variations in orientation,
shrinkage, or residual stress during molding. For each of these root causes,
Dynamic Feed can be used to induce a cavity pressure distribution to help
resolve the warpage.
Design Optimization
With the process relations defined, the design evaluation will vary the
cavity pressure distribution and history to maximize the overall production yield.
If two goals are conflicting, xi,j and xi’,j’, then a set of pressures, Pk, will be
selected that makes a compromise between the two to maximize the overall
yield:
155
⎧ −1 ⎛1 1 n ⎫
ℜ = maximize ⎨ Φ −1 ⎜ − Π ( 1 − 2 Φ ( −3ℜ i
⎝ 2 2 i =1
) ) ⎞⎟⎠ ⎬ ,
Pk ⎩ 3 ⎭ (6-4)
where:
USL i − LSL i
ℜi =
( )
2
6 σ 2x i − µ x i − τ i
~ ~ ~
x i = Wi ⋅ P j + V i ⋅ T j + U i
subject to:
P j ≥ LPL j
P j ≤ UPL j
x i ≥ LSL i
x i ≤ USL i
There are two types of constraints to be applied in the optimization: design

constraints and process constraints. Design constraints are imposed by the
designer on the allowable range of adjustable design variables, xi – for instance,
to guarantee that a length must be greater than 0 or a wall thickness must be
less than 0.5 cm. Process constraints stem from the physical limitations of the
molding process as described in Chapter 5 – that cavity pressure must fall within
0 and 150 MPa, for instance
Robustness
From the Monte Carlo simulation, the product robustness is evaluated. A

robustness of 1.0 corresponds to the predicted molded part property being
centered between the upper and lower specification limits, with a predicted
standard deviation of one sixth of the tolerance band – this corresponds to a
production yield of 99.3%. For higher values of robustness, the production yield
approaches 100%, indicating that the cavity pressure distribution can be adjusted
in every case to produce parts within the required specifications. If a design is
infeasible, then the robustness will be very close to 0.
156
In the molding of complex parts, such as instrument panels, production

yields of ~95% are considered excellent.130,131 By utilizing process relationships
with wide probabilistic spreads in the simulation, a predicted robustness near 1
indicates that the process flexibility exists to meet the required product
specifications and that re-tooling should not be necessary in production.
If the predicted robustness is significantly lower than 1, rework of the mold

or part design may be necessary to increase the robustness of the product; the
results of the simulation will indicate which constraint or objective is causing the
loss in the product robustness, suggesting a starting point for the redesign. This
may involve changing the gating scheme, varying the thickness, increasing
allowable tolerances, or other numerous actions. When corrective actions have
been completed, the relative success of the new design may be evaluated.
A CASE STUDY
To illustrate the use of this design methodology in molded part design, the
example from Chapter 2 will be further investigated. In this particular example,
the robustness of different gating designs will be evaluated with simulated
stochastic shrinkage variation.
Material suppliers have historically quoted shrinkage as a range, for

instance 0.5 to 0.7%. Also, molders have long used a rule of thumb that
shrinkage can be decreased 0.05% for every 7 MPa (1,000 psi) increase in cavity
pressure – these characteristics have been experimentally demonstrated in
Figure 5-6. However, the modeling of material shrinkage as a stochastic function
of cavity pressure has not been previously performed. As the example shows,
the stochastic model leads to a very clear evaluation of design robustness for
injection molding. Both concepts – stochastic modeling and robustness – are
directly extensible to the design and manufacture of more complex molded parts
with different kinds of specifications.
157
Design
Figure Chapter 6: -6 shows a typical molded housing with four bosses

used to attach a cover to the base. In this application, three critical dimensions
have been specified to provide a proper fit with a mating part as shown in the
right hand figure. The required tolerances are listed below in Table Chapter 6: -3
– note that these tolerances are not stringent and represent a typical industry
standard of ~0.2%.132
L1
L2
L3
Figure Chapter 6: -6: Typical molded part and specified dimensions
Table Chapter 6: -3: Specified dimensions for molded part

Dimension Required (in)
L1 6.00 ± 0.010
L2 10.00 ± 0.015
L3 6.00 ± 0.010
Design Optimization
In this example, the goal of the designer and molder is to produce molded
parts within the tolerances listed in Table Chapter 6: -3. With the geometry and
specifications of the design already determined, the designer can influence the
product robustness through estimation of linear shrinkages and choice of gating
locations.
The published linear shrinkage for this grade of amorphous polycarbonate

is between 0.4 and 0.7%. In this application, the mold designer has assumed a
158
constant linear shrinkage of 0.5% for initial tooling. As has been previously
illustrated, shrinkage is a complex function of the material properties and process
dynamics. Experimental work has shown that linear shrinkage is driven primarily
by cavity pressure – molders frequently utilize the rule that shrinkage is
decreased 0.05% for every 7 MPa (1000 psi) increase in cavity pressure. This
relationship is also exhibited from the experimental data of Chapter 5 (Figure 5-7)
which yields:
Pressure
shrinkage = 0.8% − 0.05% ⋅ (6-5)
1000 psi
This relationship, however, can not be used with confidence to tune a mold since
stochastic variations in material properties and process conditions can lead to
significant deviation in material shrinkage.133 As such, a stochastic
representation of shrinkage may be used to emulate this variation:
~
s k (%) = ~
s kmax − ~
s kslope ⋅ Pk ,
where:
(6-6)
s kmax = N ( µ = 0.8 , σ = 0.1)
~
s slope = N ( µ = 0.05, σ = 0.01)

~
k
By taking a number of normally distributed samples, the stochastic

variations in the molding process can be simulated. Figure Chapter 6: -7 plots
the relationship between cavity pressure and linear shrinkage. The gray region
indicates the shrinkage range commonly assumed by designers and molders.
The stochastic shrinkage models differ both in y-axis intercept and slope,
providing significant behavioral differences in the dimensions of molded parts.
159
0.9
0.8
0.7
Shrinkage (% in/in)
0.6
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60
Figure Chapter 6: -7: Stochastic shrinkage behavior of polycarbonate
With the specifications and process relations defined, only the constraints
for cavity pressure remain to be defined. The cavity pressures are constrained
by the machine capabilities and process dynamics. Two straightforward
constraints are the maximum cavity pressure, constrained by the maximum
packing pressure of the machine, and the minimum cavity pressure, equal to
zero.
Pi ≥ 0 MPa
(6-7)
Pi ≤ 80 MPa
An additional, more complex constraint exists due to the interaction of

cavity pressures at different positions in the cavity. For a given wall thickness
and material viscosity, there is a maximum pressure differential between gates
that can exist. If pressure at one gate is much higher than nearby gates, then the
lower pressure at the nearby gate will increase due to material flow from the high
pressure gate. Experimental work in Chapter 5 has shown that this pressure
differential is roughly 2MPa/cm and can be represented by the constraint:
160
Pi − Pj
−2 ≤ ≤ 2 , where: (6-8)
Li → j
Pi and Pj are pressures at different points in the cavity, and Li → j is the distance
between those two points.
Objectives and Constraints
With the process relations, objectives, and process constraints defined,

the design evaluation for this example can be summarized as:
⎧ −1 ⎛1 1 3 ⎞⎫
ℜ = maximize ⎨ Φ −1 ⎜ − Π ( 1 − 2 Φ ( −3ℜ i ) ) ⎟ ⎬ , where:
Pi , k ⎩ 3 ⎝ 2 2 i =1 ⎠⎭
USLi − LSLi
ℜi =
( )
2
6 σ 2x i − µ x i − τ i
⎡ 5.990 6.000 6.010 ⎤
[USL τ LSL ] = ⎢⎢ 9.985 10.000 10.015⎥⎥
⎢⎣ 5.990 6.000 6.010 ⎥⎦
N
∑x i ,k
µx = k =1
(6-9)
i
N
∑( x )
N
2
i ,k − µ xi
σ 2x = k =1
i
N
⎡ 1 0 0 ⎤
x i , k = 0.25 0.5 0.25⎥ ⋅ τ ⋅ ( 1 − ~
⎢ sk ) ⋅ Pi , k
⎢ ⎥
⎢⎣ 0 0 1 ⎥⎦
~
sk (%) = N ( 0.8 , 0.1) − N ( 0.05, 0.01) , subject to:
Pi , k ≥ 0 MPa
Pi , k ≤ 60 MPa
Pi , k − Pi) , k
−2 ≤ ≤2
Li → i)
It should be noted that this optimization problem is non-linear: the

objective function is a product of multiple yields and the yields are a non-linear
161
function of the molded part dimension. However, it is well behaved and was
easily solved by a quasi-Newton search method with a central differencing
technique.134,135 Appendix D lists the data tables, macro-language, and
spreadsheets utilized in the Monte Carlo simulation.
Design Evaluation
The cavity pressure distributions for any candidate design can now be
optimized for each set of stochastic shrinkage properties, allowing prediction of
the molded part’s dimensions. After many iterations, a distribution of dimensions
will emerge, allowing the design’s robustness to be evaluated. The next four
examples illustrate the effectiveness of different gating designs and process
strategies in obtaining the desired tolerances.
One Center Gate, Constant Pressure
As previously introduced in Chapter 2, a common gating scheme for the

housing would involve direct sprue-gating into the center of the part as shown in
Figure Chapter 6: -8. In this example, the cavity pressure was not optimized.
Rather, the cavity inlet pressure was maintained at a constant 30 MPa, the
pressure for which the deterministic shrinkage would correspond exactly to that
of the assumed linear shrinkage, 0.5%. Since this is typical of molding
operations, this design provides an industry benchmark against which to
compare more advanced designs and process control strategies.
162
L1
L2
L3
Figure Chapter 6: -8: Center-gated box
Table Chapter 6: -4 lists the molded part dimensions resulting from the
stochastic process relationship previously described. For this design, L2 is the
most well-behaved dimension, with a centered mean and the lowest deviation
(0.006 with a ±0.015 tolerance). The other dimensions generally exhibit more
shrinkage since they are farther from the gate and experience cavity pressures
slightly lower than 30MPa. L1 exhibits a low predicted yield of 76.5% due to
undersized molded part dimensions which are approaching the lower
specification limit.
Table Chapter 6: -4: Performance of one center gate at constant pressure

L1 L2 L3
Ave. Dimension 5.995 10.000 5.999
Std. Deviation 0.005 0.006 0.006
Max. Dimension 6.008 10.014 6.013
Min. Dimension 5.981 9.987 5.988
Predicted Yield 76.5 94.8 87.3
As previously defined, the robustness is the expected yield of parts which

meet the required dimensional tolerances. According to our definition, this
design has a robustness of 0.22. This corresponds to a process yield of
approximately 63% – roughly 37% of all molded parts will be rejected due to
stochastic variation in the manufacturing process. This result is consistent with
163
industry yields during the startup manufacture of complex molded parts with
multiple requirements.130,131
One Center Gate, Optimal Pressure
The previous design utilized a constant pack pressure across all the runs.
A more capable approach, becoming more common in industry, is to qualify the
process for a given mold geometry on a specific machine with a specific lot of
material – to optimize the process to achieve higher manufacturing yields. While
there is stochastic variation between molding machines and material lots, the
variation within a batch of parts is greatly reduced.
For each set of stochastic relationships, the input cavity pressure was re-
optimized to maximize the total yield. This might result in a resulting gate cavity
pressure of 34 MPa for the first set, 29 MPa for the second, and so on. Table
Chapter 6: -5 summarizes the resulting dimensions. The robustness of this
process is 0.56 with a yield of nearly 80%. However, nearly 20% of the molded
parts will have at least one dimension out of tolerance. While this is a
considerable fraction, it is a significant improvement over the previous example
and does not require additional investment, just a change in the operation of the
molding machine.
Table Chapter 6: -5: Performance of one center gate at optimal pressure

L1 L2 L3
Ave. Dimension 5.996 10.003 6.002
Std. Deviation 0.004 0.004 0.004
Max. Dimension 6.006 10.013 6.011
Min. Dimension 5.987 9.993 5.993
Note that dimensions L2 and L3 are slightly oversize – the cavity pressure
has been increased to bring L1 closer to the center. The resulting decrease in
the process yield of L2 and L3 are greatly offset by the large increase in the
process yield of L1. Moreover, the standard deviation of the dimensions has
164
been reduced along with the range of dimensions. Further analysis has shown
that if the mold designer estimates this variance in shrinkage and cuts the tool to
non-constant dimensions (make L1 slightly longer and L2 and L3 slight shorter),
then the design robustness increases to 0.8 – a yield of nearly 90%!
Three Gates, Optimal Pressure
An 80% product yield is rarely acceptable in high volume production,

however. There are three possible ways to possibly improve the yield:
• increase allowable dimensional tolerances through product design
• remove stochastic variability by qualifying materials and machines
• add degrees of freedom to compensate for variability
Presumably, the specifications have originated from a design which has

considered these issues; the tolerances are fairly loose. Also, stochastic
variation can not be completely removed; the model presented in equation (6-6)
is a good representation of stochastic process variation. Dynamic Feed Control
can be used to directly control the cavity pressure at multiple locations in the
cavity and adjust for variation. An intuitive design would utilize three gates in the
center of each of the critical dimensions as shown in Figure Chapter 6: -9,
reproduced from Chapter 2.
L1
L2
Time
Adjustable
Valves
L3
Figure Chapter 6: -9: Dynamic Feed Control of three gates
165
The pressure at each gate can then be optimized for each iteration of the
Monte Carlo simulation to adjust for the varying levels of shrinkage, subject to the
process constraints previously described in Chapter 5. Table Chapter 6: -6 lists
the predicted dimensional properties. The resulting robustness was 1.47,
corresponding to a process yield of roughly 99% for the same set of stochastic
shrinkages assumed in analysis of the other designs. Note that the maximum
and minimum part dimensions are completely within the tolerance limits. In fact,
the dimensional deviations are such that the tolerances can be tightened
significantly without incurring extra cost, possibly enabling enhanced product
quality or new product capabilities.
Table Chapter 6: -6: Performance of three gates at optimal pressures
L1 L2 L3
Ave. Dimension 6.000 10.000 6.000
Std. Deviation 0.001 0.002 0.001
Max. Dimension 6.003 10.007 6.004
Min. Dimension 5.997 9.994 5.996
Two Gates, Optimal Pressure
0.93 For completeness, a two gate design was synthesized as shown in

Figure Chapter 6: -10. This gating design allows for excellent control of
dimensions L1 and L3 while providing good ‘averaging’ control of L2. The resulting
process robustness was 0.93 for the predicted dimensions listed in Table
Chapter 6: -7. This level of robustness corresponds to a process yield of 93%.
166
L1
L2
Time
Adjustable
Valves
L3
Figure Chapter 6: -10: Two gates, optimal pressures

Table Chapter 6: -7: Performance of two gates at optimal pressures
L1 L2 L3
Ave. Dimension 6.001 10.002 5.998
Std. Deviation 0.003 0.004 0.004
Max. Dimension 6.018 10.010 6.005
Min. Dimension 5.994 9.995 5.991
It is important for the designer to apply a cost-benefit analysis to identify

the level of robustness required by comparing the cost of compliance (cutting
non-uniform tool dimensions or adding extra valves in a Dynamic Feed system)
to the costs of quality (reduced yields, re-tooling estimates, end-use performance
degradation).136137
PROCESS CONTROL FOR MULTIOBJECTIVE OPTIMIZATION

As Chapter 5 has demonstrated, Dynamic Feed Control enables the
injection molding process with the flexibility to tune the feed system and achieve
the process dynamics required to produce acceptable parts. This is critical
during the mold development stage of the product development cycle, allowing
for lower development and tooling costs as well as faster time to market.
However, this process flexibility can also be leveraged in the production stage,
when molders are trying to maintain high production yields while keeping multiple
product properties within specification.
167
When there are multiple objectives, a tradeoff boundary can be defined.

This boundary corresponds to those operating points where performance in one
objective must be given up in order to achieve better performance in another
objective. In the conventional injection molding process, it is up to the operator
to determine the best operating point in a particular set of circumstances – the
operator has internalized an implicit objective function and chooses an operating
point he believes to be optimal.
If the operator is a process expert and familiar with the current set of
circumstances, the operator’s decision may bring about a true optimal operating
point. However, the machine operator rarely understands the process dynamics
which bring about the molded part’s characteristics. Moreover, even process
experts have difficulties in operating the process efficiently when faced with an
unfamiliar situation138 or when the number of objectives increase.139 In these
situations, the operator’s choice of ‘efficient’ operating points may be far from
optimal.
For the previously investigated example of the case utilizing three gates,
the operator would likely have a difficult time choosing the optimal pressures to
maximize the production yield. If one dimension is clearly out of specification,
the operator would likely know how to change the cavity pressure to improve that
one dimension. However, the operator would not likely consider the effect which
that change will have on the other part dimensions. Given that the optimal
operating point lies on a tradeoff boundary, several cavity pressures would need
to be simultaneously changed to quickly optimize the process. Since the
operator must find the optimal point manually, process optimization becomes a
tedious and expensive chore for the operator, even if he understands the
fundamental relation between cavity pressures and part dimensions. As such,
molders commonly run at sub-optimal operating points.
168
What is needed is an on-line system that controls the process conditions

subject to feedback from the operator regarding part quality. A control scheme
for accomplishing this goal is shown in Figure Chapter 6: -11.
Operator inputs
part properties
Process Relations Optimize

cavity pressures
Dynamic Feed
Injection Molding
Reject
Inspect
Accept
Figure Chapter 6: -11: Process control for multi-objective optimization
This process is very similar to the design evaluation process of Figure Chapter 6:
-5 on page 149. The primary difference is that the feedback from the operator
regarding the actual molded part properties replaces the performance predictions
from the Monte Carlo simulation of stochastic properties. Interestingly, this
process control scheme can utilize the same process relations and optimization
algorithm which have been previously developed. Again, the process model
does not need to be a precise representation of the molding process – just
representative of tendency which cavity pressure has upon the part properties.
Since the control system for closed loop control of cavity pressures
already exists within the Dynamic Feed Control system, the implementation of
this process control scheme is straightforward – the output from the optimization
can be fed directly to the input of the Dynamic Feed Control system which will
deliver the desired process dynamics. This implementation removes the
necessity for the operator to input the cavity pressure traces at each gate which
required expert process knowledge. With this added layer of control placed on
169
top of the closed loop control system, an unskilled operator may operate the
molding process at an optimal point, without lengthy guesswork. As an example,
consider the dimensions shown Table Chapter 6: -8 for the previous example
with three drops.
Table Chapter 6: -8: Performance of three gates at non-optimal pressures

Iteration L1 P1 L2 P2 L3 P3
Initial 5.994 30 MPa 10.000 30 MPa 6.000 30 MPa
Intermediate 5.998 36 MPa 10.001 28 MPa 6.000 30 MPa
Final 6.000 38 MPa 10.000 26 MPa 6.000 30 MPa
For this example, the dimension L1 was found to be within tolerance but
off-center (due to mold temperatures, material variation, ...) – this data was fed
back to the process control system. Using the internal process relations, the
control system increases the pressure around L1 to reduce the shrinkage.
However, this would also slightly increase L2 so the control system reduces the
cavity pressures in the center of the part. Several parts are molded with these
new process conditions, the new dimensions are measured, and the results fed
back to the control system, if necessary. Using this technique for multi-objective
process control, the molding process can be maintained at an optimal operating
point without incurring lengthy or costly production delays.
SUMMARY
This chapter developed a methodology for the design and production of
high quality, molded plastic parts. The estimation of robustness in the design
stage required modeling of the process relations between cavity pressure and
the part attributes with stochastic representations. This permits slack in the
precision of model parameters and lends itself directly to the estimation of design
robustness. This methodology was then utilized for the analysis of a four-sided
housing with three critical dimensions. Different gating scenarios were examined
and their robustness identified. The results provide the first theoretical
170
confirmation of the tight tolerance design philosophy currently being expounded

in industry.69
Finally, it was illustrated how this methodology may be utilized in a

process control scheme for on-line multi-objective optimization. This process
control scheme explicitly utilizes the capabilities of Dynamic Feed Control without
requiring the operator to be a process expert and specify the cavity pressure
distribution directly. While the examples of this chapter focused on dimensional
control, the extension of this methodology to design and production of complex
parts with other types of specifications is straightforward.
171
172
101
Dixon, J.R., Howe, A., Cohen, P.R., Simmons, M.K., Dominic I: Progress towards Domain Independence
in Design by Iterative Redesign, Proceedings of the 1986 ASME Computers in Engineering, v. 4, pp. 199,
1986.
102
Personal communication with Larry Leifer, Professor, Stanford University (January, 1994).
103
Rinderle, J.R., Implications of Function-Form-Fabrication Relations on Design Decomposition Strategies,
Proceedings of the 1986 ASME Computers in Engineering, v. 4, pp. 193, (1986).
104
Lorenzo, L., McIntyre, R. K., Engineering design and Optimization of a Thermoplastic Mower Deck,
Proceedings from the 1994 Annual Technical Conference of the Society of Plastics Engineers, v. 52, pp.
3166 (1994).
105
Kyle, B.R.M., Using Taguchi Methodology for Optimizing Products and Processes, Proceedings from the
1989 Annual Technical Conference of the Society of Plastics Engineers, v. 47, pp. 1704 (1989).
106
Taguchi, G., Tsai, S.C., Taguchi on Robust Technology Development, ASME Press, New York, 1993.
107
Lewis, L., Parkinson, A., Robust Optimal Design Using a Second Order Tolerance Model, Research in
Engineering Design, v. 6, pp. 25 (1994).
108
Otto, K. N., Antonsson, E. K., The Method of Imprecision Compared to Utility Theory for Design Selection
Problems, Design Theory and Methodology, v. 53, pp. 167 (1993).
109
Eggert, R. J., Mayne, R. W., Probabilistic Optimal Design Using Successive Surrogate Probability Density
Functions, Journal of Mechanical Design, v. 115, pp. 385 (1993).
110
Eggert, R. J., Quantifying Design Feasibility Using Probabilistic Feasibility Analysis, Advances in Design
Automation, v. 32, n. 1, p. 235 (1991).
111
Siddall, J. N., Optimal Engineering Design: Principles and Applications, Marcel-Dekker, Inc, New York,
1982.
112
Balachandran, M., Knowledge-Based Optimum Design, Computational Mechanics Publications, Hobbs
the Printers Ltd., Southampton, Great Britain, 1993.
113
Ford, R. B., Quality Improvement of Mechanical Systems through Design Innovation at the Conceptual
Phase, Doctoral Dissertation submitted to the Stanford University ME Design Division, 1995.
114
DeVor, R. E., Chang, T., Sutherland, J. W., Statistical Quality Design and Control: Contemporary
Concepts and Methods, Macmillan Publishing Company, New York, 1992.
173
115
Grove, Mock, Formal Definition and Characteristics of Measurement Systems: A Factual View of
Measurement, Chapter 2 of Measurement, Accounting, and Organizational Information, John Wiley, New
York, 1979.
116
Smith, G., Statistical Process Control and Quality Improvement, Merrill Publishing Co., New York, 1991.
117
Balf, T., Motorola Nears Quality Benchmark after Twelve Year Evolutionary Effort, Aviation Week and
Space Technology, v. 135, n. 23, pp. 64 (December 9, 1991).
118
Siddall, J. N., Probabilistic Engineering Design: Principles and Applications, Marcel Dekker, New York,
1983.
119
Takeshima, M., Funakoshi, N., Molecular Orientation Distribution in Injection Molded Polycarbonate Disks,
Journal of Applied Polymer Science, v. 32, pp. 3457, (1986).
120
Personal communication with D. Lee, Professor of Mechanical Engineering, Rensselaer Polytechnic
Institute, May, 1991.
121
Yokoi, H., Nagami, S., Kawasaki, A., Murata, Y., Visual Analyses of Flow Marks Generation Process
Using Glass-Inserted Mold, Proceedings from the 1994 Annual Technical Conference of the Society of
122
Hawkins, W.L., Polymer Degradation and Stabilization, Springer-Verlag, Berlin, Germany, 1984.
123
Irani, R.K., Integrating Features, Iterative Redesign, and CAE Technology to Automate the Design of the
Feed System of Injection Molds, Doctoral Dissertation submitted to the Department of Mechanical
Engineering, University of Massachusetts at Amherst, 1989.
124
Fisa, B., Rahmani, M., Weldline Strength in Injection Molded Glass Fiber Reinforced Polypropylene,
Polymer Engineering and Science, v. 31, n. 18, pp. 1330 (1991).
125
Roe, D. S., The Effect of Process Conditions on Knit-Line Strength, Masters Thesis submitted to
University of Lowell Department of Plastics Engineering, 1992.
126
Kazmer, D. O., Roe, D. S., Increasing Weld-Line Strength through Dynamic Control of Volumetric
Shrinkage, Proceedings from the 1994 Annual Technical Meeting of the Society of Plastics Engineers, v.
52, pp. 631 (1994).
127
Väätäinen, O., Järvelä, P., Valta, K., Järvelä, P., The Effect of Processing Parameters on the Quality of
Injection Moulded Parts by using the Taguchi Parameter Design Method, Plastics, Rubber and
Composites Processing and Applications, v. 21, n. 4, pp. 211 (1994).
128
Siegmann, A., Buchman, A., Kenig, S., Residual Stresses in Polymers III: The Influence of Injection-
Molding Process Conditions, Polymer Engineering and Science, v. 22, n. 9, pp. 560 (1982).
174
129
Bushko, W. C., Stokes, V. K., The Effects of Differential Mold Temperatures on the Warpage of Packed
Injection-Molded Parts, Proceedings from the 1994 Annual Technical Conference of the Society of
130
Personal communication with Paul Choineire, Technical Development Engineer, GE Plastics, Pleasanton,
California (January 16, 1994).
131
Personal communication with Anthony Cavotta, Engineering Manager, Trend Plastics, Fremont,
California (March 7, 1995).
132
Beris, M. C., Standards and Practices of Plastics Custom Molders, SPI Plastics Engineers
Handbook of the Society of the Plastics Industry, 5th Edition, Van Nostrand Reinhold Co., New
York, 1991
133
Schulz, M., Alig, I., Influence of Stochastic Environments of Gaussian Chains on Dynamic Shear and
Bulk Properties, Journal of Chemical Physics, v. 97, n. 4, pp. 2772 (1992).
134
Davidon, W. C., Optimally Conditioned Optimization Algorithms without Line Searches, Math. Prog., v. 9,
n. 1, pp. 1, (1975)
135
Fletcher, R., Powell, M. J., A Rapidly Convergent Descent Method for Minimization, Computer J., v. 6, n.
2, pp. 163 (1963).
136
Morse, W.J., Measuring, Planning, and Controlling Quality Costs, National Association of Accountants,
Montvale, New Jersey, 1987.
137
Campanella, J., Corcoran, F., Principles of Quality Costs, 16, 16, 1983.
138
Morris, N. M., Rouse, W. B., The Effects of Type of Knowledge upon Human Problem Solving in a
Process Control Task, IEEE Trans. Syst., Man, and Cybernetics, v. SMC-15, pp. 698, (November, 1985).
139
Moray, N., Lootsteen, P., Pajak, J., Acquisition of Process Control Skills, IEEE Trans. Syst., Man, and
Cybernetics, v. SMC-16, pp. 697, (July, 1986).
175
Chapter 7: Concluding Remarks
TABLE OF CONTENTS
CHAPTER 7: CONCLUDING REMARKS...........................................................................................172

CONCLUSIONS ........................................................................................................................................172
CONTRIBUTIONS...................................................................................... Error! Bookmark not defined.
Implementation and Validation ......................Error! Bookmark not defined.
Design and Manufacturing Robustness.........Error! Bookmark not defined.
FUTURE WORK .......................................................................................................................................176
Design Methodology................................................................................. 176
Control System......................................................................................... 176
System Design ......................................................................................... 177
Commercial Validation.............................................................................. 178
TABLE OF TABLES
TABLE 7-1: ALTERNATE SOLUTION APPROACHES ........................................................................................177
1
CONCLUSIONS
Improvement in the design and production of high quality, molded plastic

parts has become increasingly urgent as thermoplastics continue their
penetration into technical applications which demand conformance to stringent
specifications and the highest quality control levels. At the same time, the
plastics industry is hard pressed to improve performance, reduce cost, and
shorten delivery time to meet escalating global competition. In many instances,
today’s product requirements exceed the capabilities of the product development
and manufacturing processes available in the plastics molding industry. These
shortcomings are evidenced by long product development cycles, excessive
tooling costs, low process yields, and inferior product quality.
To improve the quality of molded parts, recent development efforts have

focused on resolving the inherent limitations of the injection molding process.
The results have been mixed. Simulation has reduced some of the difficulties
associated with product design and processing but has not increased the
capability of the actual molding process. Similarly, closed loop feedback control
and process optimization techniques have reduced the sensitivity of the process
to a limited degree but have not improved the process flexibility. These
approaches have had only limited success because they lack the necessary
degrees of freedom. On site modifications are still difficult since the process
dynamics can not be easily changed once tooled in hard steel.
172
CONTRIBUTIONS OF THIS RESEARCH
Dynamic Feed Control is a radical alternative to the conventional “design,

build, test” methodology standard in the industry today. The research introduces
a new methodology, exploring a closed loop process that offers greatly increased
process flexibility and improved consistency compared to current molding
technology. This approach introduces additional degrees of freedom at several
locations in the mold by dynamically changing the mold characteristics during
critical phases of the molding cycle. These degrees of freedom can be used to
compensate for complex material properties, input variations, and changing
production requirements. With this production stage flexibility, the product time
to market can be dramatically reduced while ensuring significantly improved
levels of product quality and process yields.
A gain-scheduling approach was employed to control the cavity pressure

throughout the filling and packing stages. An adaptive mechanism was
developed to identify the point of switchover from the filling to the packing stage
and to provide a smooth and stable transition between stages. The development
of this control strategy was facilitated with a minimum-complexity process model.
While the model utilizes many simplifying assumptions, the process model
provides significant insight into the dynamics of the molding process and is, in
fact, an excellent representation of the aggregate process dynamics for the mold
geometries that were utilized in the experimental portion of the research.
For the first time, cavity pressure during the filling and packing stages
were controlled at multiple locations simultaneously. This is particularly
significant considering the number of previous ineffective
efforts.27,33,35,36,38,40,41,42,43,44,45,46 This was accomplished utilizing an adaptive
gain scheduling method. Dynamic closed loop cavity pressure control enables
the decoupling of the filling dynamics from the packing dynamics of the molding
machine. The multiple valves permit dynamic decoupling of different areas of the
173
cavity whether in a multi-gated part or in a multi-cavity mold. In essence, each

sub-volume of the mold has its own, locally controlled injection unit.
Experimental investigation confirmed the process flexibility and

consistency of the Dynamic Feed Control process. Results demonstrated that
the flow rates and packing pressures may be selectively specified at each gate
for a variety of part geometries and wall thicknesses. The flexibility to control the
flow rate and cavity pressures in a very unbalanced multi-gated tool was
demonstrated. Dynamic Feed metered the flow in such a way as to avoid over-
packing of a small cavity while completely filling a much larger cavity, producing
high quality parts from both cavities. This is impossible with conventional
molding technology. Moreover, the flexibility to bring a molded part within
specified tolerances without re-cutting mold steel by manipulating cavity packing
pressures was demonstrated. For polycarbonate, the packing stage flexibility
allowed multiple dimensions to be selectively altered within a shrinkage range of
0.2% to 0.8%.
Another experiment was designed to investigate the effects of typical

“noise” during the commercial production of a molded part, simulating on a
variety of machines operated by different molders. Noise was simulated by
conducting molding experiments over a select range of melt temperatures, mold
temperatures, injection speeds, and packing pressures. Process variations were
reduced by 75% when compared to the conventional injection molding process.
For a conventional commercial grade tolerance of 0.2%, the conventional
molding process with noise was found to have a process capability (Cp) of 0.52.
Utilizing the Dynamic Feed system resulted in a Cp of 1.67.
Dynamic Feed Control provides the product designer additional freedom

while simplifying the tasks of the tooling engineer and machine operator. Much
of the difficulty in obtaining high quality, molded parts has traditionally arisen from
the effects of thermoplastics’ complex material properties and unknown process
174
variation. These effects are not considered in the design and product
development stage, but arise later to require product and tool changes to
produce parts of acceptable quality. Dynamic Feed eliminates the need for
design and tool changes by providing process flexibility in the production stage.
By modeling the relationships between cavity pressures and part

properties as stochastic functions, the research has shown that the effect of
uncertainty in material properties and process dynamics on the manufacturability
of the part can be estimated at the design stage. With an estimation of
uncertainty, the robustness of candidate design and processing strategies can be
evaluated during product development. This methodology was then investigated
for a four-sided box with three critical dimensions. Different gating scenarios and
process strategies were examined and their robustness identified. The results
were consistent with recent industry findings – this is the first analytic research to
explicitly support the tight-tolerance design guidelines becoming common in
industry.70 For each critical specification in the design, a gate should be located
to provide a degree of freedom to adjust the flow rate and cavity pressure in that
area which determines the molded part’s property.
Finally, it was illustrated how this same methodology may be utilized on-
line for multi-objective optimization during production. This process control
scheme explicitly utilizes the capabilities of Dynamic Feed Control without
requiring the operator to understand the process dynamics and specify the cavity
pressure profiles directly. While the examples of this methodology focused on
dimensional control, the extension to design and production of more complex
parts with multiple types of specifications is straightforward. An approach is
suggested in the next section.
175
DIRECTIONS FOR FUTURE WORK
Design Methodology
The design methodology utilized a multi-objective optimization approach

with Monte Carlo simulation to determine the robustness of candidate designs
when subject to stochastic material or process variation. A specific example was
developed using an empirical relationship for material shrinkage from the
experimental results. This methodology can be extended to include arbitrarily
complex part geometries, all types of product specifications, more complex
material and process physics, and different kinds of input variation. However,
this implementation requires significantly more detail and complexity in the
estimation of the molded part properties.
Current commercially available simulations model the relationship

between process physics and molded part properties for arbitrarily complex part
geometries. For industry applications, such simulations should be utilized in our
design methodology to evaluate the design and process robustness of candidate
designs, before the design reaches tool steel. Alternatively, the commercial
simulations could be extended to include stochastic modeling of probable
sources of input variation (from material, machine, or operator) to provide a
measure of robustness – a more likely approach since industry will financially
support this form of implementation. In fact, such a design tool is currently under
development to automatically select molding machine process conditions to
optimize part properties.140
Control System
While our control system provides closed loop control of multiple cavity
pressures, few process experts understand the relationship between cavity
pressure and resulting part properties. Thus, the control system must be
extended to utilize rule- or model-based control laws to internally determine the
176
cavity pressure history and distribution which will produce the desired part
properties; feedback of the molded part properties (dimensions, appearance,
etc.) can be provided by the process operator. Once the control system has
been extended to part properties, a multi-objective control algorithm can be
explicitly modeled for production and operator support.
System Design
While the Dynamic Feed concept demonstrated unparalleled capabilities,

a future design should seek to improve those capabilities while reducing the
system cost and complexity. Currently, the use of Dynamic Feed is costly –
requiring capital investment in hydraulic sub-systems, pressure and position
transducers, and a sophisticated control system. Moreover, the system currently
requires lengthy installation with multiple hydraulic and electrical interconnects
and specialized expertise to utilize Dynamic Feed’s capabilities. These cost and
complexity issues must be eliminated.
The capabilities of the current system design is also limited by the multiple
valve interactions. This interaction produces undesirable effects on cavity
pressure, making it especially difficult to control the cavity pressure dynamics in
the packing stage. Several alternative system designs, listed below in Table
Chapter 7: -1 and illustrated in Appendix C, have been proposed to improve the
process capability while significantly reducing the system cost and complexity.
Table Chapter 7: -1: Alternate Solution Approaches of Appendix C
Design Advantage
Reverse taper Melt displacement consistent with valve
movement
Digital modulation Faster dynamics, no valve interaction
Rotary actuation Faster dynamics, no valve interaction
177
Commercial Validation
The experimental work performed thus far has demonstrated the process
flexibility and consistency enabled by Dynamic Feed for simple mold geometries.
While these capabilities have never before been possible, the technology is not
yet commercially ready. As such, ongoing validation research is proposed,
applying Dynamic Feed in a previous commercial application which experienced
difficulties during full-scale production. Such a demonstration is needed to
validate the Dynamic Feed Control principles at the level of complexity of
challenging commercial applications. Moreover, the experimental validation
should be expanded to include a variety of material properties and wide range of
process conditions.
178
140
Personal communication with John Rowland, General Manager, Moldflow Pty Ltd, Kilsyth, Victoria,
Australia (April, 1995).
179
References
TABLE OF CONTENTS
REFERENCES ..........................................................................................................................................190
189
References
References
1. Hyatt, J. W., A Celluloidal Material for Ivory Replacement, United States Patent
#892,300, 1869.
2. Friedel, R., The First Plastic, American Heritage of Invention and Technology, v. 2, n. 2,
pp. 18 (Summer, 1987).
3. Time magazine, Plastics in the War, v. 40, n. 9, pp. 72 (31 August, 1942).
4. McCrum, N.G., Buckley, C.P., Bucknall, Principles of Polymer Engineering, Oxford

Science Publications, Oxford, 1989.
5. Foy, G. F., Engineering Plastics and Their Commercial Development, American

Chemical Society, Washington D. C., 1969.
6. 1986 United States Industrial Outlook, United States Department of Commerce, Bureau
of Industrial Economics, Washington D. C., 1986.
7. Brown, R.H., Garter, J.E., 1994 United States Industrial Outlook, United States
Department of Commerce/International Trade Administration, Washington D. C., 1994.
8. Muccio, E.A., Plastic Part Technology, ASM International, Material Park, Ohio, 1992, pp.
176.
9. Rosenthal, S. R., Tatticonda, M. V., Time Management in New Product Development:

Case Study Findings, IEEE Engineering Management Review, v. 21, n. 3, pp. 13, 1993.
10. Halper, M., Apple Delays Adding PowerPC to Portables, Computerworld, v. 28, n. 47, pp.
6 (1994).
11. Beal, C. I., Concept to Production in Three Weeks - Is it Possible?, Proceedings from the
1993 Annual Technical Meeting of the Society of Plastics Engineers, v. 51, pp. 3158
(1993).
12. Groover, M. P., Zimmers, E. W., Computer-Aided Design and Manufacturing, Prentice-
Hall, Englewood Cliffs, NJ, 1984.
13. Bernhardt, E. C., Computer Aided Engineering for Injection Molding, Hanser Publishers,
Munchen, Germany, 1983.
14. Personal communication with Jack Watts, President and CEO of Cap-Snap Moldings,
San Jose, California (April 11, 1994).
190
References
15. Noaker, P. M., Mirror Image Moldmaking, Manufacturing Engineering, v. 113, n. 3, pp.
36-41 (September, 1994).
16. Muccio, pp. 181.
17. Spencer, R. S., Gilmore, G. D., Some Flow Phenomena in the Injection Molding of
Polystyrene, Journal of Colloidal Science, v. 6, pp. 118 (1951).
18. Harry, D., Parrot, R., Numerical Simulation of Injection Mold Filling, Polymer Engineering
and Science, v. 10, pp. 209 (1970).
19. Williams, G., Lord, H., Mold-Filling Studies for the Injection Molding Process with PVC in
Injection Molding, Polymer Engineering and Science, v. 15, pp. 553 (1975).
20. Austin, C., Chapter 9 of CAE-Computer Aided Engineering for Injection Molding, E.C.
Berhnhardt (ed.), Hanser Publishers, Munchen, Germany, 1983.
21. Hieber, C.A., Shen, S. F., Flow Analysis of the Non-Isothermal Two-Dimensional Filling
Process in Injection Molding, Israel Journal of Technology, v. 16, pp. 248 (1978).
22. Wang, V. W., Hieber, C. A., Wang, K. K., Filling of an Arbitrary Three Dimensional Thin
Cavity, Journal of Polymer Engineering, v. 7, pp. 21 (1986).
23. Personal Communication with Suresh Shah, Project Manager, GM Inland-Fisher Guide,
Troy, Michigan (November 11, 1994).
24. Austin, C., Industrial Metamorphosis, Proceedings from the 1994 Annual Technical
25. Kamal, M., Goyal, S., Chu, E., Simulation of Injection Mold Filling of a Viscoelastic
Polymer with Fountain Flow, The American Institute of Chemical Engineers, v. 34, pp. 94
(1988).
26. Cross, M., Relation between Viscoelasticity and Shear Thinning Behavior in Liquids,
Rheological Acta, v. 18, pp. 609 (1979).
27. Hu, J., Vogel, J.H., Dynamic Modeling and Control of Packing Pressure in Injection
Molding, Journal of Engineering Materials and Technology, v. 116, n. 2 (April 1994).
28. Chiu, W.-Y., Wang, C., Wang, D.-C., Analysis of Filling, Packing and Cooling Stages in
Injection Molding of Disk Cavities, Journal of Applied Polymer Science, v. 43, n. 1 (July
1991).
29. Rezayat, M., Stafford, R.O., A Thermoviscoelastic Model for Residual Stress in Injection
Molded Thermoplastics, Polymer Engineering and Science, v. 31, n. 6 (March 1991).
30. Akbar, S., Altan, M. C., On the Solution of Fiber Orientation in Two-Dimensional
Homogeneous Flows, Polymer Engineering and Science, v. 32, n. 12, (June 1992).
191
References
31. Greener, J., Pearson, G., Orientation, Residual Stresses, and Birefringence in Injection
Molding, Journal of Rheology, v. 27, pp. 115 (1983).
32. Tseng, S.C., Osswald, T.A., Predicting Shrinkage and Warpage of Fibre-Reinforced
Composite Parts, Polymer Composites, v. 15, n. 4 (August 1994).
33. Matsuoka, T., Takabatake, J.I., Koiwai, A., Integrated Simulation to Predict Warpage of
Injection Molded Parts, Polymer Engineering and Science, v. 31, n. 14 (July 1991).
34. Stempnik, L., J., Automatic Control for Plastic Machinery, Proceedings of the National
Conference on Fluid Power, v. 28, pp. 539 (1972).
35. Ma, C., Y., A Design Approach to A Computer-Controlled Injection-Molding Machine,

Polymer Engineering and Science, v. 14, n. 11, pp. 768 (1974).
36. O’Bryan, J. E., Proportional Valves, Microprocessors, and Closed-Loop Control Keeps
Plastics Molders Competitive, Hydraulic and Pneumatics, v. 42, n. 3, pp. 95 (1989).
37. Mann, J., W., Process Parameter Control: The Key to Optimization, Plastics Engineering,
pp. 25 (January, 1974).
38. Okeke, E. J., Cosma, L., Dimensional Repeatability of Gas Assisted Injection Molding
and Cavity Pressure Controlled Closed Loop Injection Molding for Structural Parts,
Proceedings from the 1993 Annual Technical Meeting of the Society of Plastics
Engineers, v. 51, pp. 79 (1993).
39. Shankar, A., Dynamic Modeling and Control of Injection Molding Machines, Doctoral
Dissertation submitted to Carnegie-Mellon University Department of Mechanical
Engineering, 1978.
40. Agrawal, A. R., Pandelidis, I. O., Pecht, M., Injection-Molding Process Control — A
Review, Polymer Engineering and Science, v. 27, p. 18 (1987).
41. Nunn, R. E., Grolman, C. P., Adaptive Process Control for Injection Molding, Journal of
Reinforced Plastics and Composites, v. 9, pp. 2121 (1991).
42. Gao, F., Patterson, W. I., Kamal, M. R., Self-Tuning Cavity Pressure Control of Injection
Mold Filling, Advances in Polymer Technology, v. 13, n. 2, pp. 111 (1994).
43. Smud, S. M., Harper, D. O., Deshpande, P. B., Advanced Process Control for Injection
Molding, Polymer Engineering and Science, v. 31, n. 15 (1991).
44. Chiu, C. P., Wei, J. H., Shih, M. C., Adaptive Model following Control of the Mold Filling
Process in an Injection Molding Machine, Polymer Engineering and Science, v. 31, n. 15
(1991).
45. Seaman, C., M., A Multiple Objective Optimization Approach to Quality Control, Doctoral
Dissertation submitted to Rensselaer Polytechnic Institute Department of Electrical,
Computer, and Systems Engineering, 1991.
192
References
46. Rowland, J. C., Ho-Le, K., Process Quality Assurance for Injection Molding of
Thermoplastic Polymers Proceedings from the 1994 Annual Technical Meeting of the
Society of Plastics Engineers, v. 52, pp. 342 (1994).
47. De Gaspari, J. D., Melt-Flow Oscillation Improves Part Properties, Plastics Technology,
pp. 78 (March, 1994).
48. Ibar, J. P., Rheomolding – A New Process to Mold Polymeric Materials, Proceedings
from the 1994 Annual Technical Meeting of the Society of Plastics Engineers, v. 52, pp.
3034 (1994).
49. Vernyi, B., New Method Improves Plastics Mechanically, Plastics News, pp. 33 (October
18, 1993).
50. Gardner, G., Malloy, R., A Moving Boundary Technique to Strengthen Weld Lines in
Injection Molding, Proceedings from the 1994 Annual Technical Meeting of the Society of
51. Becker, H., Fischer, G., Muller, U., Push-Pull Injection Moulding of Industrial Products,
Kunstoffe German Plastics, v. 83, n. 3 (1993).
52. Parekh, S., Desai, S., Brizzolara, J., Monitoring the Multi-Live Feed Injection Molding
Process with Cavity Instrumentation, Proceedings from the 1994 Annual Technical
53. Ogbonna, C., I., Production of Self-Reinforced Polyethylene using the Multi Live-Feed
(Injection) Moulding Technique, Doctoral Dissertation submitted to Brunel University
Department of Materials Technology, 1989.
54. Kazmer, D. O., Roe, D. S., Increasing Weld-Line Strength through Dynamic Control of
Volumetric Shrinkage, Proceedings from the 1994 Annual Technical Meeting of the
55. De Gaspari, J. D., Low-Pressure Alternatives for Molding Large Automotive Parts,
Plastics Technology, v. 39, n. 10, (September 1, 1993).
56. Rusch, K. C., Gas-Assisted Injection Molding — A New Technology is Commercialized,

Plastics Engineering, v. 35, (July, 1989).
57. Okeke, E. J., Cosma, L., Dimensional Repeatability of Gas Assisted Injection Molding
and Cavity Pressure Controlled Closed Loop Injection Molding for Structural Parts,
Proceedings from the 1993 Annual Technical Meeting of the Society of Plastics
Engineers, v. 51, pp. 79 (1993).
58. Molding Impossible Possibilities, Injection Molding Magazine, October, 1994.
59. Binary Injection Molding Machine, Article received from Dan Furlano, Applications
Engineer, GE Plastics, Portland, Oregon (May 15, 1994).
193
References
60. Trantina, G. G., Ysseldyke, D. A., An Engineering Design System for Thermoplastics,
1989 Society of Plastics Engineers Annual Technical Conference Proceedings, v. 48, pp.
635-639 (1989).
61. Spencer, R., Dillon, R., The Viscous Flow of Molten Polystyrene, Journal of Colloidal
Science, v. 3, pp. 163 (1948).
62. Woodruff, D., Bug Control at Chrysler, Business Week, August 22, 1994, pp. 26.
63. Personal communication with Ken Debronsky, Vice President, Mustang Division of Ford
Motor Company, Palo Alto, California (December, 1993).
64. Poslinski, A. J., Aslam, S., Kazmer, D. O., The Effects of Viscosity Variation on Injection
Molding, GE Research & Development Technical Information Series, 92CRD146, 1992.
65. Personal communication with Dan Furlano, Design Engineer, GE Plastics, Portland,
Oregon (May 15, 1994).
66. Personal communication with Kosuke Ishii, Associate Professor, Stanford University
(October 11, 1994).
67. Suh, N.P., Wilson, D.R., Bell, A.C., Van Dyck, F., Tice, W.W., Manufacturing Azioms and
Their Corollaries, Presented at the Society of Manufacturing Engineer’s Seventh North
American Metalworking Research Conference, pp. 113 (May, 1979).
68. Burgeson, J., Tight Tolerance Design, Plastics Engineering, v. 47, n. 5, pp. 23, (1991).
69. Suh, N.P., Bell, A.C., Gossard, D.C., On an Axiomatic Approach to Manufacturing and
Manufacturing Systems, Presented at the American Society of Mechanical Engineers
Winter Annual Meeting, pp. 124 (December 1977).
70. O’Neill, M. E., Chorlton, F., Viscous and Compressible Fluid Dynamics, John Wiley &
Sons, New York, 1989.
71. Nagy, M. R., Long-Term Shrinkage of Polypropylene, Proceedings from the 1993 Annual
72. Schlitling, H., Boundary Layer Theory, Seventh Edition, McGraw-Hill, New York, New
York, 1979.
73. Lee, H.S., Thin-Cavity Filling Analysis using the Finite Element Method with Control
Volume Techniques, Doctoral Dissertation to Rensselaer Polytechnic Institute
Department of Mechanical Engineering, 1989.
74. Chiang, H.H., Simulation and Verification of Filling and Post-Filling Stages of the Injection
Molding Process, Doctoral Dissertation to Cornell University Sibley School of Mechanical
Engineering, 1989.
194
References
75. Kazmer, D. O., Willey, S. J., An Examination of Material Characterization and Modeling
Techniques for Injection Molding Simulation, Proceedings from the 1992 Annual
76. Williams, M. L., Landel, R. F., Ferry, J. D., A Model for the Viscosity of Molten
Polystyrene, Journal of American Chemical Society, v. 77, n. 8, pp. 3701 (1955).
77. Zoller, P., Bolli, P., Pahud, V., Ackermann, H., Apparatus for Measuring Pressure-
Volume-Temperature Relationships of Polymers to 350 °C and 2200 kg/cm2, Review
Scientific Instrumentation, v. 47, pp. 948 (August 1976).
78. Zoller,P., Pressure-Volume-Temperature Relationships of Solid and Molten

Polypropylene, Journal of Applied Polymer Science, v. 23, pp. 1057 (1979).
79. Kamal, M.R., Lafleur, P.G., Computer Simulation of Injection Molding, Polymer
Engineering and Science, v. 22, n. 17, pp. 1069 (1982).
80. Seborg, D.E., Edgar, T.F., Shah, S.L., Adaptive Control Strategies for Process Control: A
Survey, AICheJ, A Publication of the American Institute of Chemical Engineers, v. 32, pp.
881-913 (June 1986).
81. Sanschagrin, B., Process Control for Injection Molding, Polymer engineering and
Science, v. 23, n. 8, pp. 431 (1983).
82. Korem, Y., Adaptive Control Systems for Machining, Manufacturing Review, v. 2, n. 1,
pp. 6 (March 1989).
83. deVries, B. A. J., Verbruggen, H. B., Multivariable Process and Prediction Models in
Control, International Journal of Adaptive Control, v. 8, n. 2, pp. 261, (1994).
84. Roffel, B., Vermeer, P.J., Chin, P.A., Simulation and Implementation of Self-Tuning
Controllers, Prentice Hall, Englewood Cliffs, New Jersey, 1989.
85. Butler, H., Model Reference Adaptive Control, Prentice Hall-Hemel Hapstead,
Hertfordshire, 1992.
86. Sastri, S.S., Isidori, A., Adaptive Control of Linearizable Systems, IEEE Transactions on
Automatic Control, v. AC-34, pp. 1123 (1989).
87. Franklin, G.F., Powell, J.D., Workman, M.L., Digital Control of Dynamic Systems,
Addison-Wesley, New York, New York, 1992, pp. 462.
88. Franklin, G.F., Powell, J.D., Workman, M.L., Digital Control of Dynamic Systems,
Addison-Wesley, New York, New York, 1992, pp. 468.
89. Ljung, L., System Identification Toolbox, The Math Works, Cambridge, Massachussets,
1988.
195
References
90. Franklin, G. F., Powell, J. D., Emami-Naeini, A., Feedback control of Dynamic Systems,
Addison-Wesley, New York, New York, 1988.
91. Temperature Control of Kona VG Series Manifold, Kona Corporation Manifold Design
Manual, Kona Corporation, Gloucester, MA, 1994.
92. Time Response of the PT-XL Amplified Pressure Transducer, Dynisco Instruments
Technical Literature, Dynisco Instruments, Sharon, Ma, 1988.
93. Mold Shrinkage Estimates, GE Plastics Lexan™ Resin Design Guide, GE Plastics,
Pittsfield, MA, p. 31, (1987).
94. Box, G.E.P., Hunter, W.G., Hunter, J.S., Statistics for Experimenters, John Wiley & Sons,
New York, 1978.
95. RS/Explore, Release 3.01, Sub-module of RS/1, Release 5.0.1, Used under license from
Bolt Bareneck and Newmann, Inc., 1994.
96. Beris, M. C., Standards and Practices of Plastics Custom Molders, SPI Plastics
Engineers Handbook of the Society of the Plastics Industry, 5th Edition, Van Nostrand
Reinhold Co., New York, 1991.
97. Farnum, N.R., Modern Statistical Quality Control and Improvement, Duxbury Printers,
Belmont, California, 1994.
98. Wang, V.W., Hieber, C. A., Wang, K. K., Dynamic Simulation and Graphics for the
Injection Molding of Three-Dimensional Thin Parts, Journal of Polymer Engineering, v. 7,
n. 1, pp. 21-45 (1986).
99. Kim, B., Low Thermal Inertia Injection Molding, Doctoral Dissertation submitted to the
Massachusetts Institute of Technology Department of Mechanical Engineering, 1983.
100. Personal conversation with Gary Stanton, Field Service Representative, Mitsubishi
Heavy Industries, Learson, Texas, Pleasanton, California (February 11, 1995).
101. Dixon, J.R., Howe, A., Cohen, P.R., Simmons, M.K., Dominic I: Progress towards
Domain Independence in Design by Iterative Redesign, Proceedings of the 1986 ASME
Computers in Engineering, v. 4, pp. 199, 1986.
102. Personal communication with Larry Leifer, Professor, Stanford University (January,
1994).
103. Rinderle, J.R., Implications of Function-Form-Fabrication Relations on Design

Decomposition Strategies, Proceedings of the 1986 ASME Computers in Engineering, v.
4, pp. 193, (1986).
104. Lorenzo, L., McIntyre, R. K., Engineering design and Optimization of a Thermoplastic
Mower Deck, Proceedings from the 1994 Annual Technical Conference of the Society of
196
References
105. Kyle, B.R.M., Using Taguchi Methodology for Optimizing Products and Processes,
Proceedings from the 1989 Annual Technical Conference of the Society of Plastics
Engineers, v. 47, pp. 1704 (1989).
106. Taguchi, G., Tsai, S.C., Taguchi on Robust Technology Development, ASME Press, New
York, 1993.
107. Lewis, L., Parkinson, A., Robust Optimal Design Using a Second Order Tolerance Model,
Research in Engineering Design, v. 6, pp. 25 (1994).
108. Otto, K. N., Antonsson, E. K., The Method of Imprecision Compared to Utility Theory for
Design Selection Problems, Design Theory and Methodology, v. 53, pp. 167 (1993).
109. Eggert, R. J., Mayne, R. W., Probabilistic Optimal Design Using Successive Surrogate
Probability Density Functions, Journal of Mechanical Design, v. 115, pp. 385 (1993).
110. Eggert, R. J., Quantifying Design Feasibility Using Probabilistic Feasibility Analysis,
Advances in Design Automation, v. 32, n. 1, p. 235 (1991).
111. Siddall, J. N., Optimal Engineering Design: Principles and Applications, Marcel-Dekker,
Inc, New York, 1982.
112. Balachandran, M., Knowledge-Based Optimum Design, Computational Mechanics

Publications, Hobbs the Printers Ltd., Southampton, Great Britain, 1993.
113. Ford, R. B., Quality Improvement of Mechanical Systems through Design Innovation at
the Conceptual Phase, Doctoral Dissertation submitted to the Stanford University ME
Design Division, 1995.
114. DeVor, R. E., Chang, T., Sutherland, J. W., Statistical Quality Design and Control:
Contemporary Concepts and Methods, Macmillan Publishing Company, New York, 1992.
115. Grove, Mock, Formal Definition and Characteristics of Measurement Systems: A Factual
View of Measurement, Chapter 2 of Measurement, Accounting, and Organizational
Information, John Wiley, New York, 1979.
116. Smith, G., Statistical Process Control and Quality Improvement, Merrill Publishing Co.,
New York, 1991.
117. Balf, T., Motorola Nears Quality Benchmark after Twelve Year Evolutionary Effort,
Aviation Week and Space Technology, v. 135, n. 23, pp. 64 (December 9, 1991).
118. Siddall, J. N., Probabilistic Engineering Design: Principles and Applications, Marcel
Dekker, New York, 1983.
119. Takeshima, M., Funakoshi, N., Molecular Orientation Distribution in Injection Molded
Polycarbonate Disks, Journal of Applied Polymer Science, v. 32, pp. 3457, (1986).
197
References
120. Personal communication with D. Lee, Professor of Mechanical Engineering, Rensselaer

Polytechnic Institute, May, 1991.
121. Yokoi, H., Nagami, S., Kawasaki, A., Murata, Y., Visual Analyses of Flow Marks
Generation Process Using Glass-Inserted Mold, Proceedings from the 1994 Annual
Technical Conference of the Society of Plastics Engineers, v. 52, pp. 368 (1994).
122. Hawkins, W.L., Polymer Degradation and Stabilization, Springer-Verlag, Berlin,

Germany, 1984.
123. Irani, R.K., Integrating Features, Iterative Redesign, and CAE Technology to Automate
the Design of the Feed System of Injection Molds, Doctoral Dissertation submitted to the
Department of Mechanical Engineering, University of Massachusetts at Amherst, 1989.
124. Fisa, B., Rahmani, M., Weldline Strength in Injection Molded Glass Fiber Reinforced
Polypropylene, Polymer Engineering and Science, v. 31, n. 18, pp. 1330 (1991).
125. Roe, D. S., The Effect of Process Conditions on Knit-Line Strength, Masters Thesis
submitted to University of Lowell Department of Plastics Engineering, 1992.
126. Kazmer, D. O., Roe, D. S., Increasing Weld-Line Strength through Dynamic Control of
Volumetric Shrinkage, Proceedings from the 1994 Annual Technical Meeting of the
127. Väätäinen, O., Järvelä, P., Valta, K., Järvelä, P., The Effect of Processing Parameters on
the Quality of Injection Moulded Parts by using the Taguchi Parameter Design Method,
Plastics, Rubber and Composites Processing and Applications, v. 21, n. 4, pp. 211
(1994).
128. Siegmann, A., Buchman, A., Kenig, S., Residual Stresses in Polymers III: The Influence
of Injection-Molding Process Conditions, Polymer Engineering and Science, v. 22, n. 9,
pp. 560 (1982).
129. Bushko, W. C., Stokes, V. K., The Effects of Differential Mold Temperatures on the
Warpage of Packed Injection-Molded Parts, Proceedings from the 1994 Annual
Technical Conference of the Society of Plastics Engineers, v. 52, pp. 506 (1994).
130. Personal communication with Paul Choineire, Technical Development Engineer, GE

Plastics, Pleasanton, California (January 16, 1994).
131. Personal communication with Anthony Cavotta, Engineering Manager, Trend Plastics,
Fremont, California (March 7, 1995).
132. Beris, M. C., Standards and Practices of Plastics Custom Molders, SPI Plastics
Engineers Handbook of the Society of the Plastics Industry, 5th Edition, Van Nostrand
Reinhold Co., New York, 1991
133. Schulz, M., Alig, I., Influence of Stochastic Environments of Gaussian Chains on
Dynamic Shear and Bulk Properties, Journal of Chemical Physics, v. 97, n. 4, pp. 2772
(1992).
198
References
134. Davidon, W. C., Optimally Conditioned Optimization Algorithms without Line Searches,
Math. Prog., v. 9, n. 1, pp. 1, (1975)
135. Fletcher, R., Powell, M. J., A Rapidly Convergent Descent Method for Minimization,
Computer J., v. 6, n. 2, pp. 163 (1963).
136. Morse, W.J., Measuring, Planning, and Controlling Quality Costs, National Association of
Accountants, Montvale, New Jersey, 1987.
137. Campanella, J., Corcoran, F., Principles of Quality Costs, 16, 16, 1983.
138. Morris, N. M., Rouse, W. B., The Effects of Type of Knowledge upon Human Problem
Solving in a Process Control Task, IEEE Trans. Syst., Man, and Cybernetics, v. SMC-15,
pp. 698, (November, 1985).
139. Moray, N., Lootsteen, P., Pajak, J., Acquisition of Process Control Skills, IEEE Trans.
Syst., Man, and Cybernetics, v. SMC-16, pp. 697, (July, 1986).
140. Personal communication with John Rowland, General Manager, Moldflow Pty Ltd,
Kilsyth, Victoria, Australia (April, 1995).
199
TABLE OF CONTENTS
A-1 MATLAB ADAPTIVE CODE.............................................................................................................179
A-2 MATLAB CLOSED LOOP SIMULATION........................................................................................180
A-3 MATLAB CODE FOR OPEN LOOP COMPARISON .......................................................................181
B-1 DATA FOR REGRESSION OF CONVENTIONAL PROCESS.........................................................182
B-2 DATA FOR REGRESSION OF DYNAMIC FEED.............................................................................183
C-1 REVERSE TAPER...............................................................................................................................184
C-2 DIGITAL MODULATION ..................................................................................................................185
C-3 ROTARY ACTUATION......................................................................................................................186
D-1 MONTE CARLO SIMULATION........................................................................................................187
D-2 RUNROBUST.XLS .............................................................................................................................188
D-3 ROBUST.XLS .....................................................................................................................................189
START OF CHAPTER: 179

Appendix A: Matlab Metafile Code
Appendix B: Experimental Data for 220 mm x 50 mm Plaque
Appendix C: Alternative Valve Designs
Appendix D: Monte Carlo Simulation
1
Appendices
A-1 MATLAB ADAPTIVE CODE

function run(Tstart,Pslope,Phold)
% adaptive closed loop simulations to obtain desired cavity pressure trace
% inputs define start of fill, slope of pressure, and pack pressure level
global UU TT Pin X Pe Px
Tswitch=Tstart+2.6-0.03*Pslope;
Xpack=.010+.00050*Phold;
Tpack=4.0;
dt=0.02;
Tend=6.98;
Pslope=Pslope*1e6;
Phold=Phold*1e6;
goT=1;goX=1;
xp=1;ii=0;
while (goT+goX)>0,
% Run the Closed Loop Simulation

close(Tstart,Pslope,Phold,Tswitch,Xpack,Tpack,Tend,dt)
% Adapt swtictching time and valve positions

if abs(Pe)<6e6 goT=0;
if abs(Px)<4e6 goX=0;
else ddx=-.00050*xp*Px/1e6,Xpack=Xpack+ddx,end;
elseif abs(Pe)<25e6 ddt=-0.004*xp*Pe/1e6,Tswitch=Tswitch+ddt,
else ddt=-0.007*xp*Pe/1e6;Tswitch=Tswitch+ddt,end;
% If difficulty in adapting, cut down step size

ii=ii+1;
if ii>10 xp=xp*0.5;ii=0;Tswitch=Tstart+2.6-0.03*Pslope/1e6;
Xpack=.010+.00050*Phold/1e6;goT=1;goX=1;end;
if Xpack<0 xp=xp*0.5;ii=0;Tswitch=Tstart+2.6-0.03*Pslope/1e6;
if Tswitch<Tstart xp=xp*0.5;ii=0;Tswitch=Tstart+2.6-0.03*Pslope/1e6;
if Tswitch>5.0 xp=xp*0.5;ii=0;Tswitch=Tstart+2.6-0.03*Pslope/1e6;
% Plot results
figure(1),clf
plot(TT,Pin/1e6,'m-',TT,X(:,2)/1e6,'m-',TT,1000*X(:,3),'y-',TT,100*UU(2,:),'w--');
xlabel('Time (sec)');
ylabel('Cavity Pressure (MPa) & Valve Position (mm)');
end;
179
Appendices
A-2 MATLAB CLOSED LOOP SIMULATION

function dfc(Tstart,Pslope,Phold,Tswitch,Xpack,Tpack,Tend,dt)
% closed loop simulation of packing and filling stages
% called by adaptive procedure run.m
global UU TT Pin X Pe Px
c=0.01;
cc=c*c;
wh=5*.2*cc; % Cross-sectional area of cavity in m*m
Ahyd=3.141*8*8*cc; % Area of injection cylinder in m*m
Aram=0.1*Ahyd; % Area of front of ram in m*m
Avalve=3.141*0.5*0.5*cc; % Area of valve stem in m*m
Mram=60; % Mass of ram in kg
Bp=4.9e-4; % Compressibility in cc/g/MPA
Bt=100*4.1e-4; % Thermal shrinkage in cc/g/C
Bmelt=9e8; % Melt compressibility - units??
Kvalve=800*12*c*2/(2*3.141*0.4*c^4); % Valve resistance for 1 cm gap
Kcavity=800*12/(5*c^5*0.2^3); % Cavity resistance for 1 meter flow length
%states are Vram, Pcavity, Xvalve, Lflow --- inputs are Phydraulic, Vvalve
% measured in m/s, MPa, m, and m --- MPa and m/s
Pin=[];UU=[];TT=[];sint=0;DP=0;
X=[0.005 .2e6 0.001 0.001]; % Initial conditions: Vram, Pcavity, Xvalve, Lflow
A=[0 -Aram/Mram 0 0; % NOTE: Zero entries are computed in UpdatePlant
0 0 0 0;
0 0 0 0;
0 0 0 0];
B=[Ahyd/Mram 0;
0 0;
0 1;
0 0];
for i=2:1:(Tend/dt)
TT(i)=(i-1)*dt;time=TT(i);
UU(1,i)=10.0; % Hydraulic pressure in MPa
Pin(i)=Phold;
if time<Tswitch Pin(i)=Pslope*(time-Tstart);end;
if time<Tstart Pin(i)=0;end;
if X(i-1,2)<0 X(i-1,2)=0.1;end;
if X(i-1,3)<=0.001 X(i-1,3)=0.001;end;
if X(i-1,3)>0.1 X(i-1,3)=0.1;end;
Rvalve=Kvalve/X(i-1,3)^3; % Valve resistance
Rcavity=Kcavity*X(i-1,4); % Cavity resistance
Vmelt=X(i-1,4)*wh; % Melt volume in m*m*m
Cmelt=Bmelt/Vmelt; % Useful compressibility factor
if X(i-1,4)>0.2 X(i-1,4)=0.2;Rcavity=1e15;full=1;end;
A(1,1)=-Rvalve*Aram*Aram/Mram;A(2,1)=Aram*Cmelt;A(2,4)=-Bt*wh*Cmelt;
A(2,2)=-Cmelt/Rcavity;A(4,2)=1/(wh*Rcavity);B(2,2)=-0.5*Cmelt*Avalve;
if i>4 DP=(X(i-1,2)-X(i-3,2))/(2*dt);end;
sint=0.99*sint+(Pin(i)-X(i-1,2));
if time < Tswitch
UU(2,i)=0.0175*(Pin(i)-X(i-1,2))/1e6+.07-1*X(i-1,3)+0.0005*(Pslope-DP)/1e6;end;
if time >= Tswitch
UU(2,i)=0.0015*(Pin(i)-X(i-1,2))/1e6+1*(Xpack-X(i-1,3))-0.001*DP/1e6;end;
if time < (Tstart-1) UU(2,i)=0;end
if time >Tswitch if time<(Tswitch+0.5)
UU(2,i)=8*(Xpack-X(i-1,3));Pe=X(i-1,2)-Phold;end;end;
if time < (Tswitch+0.5*Tpack) Px=X(i-1,2)-Phold;end;
if UU(2,i)>0.2 UU(2,i)=0.2;end;
if UU(2,i)<-.2 UU(2,i)=-0.2;end;
if X(i-1,3)<=0.001 if UU(2,i)<=0 UU(2,i)=0;end;end;
if time>(Tswitch+Tpack) UU(2,i)=-0.006;end;
[a,b]=c2d(A,B,dt);X(i,:)'=a*X(i-1,:)'+b*UU(:,i)'; % Advance time step
end
180
Appendices
A-3 MATLAB CODE FOR OPEN LOOP COMPARISON

function dfc
% open loop simulation of model
lt=9.6; % Length of simulation
dt=0.05; % Time step
c=0.01; % Conversion factor from cm to m
cc=c*c; % Conversion factor from cm*cm to m*m
Lcav=0.1; % Length of rectangular cavity in m
wh=5*.3*cc; % Cross-sectional area of cavity in m*m
Ahyd=3.141*8*8*cc; % Area of injection cylinder in m*m
Aram=0.1*Ahyd; % Area of front of ram in m*m
Avalve=3.141*0.5*0.5*cc; % Area of valve stem in m*m
Mram=60; % Mass of ram in kg
Bp=4.9e-4; % Compressibility in cc/g/MPA
Bt=100*4.1e-4; % Thermal shrinkage in cc/g/C
Bmelt=9.2e8; % Melt compressibility
Vis=400; % Material viscosity in Pa Sec
Kvalve=Vis*12*c*2/(2*3.141*0.4*c^4); % Valve resistance for 1 cm gap
Kcavity=Vis*12/(5*c^5*0.3^3); % Cavity resistance for 1 meter flow length (3mm)
% Process inputs: hydraulic pressure and valve velocity

U(1,:)=[0 5 58 60 62 66 89 94 100 82 78 75 74 73 72 71 70 70 70 70 70 70 10 0]*1e6; %MPa
U(2,:)=[.113 .02 0.006 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; %m/s
T=(lt*(0:(length(U(1,:))-1))/(length(U(1,:))-1))';
%Initialization: states are Vram, Pcavity, Xvalve, Lflow

X=[.002 .2e6 0.001 0.001];
A=[0 -Aram/Mram 0 0; 0 0 0;0 0 0 0;0 0 0 0];
B=[Ahyd/Mram 0;0 0;0 1;0 0];
for i=2:1:(lt/dt)
TT(i)=(i-1)*dt;time=TT(i);
UU(1,i)=INTERP1(T,U(1,:),TT(i));
UU(2,i)=INTERP1(T,U(2,:),TT(i));
if X(i-1,2)<0 X(i-1,2)=0.1;end;
if X(i-1,3)<=0.001 X(i-1,3)=0.001;end;
if X(i-1,3)>0.5 X(i-1,3)=0.5;end;
Rvalve=Kvalve/X(i-1,3)^3; % Valve resistance
Rcavity=Kcavity*X(i-1,4); % Cavity resistance
Vmelt=X(i-1,4)*wh; % Melt volume in m*m*m
Cmelt=Bmelt/Vmelt; % Useful compressibility factor
if X(i-1,4)>Lcav X(i-1,4)=Lcav;Rcavity=1e15;end; % Cavity is full
A(1,1)=-Rvalve*Aram*Aram/Mram;
A(2,1)=Aram*Cmelt;
A(2,4)=-Bt*wh*Cmelt;
A(2,2)=-Cmelt/Rcavity;
A(4,2)=1/(wh*Rcavity);
B(2,2)=-.75*Cmelt*Avalve;
X(i,:)=sim(A,B,UU(:,i)',dt,X(i-1,:)');
if TT(i)<0.5,X(i,2)=0;end;
end
clf,plot(TT,X(:,2)/10e6,'w:');hold on; % Plot simulation results

fin=fopen('1_7.1','r'); % Get process data
fgetl(fin);fgetl(fin);fgetl(fin);line=fgetl(fin);disp(line)
np=size(findstr(line,' '),2)+1;D=zeros(1,np);
i=1;
while ~feof(fin)
i=i+1;D(i,:)=fscanf(fin,'%f',np)';
end
fclose(fin);
plot(D(:,1),D(:,3)/145,'w-'); % Plot process data
axis([0 9.6 0 100]);xlabel('Time (sec)');ylabel('Cavity Pressure (MPa)')
181
Appendices
B-1 DATA FOR REGRESSION OF CONVENTIONAL PROCESS

Run Melt Mold Speed Pack Slope1 Pack1 Dim1 Slope2 Pack2 Dim2
1 0 0 0 0 533.4173 6622.453 6.4 1263.591 6816.021 7
1 0 0 0 0 556.0501 6666.787 9.4 1342.623 6847.013 9.6
1 0 0 0 0 611.7136 6652.726 9.2 1318.492 6826.438 9.5
1 0 0 0 0 551.2293 6667.25 9.2 1321.607 6843.041 9.2
1 0 0 0 0 564.3687 6639.763 7.4 1307.673 6828.808 7.5
2 0 0 1 1 561.246 7677.658 6.4 1302.192 7949.794 8.5
2 0 0 1 1 554.0664 7694.406 9 1259.11 7950.642 8.8
2 0 0 1 1 544.2932 7698.277 6 1249.037 7959.974 6.3
2 0 0 1 1 578.6433 7642.819 9 1098.543 7921.74 8.8
2 0 0 1 1 532.9829 7642.08 7.5 1187.861 7900.652 9
3 0 1 0 1 443.1286 7170.369 8.8 984.499 7467.402 9.2
3 0 1 0 1 414.1808 7184.964 8.2 952.3367 7478.603 9.2
3 0 1 0 1 468.4952 7177.651 8.5 1031.893 7485.512 8.5
3 0 1 0 1 502.9017 7217.142 8.5 1021.99 7497.834 9
3 0 1 0 1 487.4807 7208.214 8.5 977.3322 7502.136 9.25
4 0 1 1 0 591.1052 6691.181 9.8 1253.151 6883.445 9.8
4 0 1 1 0 559.1875 6700.137 8.6 1194.902 6900.077 9.5
4 0 1 1 0 541.3352 6699.357 9.6 1236.283 6898.535 9.5
4 0 1 1 0 556.3498 6710.76 9.2 1154.113 6905.284 9.6
4 0 1 1 0 559.9596 6684.694 10 1153.326 6872.786 6
5 1 0 0 0 640.2168 7666.36 9.4 1299.737 7856.322 8.2
5 1 0 0 0 719.4447 7630.886 8.6 1236.073 7811.693 8.4
5 1 0 0 0 917.5229 7630.941 9.5 1219.507 7822.277 8.8
5 1 0 0 0 690.2573 7611.827 9.5 1315.302 7808.559 8.8
5 1 0 0 0 699.4974 7601.822 9 1162.906 7816.326 8.8
6 1 0 1 1 1254.922 6789.352 9 1323.741 6954.754 9.6
6 1 0 1 1 866.5921 6769.1 9.6 1483.237 6936.567 9.5
6 1 0 1 1 852.7873 6769.912 10.4 1438.567 6918.252 9.6
6 1 0 1 1 1224.015 6795.898 9.8 1406.99 6953.764 9.4
6 1 0 1 1 1081.078 6790.593 9.8 1439.994 6948.997 8.8
7 1 1 0 1 585.4033 6669.602 11 1198.891 6772.937 10.5
7 1 1 0 1 553.0745 6661.201 10.1 1238.263 6766.128 10.2
7 1 1 0 1 646.0797 6654.644 11 1319.747 6767.785 10.2
7 1 1 0 1 689.7885 6638.808 11 1276.958 6749.106 7.5
7 1 1 0 1 719.0882 6667.622 9.2 1292.312 6767.79 8.8
8 1 1 1 0 1555.559 7940.371 9 1507.419 8070.995 9.6
8 1 1 1 0 1010.375 7952.394 7 1378.782 8074.67 7.5
8 1 1 1 0 990.0442 7952.537 7.5 1404.544 8075.716 8.5
8 1 1 1 0 1405.246 7931.54 6.8 1461.236 8053.965 9.5
8 1 1 1 0 1040.267 7925.715 10 1432.057 8064.293 9.5
Average 8.91 Average 8.88625
Std Dev. 1.246904 Std Dev. 0.986641
Max 2.09 Max 1.61375
Min -2.91 Min -2.88625
182
Appendices
B-2 DATA FOR REGRESSION OF DYNAMIC FEED

Run Melt Mold Speed Pack Slope1 Pack1 Dim1 Slope2 Pack2 Dim2
1 0 0 0 0 1014.358 3563.296 18 738.3499 3140.143 17.3
1 0 0 0 0 1020.025 3340.57 18.4 741.7952 2877.231 17.6
1 0 0 0 0 1065.88 3486.451 18.5 745.5297 3100.389 17.7
1 0 0 0 0 1185.879 3549.631 17.7 802.7811 3135.106 17
1 0 0 0 0 1088.623 3401.006 17.5 843.0025 3030.66 17.1
2 0 0 1 1 1122.692 3334.142 17.6 850.8141 3398.473 17.2
2 0 0 1 1 1066.421 3129.703 17.8 765.1777 3119.17 17
2 0 0 1 1 1061.954 3241.521 17.8 832.2777 3297.036 17.1
2 0 0 1 1 1153.374 3219.806 17.8 844.8282 3258.476 17
2 0 0 1 1 996.0888 3153.16 18 779.7479 3065.04 17.5
3 0 1 0 1 837.8166 4170.235 17.6 796.6846 4185.589 17
3 0 1 0 1 857.5266 3676.587 18 761.7916 3436.051 17.4
3 0 1 0 1 877.1539 3925.341 18.2 826.4325 3707.399 17.1
3 0 1 0 1 925.9606 3738.633 18.2 821.8661 3497.497 17
3 0 1 0 1 864.1742 3812.501 18 837.4485 3575.493 17
4 0 1 1 0 941.1838 3611.456 18.2 950.5054 3406.734 17
4 0 1 1 0 1083.639 3505.461 18.4 967.2774 3293.455 17.4
4 0 1 1 0 823.1568 3484.1 18.4 817.4084 3282.752 17.4
4 0 1 1 0 891.7884 3674.605 18.4 872.6505 3486.605 17
4 0 1 1 0 900.3536 3555.815 18.6 889.4422 3352.228 17.3
5 1 0 0 0 920.894 4117.829 18.5 896.7897 3494.802 17.8
5 1 0 0 0 969.0664 4202.894 17.8 955.8543 3559.309 17.8
5 1 0 0 0 957.0744 4094.928 18.1 954.3626 3500.203 17.9
5 1 0 0 0 955.1939 4328.478 17.8 864.5807 3822.866 17.5
5 1 0 0 0 921.7164 4129.023 18 935.919 3510.77 17.6
6 1 0 1 1 1105.478 3533.047 18.5 919.7457 3435.989 18
6 1 0 1 1 1095.382 3497.51 18.4 901.8484 3400.193 17.8
6 1 0 1 1 1009.676 3422.328 18.6 822.6954 3304.974 18
6 1 0 1 1 1078.69 3497.757 18.5 941.6131 3384.786 18
6 1 0 1 1 1041.347 3560.036 18.4 855.6604 3419.705 17.9
7 1 1 0 1 1217.367 3517.079 18.7 978.0519 3424.187 18
7 1 1 0 1 1019.773 3539.172 18.7 838.3406 3413.35 17.8
7 1 1 0 1 1031.06 3604.227 18.2 854.1598 3485.926 18
7 1 1 0 1 1013.366 3489.633 18.8 853.13 3364.788 18
7 1 1 0 1 999.5527 3574.522 18.8 810.2731 3451.016 18
8 1 1 1 0 1056.188 3921.99 17.5 876.6365 3479.043 17
8 1 1 1 0 999.4628 4088.24 17.6 804.7451 3612.005 17.5
8 1 1 1 0 997.7841 3946.874 17.5 807.5579 3446.995 17.8
8 1 1 1 0 1072.561 3939.494 17.5 887.2043 3501.874 17.2
8 1 1 1 0 985.5829 4025.463 17.5 783.3092 3554.647 17
Average 18.1125 Average 17.4675
Std Dev. 0.412116 Std Dev. 0.385897
Max 0.6875 Max 0.5325
Min -0.6125 Min -0.4675
183
Appendices
C-1 REVERSE TAPER
This design is similar to the design implemented in this research, except

that the valve stem has a reverse taper which would improve flow control. For
instance, if the cavity pressure is too high, closing the valve chokes the flow and
creates a void in the feed system to quickly reduce the cavity pressure. This
design change would make greatly simplify control throughout the filling and
packing stages.
Unfortunately, the reserse taper makes maintenance more difficult as the

entire manifold would need to be disassembled to access the valve stem.
Additionally, each drop would still require its own hydraulic circuit and related
sub-systems so system cost would not be significantly reduced.
184
Appendices
C-2 DIGITAL MODULATION
This design concept would utilize a fast acting solenoid to toggle the valve
on or off at rates on the order of 50 Hz, approximately five times the natural
frequency of the melt dynamics. The drop volume, approximately one cubic inch,
would serve as a resonance chamber to equilibrate the polymer melt before
entering the cavity. The pressure transducer located in the drop provides
process feedback without requiring core or cavity modification.
To utilize a solenoid or similar actuation device, the maximum melt force

on the valve pin must be on the order of one hundred pounds. This design fulfills
that requirement while avoiding excessive pressure drops and shearing. As
shown, a one-piece insert could be fitted through the top of the manifold bar for
ease of manufacture, assembly, and maintenance. If effective, this digital design
represents a simple, cost effective solution.
185
Appendices
C-3 ROTARY ACTUATION
This design concept would utilize a rotational actuator with a rotary valve
to meter flow to the cavity. This concept has some very nice properties: low
actuation forces due to balanced flow, fast positioning dynamics, negligible
displaced flow with valve actuation, and small valve packaging. A primary
concern, however, is the actuators potential to meter flow. In a full open position,
there is negligible pressure loss. However, the valve would need to be in a
nearly closed position before choking flow significantly.
Section
186
Appendices
D-1 MONTE CARLO SIMULATION
Macro GoForIt3 performs Monte Carlo iterations for the 3 gate Dynamic
Feed example until the counter cell of worksheet RUNROBUST.XLS is 0. Macro
OptPres3 updates ROBUST.XLS with the stochastic material properties,
optimizes the cavity pressure distribution, and copies the predicted part
properties back to RUNROBUST.XLS. Not shown are formulae for calculating
robustness and production yield from the predicted part properties (quite
straightforward).
Sub GoForIt3()
While ActiveCell.Value > 0
Application.Run Macro:="RUNROBST.XLS!OptPres3"
Wend
End Sub
Sub OptPres3()
ActiveCell.Offset(0, 1).Range("A1:C1").Select
Selection.Copy
Windows("ROBUST.XLS").Activate
ActiveCell.Offset(-7, 0).Range("A1").Select
Selection.PasteSpecial Paste:=xlValues, Operation:=xlNone, _
SkipBlanks:=False, Transpose:=True
Windows("RUNROBST.XLS").Activate
ActiveCell.Offset(0, 3).Range("A1:C1").Select
Application.CutCopyMode = False
Selection.Copy
ActiveCell.Offset(0, 1).Range("A1").Select
ActiveCell.Offset(7, -1).Range("A1").Select
Application.ExecuteExcel4Macro String:= _
"[SOLVER.XLA]SOLVER!SOLVER.OK(!R9C2,1,0,(!R12C2:R14C2))"
Application.ExecuteExcel4Macro String:= _
"[SOLVER.XLA]SOLVER!SOLVER.SOLVE(TRUE)"
ActiveCell.Offset(3, 0).Range("A1:A3").Select
Selection.Copy
ActiveCell.Offset(7, 1).Range("A1:A3").Select
Application.CutCopyMode = False
Selection.Copy
ActiveCell.Offset(-10, -1).Range("A1").Select
Selection.Copy
SkipBlanks:=False, Transpose:=False
ActiveCell.Offset(1, -13).Range("A1").Select
Selection.Copy
End Sub
187
Appendices
D-2 RUNROBUST.XLS
It should be noted that Smax and dS/dP are calculated assuming a

Gaussian distribution before the Monte Carlo simulation. This data table is then
used in all the design evaluations with the Monte Carlo simulation running down
the table to obtain part property predictions. This table shows the simulation has
performed the first 28 iterations.
Max Shrinkage dS/dP Pressures Dimensions Lot

Count @1 @2 @3 @1 @2 @3 @1 @2 @3 L1 L2 L3 Yield
1 .00832 .00840 .00870 .00009 .00010 .00011 35.2 32.4 33.9 6.000 10.000 6.000 0.992
2 .00793 .00975 .00867 .00011 .00010 .00011 27.9 42.9 35.1 6.001 9.999 6.000 0.991
3 .00980 .00776 .00765 .00009 .00012 .00010 47.7 32.7 24.7 5.998 10.004 5.999 0.972
4 .00854 .01000 .00805 .00009 .00009 .00010 40.1 48.0 33.0 6.001 9.997 6.002 0.984
5 .00949 .00705 .00844 .00009 .00009 .00010 45.6 30.6 31.8 5.998 10.003 5.999 0.984
6 .00819 .00862 .00684 .00010 .00009 .00012 31.6 32.9 17.9 6.001 9.997 6.001 0.985
7 .00880 .00979 .00800 .00010 .00008 .00010 40.8 48.3 33.3 6.001 9.997 6.001 0.984
8 .00870 .00728 .00877 .00010 .00009 .00010 37.8 24.4 37.0 6.000 10.000 6.000 0.992
9 .00754 .00670 .00677 .00011 .00010 .00009 23.3 16.4 20.7 6.000 10.000 6.000 0.992
10 .00775 .00770 .00827 .00009 .00009 .00009 31.8 29.4 35.5 6.000 10.000 6.000 0.992
11 .00791 .00856 .00703 .00010 .00008 .00011 29.8 36.3 21.3 6.001 9.997 6.001 0.986
12 .00850 .00835 .00748 .00009 .00010 .00009 40.1 32.6 28.2 6.000 10.000 6.000 0.992
13 .00742 .00787 .00600 .00010 .00011 .00009 24.8 25.7 11.5 6.000 10.000 6.000 0.992
14 .00760 .00915 .00974 .00010 .00010 .00009 26.9 41.7 51.1 6.000 10.000 6.000 0.992
15 .00904 .00702 .00921 .00010 .00010 .00010 40.3 25.3 40.3 5.999 10.001 6.000 0.990
16 .00724 .00820 .00786 .00010 .00010 .00010 22.6 31.5 29.9 6.000 10.000 6.000 0.992
17 .00880 .00954 .00820 .00011 .00011 .00010 35.6 40.4 31.3 6.000 10.000 6.000 0.992
18 .00810 .00754 .00846 .00011 .00009 .00010 29.1 27.5 35.4 6.000 10.000 6.000 0.992
19 .00757 .00809 .00782 .00012 .00011 .00009 21.4 28.7 30.5 6.000 10.000 6.000 0.992
20 .00836 .00822 .00841 .00009 .00009 .00010 36.3 37.4 33.8 6.000 10.000 6.000 0.992
21 .00848 .00923 .00770 .00010 .00009 .00011 34.2 41.0 26.0 6.000 9.999 6.001 0.990
22 .00698 .00770 .00701 .00010 .00011 .00010 19.9 25.1 19.3 6.000 10.000 6.000 0.992
23 .00633 .00874 .00649 .00010 .00009 .00010 15.9 30.9 16.8 6.002 9.997 6.001 0.979
24 .00790 .00831 .00632 .00009 .00010 .00010 31.5 29.2 14.2 6.000 9.999 6.001 0.991
25 .00978 .00749 .00960 .00010 .00010 .00009 45.9 31.7 46.7 5.999 10.003 5.998 0.983
26 .00859 .00676 .00838 .00010 .00011 .00010 35.1 20.1 34.7 5.999 10.001 6.000 0.990
27 .00649 .00885 .00784 .00011 .00009 .00009 16.3 23.8 38.8 6.002 9.993 6.004 0.912
28 .00806 .00833 .00754 .00011 .00011 .00009 26.7 31.3 27.8 6.000 10.000 6.000 0.992
29 .00600 .00973 .00637 .00010 .00011 .00011
30 .00763 .00742 .00795 .00010 .00012 .00011
31 .00812 .00972 .00780 .00009 .00010 .00010
188
Appendices
D-3 ROBUST.XLS
This is a worksheet – at the begining of eachiteration the shrinkage model

is copied, then the optimal pressure distributions recalculated. Note that this
worksheet matches that of the optimized 28th iteration of RUNROBUST.XLS
3 Gate: Dynamic Feed
Max. S dS/dP
Shrinkage 1 0.0085 0.00010
Shrinkage 2 0.0088 0.00009
Shrinkakge 3 0.0060 0.00010
Requirement Listing L Ldes allowable sigma

Dimension 1 6.00 6.030 0.0100 0.0033
Dimension 2 10.00 10.050 0.0150 0.0050
Dimension 3 6.00 6.030 0.0100 0.0033
Shrinkage Estimate 0.0050
Goal 0.991
Distances
Pressures x1 x2 x3
P1 31.50 n/a n/a n/a
P2 29.20 5.00 n/a n/a
P3 14.20 10.00 5.00 n/a
Objectives
Shrinkage Length Cost
Dimension 1 0.0055 6.0000 0.0027
Dimension 2 0.0056 9.9990 0.0032
Dimension 3 0.0045 6.0010 0.0040
0.991
Process Constraints Actual Allowed

Minimum Pressure 14.20 0.00
Maximum Pressure 31.50 150.00
Max. Pressure Slope 2.00 2.00
189
Appendices
190

Dynamic Feed Control

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Dynamic Feed Control

Загружено:

Авторское право:

Доступные форматы

DYNAMIC FEED CONTROL: A NEW METHOD FOR

INJECTION MOLDING OF HIGH QUALITY PLASTIC PARTS

David Owen Kazmer

Approved for the University Committee on Graduate Studies:

The thrust of thermoplastic materials into advanced technical applications

The invention utilizes multiple valves in the feed system of a mold to

Finally, a methodology for the design of molded parts was introduced

The work was performed with the cooperation of several industry

Finally, I’d like to acknowledge the support of the Stanford Integrated

I’m so glad you happened to me.

LIST OF TABLES .....................................................................................................................................xiv

LIST OF FIGURES ....................................................................................................................................xv

CHAPTER 1: INTRODUCTION ...............................................................................................................2

CHAPTER 3: PROCESS DYNAMICS.....................................................................................................39

CHAPTER 4: EXPERIMENTAL VALIDATION OF CONTROL STRATEGY ................................84

CHAPTER 5: EXPERIMENTAL VALIDATION OF PROCESS CAPABILITY.............................107

CHAPTER 6: DESIGN METHODOLOGY...........................................................................................137

TABLE 1-1: PROCESS TO MODIFY MOLD............................................................................................ 21

FIGURE 1-1: CLASSIC DESIGN TO PRODUCTION PROCESS ................................................................... 6

Ppack Pack pressure level

INJECTION MOLDING OF HIGH QUALITY PLASTIC PARTS

David Kazmer, Ph.D.

Adviser: Philip Barkan

The thrust of thermoplastic materials into advanced technical applications has

Approved for publication:

CHAPTER 1: INTRODUCTION ................................................................................................................2

Injection molding of thermoplastics has emerged as the premier vehicle for

THE PLASTICS INDUSTRY

Fundamental research in the 1930’s led to the development of nylon and

Even though plastics continued to be used in technical applications after

By 1985, the volume of plastics produced in the United States exceeded

The injection molding process, described later in Chapter 3, is the most

As applications become increasingly technical, competitive pressures are

Manufacturers of plastic parts are constantly striving to reduce unit cost.

begun to displace engineering thermoplastics in less demanding technical

With a greater number of product manufacturers competing for a given

Prototype Mold Design

Figure Chapter 1: -1: Classic design to production process

Advances in the Design Process

Continued competitive pressures during a recessionary period have driven

In the short life cycle typical of computer and communication markets,

Product development times have been reduced significantly in the last

The utilization of computer technologies for Computer Aided Design (CAD)

Concurrently, process simulation has become an integral tool in the

Williams and Lord advanced the simulation of the injection molding

By 1980, commercial simulations had been introduced by Moldflow Ptd20

engineers became accustomed to performing multiple analyses to avoid costly

However, computer simulations will always be, at best, an approximation

Advances in the Injection Molding Process

After the modern reciprocating screw machine was developed in 1972,

incremental changes to the molding machine’s control systems, some

Prior to 1972, injection molding machines controlled the injection pressure

The first modern computer-controlled injection molding machine was first

EXTERNAL INPUT VARIABLES

Machine Wear, Hydraulic Leakage,

Back Pressure Melt Volume INJECTION Melt Pressure Cav. Pressure

Mold Temp. Part Appearance

Figure Chapter 1: -2: Ma’s Description of the Injection Molding Process

Control systems using servo-valves to obtain closed loop control of ram

By monitoring cavity pressure as an input signal, closed loop cavity

More sophisticated model-based control systems have yielded only

Many model-based controllers have been developed. Shankar was first to

Another potential benefit of model-based control is the resolution of