You are on page 1of 83

The Pennsylvania State University The Graduate School College of Engineering

INVESTIGATION OF A MASTER AXIS OF ROTATION FOR USE IN SPINDLE METROLOGY

A Thesis in Mechanical Engineering by Robert D. Grejda

Submitted in Partial Fulfillment of the Requirements for the Degree of

Master of Science

May 1999

I grant The Pennsylvania State University the nonexclusive right to use this work for the Universitys own purposes and make single copies of the work available to the public on a not-for-profit basis if copies are not otherwise available.

______________________ Robert D. Grejda

We approve the thesis of Robert D. Grejda.

Date of Signature

________________________________ Eric R. Marsh Assistant Professor of Mechanical Engineering Thesis Advisor

______________

________________________________ Marc Carpino Associate Professor of Mechanical Engineering

______________

________________________________ Richard C. Benson Professor of Mechanical Engineering Head of the Department of Mechanical Engineering

______________

iii

ABSTRACT

The use of an alternative method for spindle and machine tool metrology is investigated in detail. A master axis of rotation is used to characterize the error motions of a machine tool spindle instead of traditional techniques that use a master sphere or flat as an artifact. The master axis method allows for comprehensive analyses to be taken of a spindle, such as measurement of structural loop stiffness at speed. The effect of spindle loading, both static and dynamic, while simultaneously measuring error motion at speed is also feasible using the master axis. In this investigation, the master axis method is used to characterize the spindle of a 3-axis vertical milling machine. Results show the spindle error motions are affected by the speed and applied load; a characteristic that may not be measured practically by traditional spindle measurement methods. Practical measurement of dynamic stiffness of a spindle at speed is also examined with the use of the master axis. These testing conditions allow for more accurate analysis of machine tool spindles using methods that are representative of practical applications.

iv

TABLE OF CONTENTS

List of Figures ..................................................................................................................... v List of Tables....................................................................................................................viii Acknowledgments.............................................................................................................. ix Chapter 1: Overview ........................................................................................................... 1 1.1: Introduction ...................................................................................................... 1 1.2: Research Objectives ......................................................................................... 6 1.3: Literature Review............................................................................................. 8 1.4: Proposed Approach ........................................................................................ 22 Chapter 2: Introduction of Master Axis ............................................................................ 23 2.1: Master Axis Design Overview ....................................................................... 23 2.2: Master Axis Validation on a Rotary Table .................................................... 28 2.3: Experimental Setup ........................................................................................ 31 Chapter 3: Experimental Procedure .................................................................................. 34 3.1: Testing Procedures ......................................................................................... 34 3.1.1: Dynamic Compliance Measurement at Speed ................................ 34 3.1.2: Radial Error Motion Under Load and Static Stiffness Analysis ..... 36 3.1.3: Structural Error Motion................................................................... 38 3.1.4: Variation of Drawbar Torque.......................................................... 39 3.2: Post Processing of Results ............................................................................. 42 Chapter 4: Results ............................................................................................................. 47 4.1: Static and Dynamic Stiffness/Compliance at Speed Results ......................... 47 4.2: Radial Error Motion Results at Speed and Under Load................................. 52 4.3: Structural Error Motion Results ..................................................................... 58 4.4: Variation of Drawbar Tension Results........................................................... 61 Chapter 5: Conclusions and Future Work ......................................................................... 63 Appendix: Supporting Figures ......................................................................................... 64 References ......................................................................................................................... 72

LIST OF FIGURES

1.1: Example of a polar plot showing spindle error measurement...................................... 3 1.2: Spindle measurement using master sphere (a) and master axis (b)............................. 6 1.3: Checking full indicator movement (FIM) of a spindle after Schlesinger .................... 9 1.4: Tlustys method for measuring spindle error............................................................. 10 1.5: Schematic for fixed and rotating sensitive direction in turning (a) and rotating sensitive direction in jig boring (b) after B89.3.4M.................................................. 11 1.6: Schematic of jig borer with structural loop................................................................ 13 1.7: Calculation of error motion in the instantaneous sensitive direction after B89.3.4M........................................................................................................... 14 1.8: Illustration of squareness reversal, (a) shows the 180 rotation from (b) resulting in a squareness error of . ............................................................ 15 1.9: Illustration of Donaldson Reversal, (b) showing the 180 rotation from (a) of both indicator and ball with respect to the spindle after Donaldson................ 15 1.10: 3-D representation of spindle error for multiple revolutions after Vanherck andPeters (a) and the traditional 2-D representation (b) ......................... 17 1.11: Three-probe spindle measurement after Gao et al. .................................................. 18 2.1: Portable master axis 3R air bearing ........................................................................... 24 2.2: Vertical spindle implementation of master axis......................................................... 25 2.3: Measuring tilt and face error motion with master axis .............................................. 27 2.4: Schematic of master axis validation on a rotary table ............................................... 28 2.5: Rotary table validation results ................................................................................... 29 2.6: Comparison of average radial error using master axis and master sphere................. 30 2.7: Bridgeport milling machine used in master axis testing............................................ 31

vi

2.8: Master axis and gage arrangement with torque arm implementation on milling machine...................................................................................................................... 32 2.9: Portable master axis on milling machine with static load application and data acquisition setup.......................................................................................... 33 3.1: Schematic of dynamic stiffness measurement on milling machine ........................... 35 3.2: Schematic for static stiffness measurement and radial error motion on milling machine...................................................................................................................... 37 3.3: Schematic for measurement of structural error motion ............................................. 39 3.4: Data analysis procedure for error motion analysis .................................................... 42 4.1: X-direction stiffness of milling machine under load and at speed............................. 48 4.2: Y-direction stiffness of milling machine under load and at speed............................. 49 4.3: Dynamic compliance comparison for different excitation forces.............................. 50 4.4: Dynamic compliance with 40 N RMS excitation for spindle running and spindle off ........................................................................................................... 51 4.5: Synchronous radial error motion for 2300 RPM, 3-D representation ....................... 53 4.6: Synchronous radial error motion for 3200 RPM, 3-D representation ....................... 54 4.7: Synchronous radial error motion for 2300 RPM, polar plot...................................... 54 4.8: Synchronous radial error motion for 3200 RPM, polar plot...................................... 55 4.9: Asynchronous radial error motion for 2300 RPM, polar plot (32 revs, 107 N load) ................................................................................................. 56 4.10: Asynchronous radial error motion for 3200 RPM, polar plot (32 revs, 107 N load) ................................................................................................. 56 4.11: Structural error motion for 2300 RPM, 3-D representation..................................... 59 4.12: Structural error motion for 2300 RPM, polar plot ................................................... 59 4.13: Structural error motion for 3200 RPM, 3-D representation..................................... 60 4.14: Structural error motion for 3200 RPM, polar plot ................................................... 60

vii

4.15: 3-D synchronous radial error motion for drawbar torque variation (89 N Load, 2300 RPM)........................................................................................... 61 A.1: Synchronous radial error motion for 60 RPM, 3-D and 2-D representation............. 64 A.2: Synchronous radial error motion for 140 RPM, 3-D and 2-D representation........... 65 A.3: Synchronous radial error motion for 220 RPM, 3-D and 2-D representation........... 66 A.4: Synchronous radial error motion for 300 RPM, 3-D and 2-D representation........... 67 A.5: Synchronous radial error motion for 380 RPM, 3-D and 2-D representation........... 68 A.6: Synchronous radial error motion for 500 RPM (LO), 3-D and 2-D representation.. 69 A.7: Synchronous radial error motion for 500 RPM (HI), 3-D and 2-D representation... 70 A.8: Synchronous radial error motion for 1400 RPM, 3-D and 2-D representation......... 71

viii

LIST OF TABLES

2.1: Master axis specifications .......................................................................................... 24 3.1: Parameters for dynamic stiffness tests....................................................................... 35 3.2: Radial error motion test parameters for milling machine .......................................... 37 3.3: Parameters for variation of drawbar torque measurements ....................................... 40

ix

ACKNOWLEDGMENTS

I would like to thank my advisor, Dr. Eric Marsh, for giving me the opportunity to work on this project and for teaching me so much about learning. Without his efforts and guidance, I would not be where I am today. I would also like to thank Mr. Dave Arneson, Mr. Phil Gerber, Mr. Mel Liebers, Mr. Dan Oss, and Mr. Steve Sanner of Professional Instruments, Co.; Mr. Don Martin and Mr. Robert Benjamin of Lion Precision; and Mr. Tim Sheridan of Lorien Consultants for allowing me to tap their seemingly unlimited supply of knowledge and resources, for their donation of hardware for this project, and for developing and building the master axis. Also, I would like to thank Jim Bryan for his insightful suggestions and direction with this project. Success of this project has also been made possible with help of the Mechanical Engineering Department Machine Shop, especially Mr. Dave McCloskey, Mr. Larry Horner, and Mr. Ron Gathagan. Also, the assistance of the entire Mechanical Engineering support staff is greatly appreciated. I would also like to thank my co-worker, Byron Knapp, for his insight and friendship throughout this project, as well as support from Adam Schaut and Peter Jargowsky. Finally, I am very grateful for the support of my parents, my fiance Marie, and my entire family. Without their support, understanding, and care I could not have achieved all that I have to this point.

CHAPTER 1 OVERVIEW

This chapter provides a brief history of spindle metrology. A comprehensive review of the literature is also presented. In addition, the problem statement and research objectives for this project are discussed.

1.1 Introduction During the past few decades, manufacturing processes have trended toward increased precision and accuracy. Weapons development during the cold war required and refined such high precision processes as single point diamond turning for nuclear weapons and guidance systems as well as machining of precision optics with minimal post-processing. For example, it is common for T-6061 aluminum to be diamond turned into parabolic reflectors for telescopes with surface finishes below 25 nanometers Ra while holding form errors less than 100 nanometers with proper machinery and tooling. These types of advances are possible because high accuracy machine tools were developed to permit the production of high accuracy parts. Hand in hand with the development of these higher accuracy machines is the development of better techniques for measurement. Machine tool metrology involves accurate identification of errors caused by vibration, geometrical issues, and thermal effects. Special programs were developed at Lawrence Livermore National Lab (LLNL), the Oak Ridge (Y-12) National Lab, and the National Institute of Standards and

Technology (NIST) that would deal with machine tool error measurement, the improvement of the accuracy of machines, and the distribution of traceable national standards [1]. From these groups, especially LLNL, major advances were made in improvement of machine tool metrology with a particular focus on error sources and separation. With the help of pioneers in the metrology field such as Jim Bryan and Bob Donaldson, even higher accuracy machines such as the Large Optics Diamond Turning Machine (LODTM) were developed. LODTM was made possible because of their work, better measurement techniques, and precision machine design fundamentals like those discussed by Moore; which include proficiency in the areas of geometry, standards of length, dividing the circle, and roundness [2]. These fundamentals must be incorporated into all of a machine tools components including axes ways, lead or ball screws, spindles, motor drives, and machine tool bases. For many machine tools, a large part of its accuracy can be attributed to the spindle. Most machine tools such as lathes, milling machines, and all types of grinding machines, use a spindle or an axis of rotation for positioning workpieces or tools or machining parts. Consequently, the accuracy of the spindles used in their design directly influence the accuracy of the entire machine. Therefore, spindle accuracy can be considered one of the most important components in the overall accuracy of a machine tool. Substantial care must be taken then in the proper analysis of spindle measurements. For appropriate spindle analysis, focus must be maintained on understanding what is actually being measured along with using correct measurement techniques.

For instance, Figure 1.1 shows the result of a spindle measurement in the form of a polar plot. This plot shows the radial deviation from a perfect axis of rotation during operation. Interpretation of this measurement could possibly identify a bearing fault or reveal an imbalance in the drive components. The measurement contains the effects of spindle error, which for the spindle metrologist, is the desired result of the process, convoluted with errors from the entire machine. Therefore, correctly identifying sources of error is a major key in unlocking the accuracy of a machine and a spindle.
90 120 2.2 150 1.1 30 3.3 Microns 60

180

210

330

240 270 Degrees

300

Figure 1.1: Example of a polar plot showing spindle error measurement.

A number of factors can affect error motions of a spindle in a machine tool. Thermal growth of a spindle occurs as a machine warms up to operating temperature. This can cause significant axial errors of a spindle that occur over a period of hours. Variations of preloads on spindle bearings can also cause significant differences in error

measurement. Bearing preload has been shown not to only affect the static and dynamic stiffness of a spindle but also affect the accuracy [3]. Spindle growth on start-up and bearing preload are characteristics relating directly to the spindle, however, additional factors not directly related to the spindle can also have an effect on a spindle measurement. Structural motion from sources such as drive imbalances and coolant pumps can have an adverse effect on spindle accuracy. These sources can cause vibration of the entire machine, which consequently appears in the spindle measurement. It is very important to recognize that these types of issues exist in spindle measurement, so as to avoid misdiagnosis the actual cause of error motions. Proper measurement techniques can also be critical in spindle analysis. Different measurement techniques are available for measuring spindles depending upon the application of the machine. For instance, the spindle on a lathe, which has a fixed sensitive direction, has to be measured differently than the spindle on a jig borer, which has a rotating sensitive direction. The correct measurement technique needs to be applied corresponding to the way that the machine is to be used. Moreover, as spindles become more accurate, error separation of spindle errors from the master used in spindle testing needs to be considered. Traditionally, spindle measurements are performed by mounting a lapped steel ball onto the spindle under test and measuring the error motions with capacitance sensors. The roundness error of the lapped ball, which is generally on the order of 25 nm, is considered to be negligible in comparison to the error motion of the spindle. However, the best modern air bearing spindles have error motions that can be an order of magnitude less than the roundness of the lapped masters. Under these conditions, the error of the master must be accounted for

with reversal or multi-probe techniques. These techniques separate the error of the master from the error of the spindle under test. Hence, spindle measurement must be performed with proper technique and understanding. In most cases, the measurement will be a combination of thermal effects, structural error motion, lost motion from joints and mounts, dynamic effects, master artifact error, and the desired result, the actual spindle error. Therefore, interpretation of results and understanding of proper machine tool metrology is paramount to proper spindle measurement.

1.2 Research Objectives The use of a master axis as an alternative method for spindle measurement is studied in detail. Traditionally, spindle metrology is performed using a lapped artifact or master usually in the form of a sphere, cylinder, or flat as shown in Figure1.2a. The master is mounted onto the spindle under test and measured using an appropriate non-contact gage arrangement. Figure 1.2b illustrates the use of the master axis for spindle measurement. The green color denotes the gaging surface for each setup. The use of the master axis allows for more comprehensive and accurate analyses to be performed on the spindle under test.

(a)

(b)

Figure 1.2: Spindle measurement using master sphere (a) and master axis (b).

It is the goal of this research project to demonstrate the use of the master axis to characterize a 3-axis vertical milling machine spindle and some of the errors that are a function of its operation. These errors are a function of load as well as speed; a characteristic that cannot be analyzed using traditional measurement methods. This information can be used to better characterize the performance of a machine tool spindle.

1.3 Literature Review Much work has been done in the past four decades regarding spindle metrology. Increased precision requirements of the manufacturing and defense industries have advanced spindle measurement; tighter tolerance on form and surface finish of parts require better performance of machine tool spindles. This has led to advances in spindle measurement. Schlesinger shows the first attempts at specifying measurement techniques of error motions and the accuracy of an axis of rotation [4]. The measuring equipment consists of a dial indicator to check the total indicator reading (TIR) or full indicator movement (FIM) of spindle components such as the outside diameter of the spindle rotor, the spindle taper, or a center held in the spindle. By mounting the indicator at a fixed position and measuring a rotating diameter on the spindle, Schlesinger claims to quantify the accuracy of the spindle. Schlesinger also notes that the measurements should be made at the normal operating temperature to accurately represent the spindle measurements of the machine during operation. This concept of measuring the FIM of spindle components as an absolute measure of spindle error is technically incorrect. The FIM is not a true measurement of the axis of rotation, rather a measure of roundness or alignment with respect to the axis of rotation. This can be visualized by measuring the error motion of a bearing that has a gaging cylinder mounted eccentric to the rotor of the spindle as shown in Figure 1.3. A dial indicator arranged to measure the error motion results in a FIM of twice the eccentricity, even if the axis of rotation is perfect. Hence, more appropriate spindle measurement techniques were developed to better quantify the error motion of an axis of rotation.

INDICATOR

eccentricity STATOR

ROTOR

Figure 1.3: Checking full indicator movement (FIM) of a spindle after Schlesinger [4].

Tlusty followed Schlesingers work twenty years later with spindle measurements that used a lapped spherical master and an oscilloscope to aid in real-time measurement and interpretation [5]. Figure 1.4 shows Tlustys typical spindle measurement setup with the output of two measurement gages displayed on an oscilloscope. Non-contact gages point toward a master sphere that is mounted slightly eccentric to the spindle under test. The eccentricity causes a circular shaped polar plot on the oscilloscope. If the spindle under test were a perfect axis of rotation, the oscilloscope would then plot a perfect circle of radius equal to the eccentricity. Hence, deviations from the perfect axis should appear as deviations from a circle on the oscilloscope.

10

MASTER SPHERE ON WOBBLE PLATE

OSCILLOSCOPE

HORIZONTAL GAGE

VERTICAL GAGE

Figure 1.4: Tlustys method for measuring spindle error.

Tlusty uses this setup to measure turning and cylindrical grinding operations, both of which are processes understood to have a fixed sensitive direction. The sensitive direction of a machine tool is defined by ANSI Standard B89.3.4M; Axes of Rotation as the direction perpendicular to the ideal generated workpiece through the instantaneous point of gaging. [6]. Figure 1.5a is a schematic indicating the sensitive and nonsensitive directions for a turning operation, which has a fixed sensitive direction. Figure 1.5b is the schematic for a jig boring operation in which the sensitive direction rotates with the tool while the workpiece is fixed. Hence, the jig boring operation is considered to have a rotating sensitive direction.

11

R TOOL TOOL WORKPIECE WORKPIECE (a)


NON SENSITIVE DIRECTION SENSITIVE DIRECTION

(b)

Figure 1.5: Schematic for fixed sensitive direction in turning (a) and for rotating sensitive direction in jig boring (b) after ANSI Standard B89.3.4M [6].

The error motions in the non-sensitive directions have a second-order effect in relation to that of the sensitive directions because the turned or bored surface is circular. When using the oscilloscope with two probes on a lathe, Tlusty measures both the sensitive and non-sensitive direction, plotting one against the other. However, the motions in the non-sensitive direction, for a lathe, have little effect on the form of the part and cannot be compared directly to the motions of the sensitive direction. Therefore, as Bryan et al. stated nearly ten years later, Tlustys experiments do not accurately represent the error that occurs in the sensitive direction of the machine [7]. This indicates that the oscilloscope information does not correlate to the error motions that would be machined into the workpiece as a result of the error motion of the spindle. Instead a single probe can be used in place of the tool in Figure 1.5a to correctly record the error motion. This information can be superimposed upon an artificially composed base circle to represent

12

the spindle error. Tlustys methods for spindle measurement, although slightly incorrect for the fixed sensitive direction, yield useful information for rotating sensitive direction measurements. Bryan et al. [7] and Donaldson [8], both researchers at Lawrence Livermore National Laboratory, are credited with major contributions the advancement of spindle measurement. Bryan provides a framework for completeness in measurement and accurate terminology while testing spindles. Also, Bryan states the accuracy of a machine is a function of the tool, the vibration characteristics of the machine, the spindle error, and accuracy of the slideways and the overall geometry [7]. Therefore, while testing spindles, care must be taken to recognize the broader issues and their contribution to the overall accuracy of the machine. Structural error motion is one of these oftenoverlooked issues that may complicate a spindle measurement. Structural error motion is unwanted vibration of the machine due to internal or external excitation of the machine tool. Figure 1.6 shows the structural loop for a jig borer. This loop begins at the tool point and includes the tool, spindle and quill, the column, the machine base, lead or ball screws and drives, the slideways on each axis, and the workpiece. The stiffness of this structural loop and hence the relative motion during operation between the workpiece and tool will affect the quality of the part the machine can produce. Unfortunately, this may often be misinterpreted as the result of a poor spindle when the actual cause of error could be from poor fixturing of parts or the influence of a drive motor.

13

STRUCTURAL LOOP

X Y

Figure 1.6: Schematic of jig borer with structural loop.

Bryan et al. shows that Tlustys method of spindle measurement can be used successfully for machine tool spindles that are used in rotating sensitive direction applications [7]. The radial error motion is simultaneously collected in two orthogonal directions. Using an artificial base circle generator, the projection of both the orthogonal directions onto the instantaneous sensitive direction can be added to produce a resultant error. Figure 1.7 illustrates how the measurements from the orthogonal gages (X() and Y()) are projected into the instantaneous sensitive direction, ().

14

SENSITIVE DIRECTION

Y ()

()

() = X() cos + Y() sin

X()

Figure 1.7: Calculation of error motion in the instantaneous sensitive direction after ANSI Standard B89.3.4M Axes of Rotation [6].

This type of error measurement is an attempt to recreate the pure but impractical form of determining the rotating sensitive measurement, in which the displacement probe is mounted in the spindle and senses the spindle error in relation to the inner diameter of a master cylinder. As methods for measuring spindles became more refined and accurate, error separation becomes a concern because neither the master nor the spindle is perfect. Donaldson is credited with a practical, yet elegant, method to separate the roundness error of master artifacts from synchronous spindle error motions [8]. The Donaldson method of reversal in spindle analysis is analogous to the method used to separate the error of a square or straightedge. Figure 1.8 illustrates the realization of the squareness reversal. The apparent perpendicular for the square is drawn while in position (b) as

15

shown. The square is turned over to position (a) and the perpendicular is redrawn, revealing the error of the square.

a
rotation from (b)

Figure 1.8: Illustration of squareness reversal, (a) shows the 180 resulting in a squareness error of .

In spindle measurement, an indicator measures the combined radial error of the spindle and the out-of-roundness of the ball, as shown in Figure 1.9. The indicator and the master ball are rotated 180 measured again.
MASTER BALL SPINDLE MASTER BALL

with respect to the spindle and the total error is

INDICATOR

INDICATOR

(a)

(b)

Figure 1.9: Illustration of Donaldson Reversal, (b) shows the 180 rotation from (a) of both indicator and ball with respect to the spindle after Donaldson [5]

16

Equation 1.1 and 1.2 describe the measurements; where X is the synchronous spindle error, e is the master error, is the angular position of the spindle, M1 is the primary measurement, and M2 is the reversed measurement.

M 1 ( ) = X ( ) + e( ) M 2 ( ) = X ( ) + e( )

(1.1) (1.2)

The spindle error, X, has the opposite sign in the M2 measurement due to the rotation of both the artifact and the indicator. Hence, Equations 1.1 and 1.2 can be solved for e and X, using addition and subtraction of equations yielding Equation 1.3 and 1.4.
e( ) = M 1 ( ) + M 2 ( ) 2 M 1 ( ) M 2 ( ) 2 (1.3)

X ( ) =

(1.4)

Since the error of the ball is then known as a function of , its error map can be recorded and used in subsequent spindle measurements. This supercedes the need for future reversals for the same master except in the instance of recalibration. Another form of error separation has been proposed for additional spindle errors. Evans, Hocken, and Estler [9] describe a method for measuring and separating face error motion of a spindle from the axial error motion. Face error is defined as the z-axis error motion of a spindle measured at some radius from the axis of rotation while axial error is measured at a radius of zero or coincident with the axis of rotation [6]. Modern implementation of the reversal process for precision spindles uses digital signal analysis methods and data sampling that occurs as function of rotation angle rather

17

than of time because of a non-constant spindle speed during testing. Vanherck and Peters first used digital methods for acquiring and storing data and representing the results of spindle measurements [10]. A computer allows for FFT methods to be used in analyzing spindle error data, which allows for compensation of master artifact eccentricity and frequency analysis of data. Vanherck and Peters also implement storage of the master artifact profile using a computer and present a 3-D representation of spindle error motion over a number of revolutions. This allows for trends in the error motion to be viewed as a function of the spindle revolution in comparison to the traditional 2-D plot as shown in Figure 1.10. Additional researchers such as Kakino et al. [11], Yao-Sun [12], and Noguchi et al. [13] use the techniques examined by Vanherck and Peters for measuring error motion using a classical type artifact for rotating sensitive tests.

(a)

(b)

Figure 1.10: 3-D representation of spindle error for multiple revolutions after Vanherck and Peters [10] (a) and traditional 2-D representation (b).

Many authors following Donaldson have elaborated on variations of his methods, such as multi-probe and multi-position techniques, which incorporate error separation for higher accuracy. Whitehouse [14] reviews in detail both multi-position (or multiorientation) methods in comparison to multi-probe methods and the advantages and

18

limitations of each. Whitehouse states that multi-probe measurement can be used for error separation when asynchronous (or variable) errors occur from one revolution of the spindle to the next. One of the better known of these multi-probe techniques is the three-probe method shown in Figure 1.11. Researchers such as Shino et al. [15], Zhang and Wang [16], Gao et al. [17], and Chapman [18] have performed spindle analyses using three probe techniques for precision spindles. In these methods, the total radial error motion of the master artifact and spindle is simultaneously measured at three angular locations. In Figure 1.11, the displacement probes are located at 0, , and , with respect to the X-axis of the fixed coordinate system.
Y 2 r()

X 1

Master Error

Figure 1.11: Three-probe spindle measurement after Gao et al. [17].

The measured error motions of the probes can be written as Equations 1.5 through 1.7,

19

S1 ( ) = r ( ) + e X ( ) S 2 ( ) = r ( ) + eY ( ) sin( ) + e X ( ) cos( ) S 3 ( ) = r ( ) + eY ( ) sin( ) + e X ( ) cos( )

(1.5) (1.6) (1.7)

where is the angle of rotation of the spindle, and are the angular position of probes 2 and 3, respectively, S1-3 are the measured displacements, r is the workpiece or master error, and eX and eY are the spindle errors in the X and Y direction at instant . These equations can be solved simultaneously for r(), eX(), and eY() for each revolution of the spindle. As previously mentioned the three-probe method works well when asynchronous errors make up a significant portion of the total spindle error. However, the technique does have limitations. The angular position of the probes results in measurements that are insensitive to specific harmonics or undulations per revolution (UPRs). Also in the analysis of precision spindles, the frequency range and linearity of the three probes must be consistent. The multi-position approach uses multiple orientations of the master with respect to the spindle. For example, a measurement process begins with the master in a specific orientation with respect to the spindle at which a measurement is taken. Then, the master is indexed a predetermined amount with respect to the spindle. This index is an integer division of 360 . The measurement is recorded for each orientation and is repeated until the original orientation is achieved. As Whitehouse [14] and Chetwynd and Siddall [19] state this form of error separation is very useful in resolving the systematic or repeatable errors of a spindle and master combination. However, the method is incapable

20

of separating spindle or master artifact errors that are integer multiples of the number of indexes as shown by Linxiang [20]. The methods previously described all have in common the use of a sphere or flat as a master. Also, as the precision of the spindle under test increases, it becomes necessary to apply error separation techniques to increase the accuracy of the analysis. For many calibration and testing purposes, these methods work adequately. However, some spindle testing requires additional information about a spindle. In these instances, the master axis may be used in place of the master artifact for analysis of the spindle error motions. Although it is obvious that spindle error motions are often a function of speed, the errors can often be a function load as well. For spindles used in machine tool applications, the characteristics of a bearing may change dramatically with the application of a load representative of cutting forces. A good example is a hydrodynamic spindle on a cylindrical grinding machine, which needs to be spinning in order to sustain a grinding load. The dynamic characteristics of this type of bearing are also a function of load. Being able to apply the load safely and accurately using artifacts may be complicated. Bryan illustrates a setup for applying a cutting force when spindle testing on a lathe; however, few additional attempts to develop a tool for more general applications have been pursued [7]. Preliminary testing shows that the master axis allows for application of loads that may simulate cutting force [21]. This allows a spindle to be tested in a manner more characteristic of its use, perhaps with a load applied in a radial direction as to simulate a cutting force.

21

Information about the dynamic compliance of the spindle at speed is also useful in the characterization of spindle system. This information may be useful in determining the effects of bearing preload or a schedule for spindle maintenance. Problems arise when trying to excite the rotating spindle rotor with a time varying load. Again, this type of analysis becomes complicated with conventional spindle measurement hardware. The master axis allows a time varying load to be applied to the rotor, allowing for dynamic compliance to be reliably tested. Finally, although it will not be investigated in this research, the master axis method for spindle measurement may prove to decrease the uncertainty of spindle analysis. Because the spindle errors of the master axis are in many cases less than the form error of a master artifact, the master axis could be used with some of the aforementioned error separation techniques to increase the accuracy of spindle analysis. This increase in accuracy would be beneficial in calibration of high precision air bearing spindles.

22

1.4 Proposed Approach

In order to advance the state of the art in spindle measurement, the use of the master axis as a spindle metrology tool is investigated. This method allows for comprehensive tests to be performed on a subject spindle in comparison to traditional spindle measurement techniques. The performance of the master axis in direct comparison to a master sphere on a reference axis is first investigated in order to check the repeatability of the master axis method as an acceptable alternative for spindle measurement. The master axis is then used to characterize the spindle of a 3-axis vertical milling machine by measuring error motions under various conditions. The rotating sensitive radial error motion of the machine is identified using the master axis over a range of spindle speeds and radial loads. Also, the structural error motion of the machine is determined over the same ranges of speed and radial loads to examine its effect on the radial error motion measurements. Static and dynamic compliance of the spindle at speed and under load is also measured. Short observations are also made on the effect of changes of drawbar torque on the spindle error motions. Brief conclusions are made concerning the use of the master axis method for spindle measurement as well as recommendations for future work.

23

CHAPTER 2 INTRODUCTION OF MASTER AXIS

This chapter introduces the master axis and the concepts motivating its use in spindle metrology. Also, the basic experiments concerning the validation of the master axis against a master sphere on a rotary table are discussed. Lastly, the basic setup of spindle measurements using the master axis on a vertical milling machine is previewed.

2.1 Master Axis Design Overview The term master axis refers to an air bearing spindle used as a master artifact for spindle measurement. Figure 2.1 shows a picture of the portable master axis used in this investigation. The air bearing is a Professional Instruments BLOCK-HEAD 3R spindle rated for speeds up to 15,000 RPM with error motions less than 25nm (1in.). The master axis is equipped with an integral 1024-count encoder with index pulse to aid in position measurement of the spindle under test. Table 2.1 shows some of the important characteristics of the master axis spindle, including weight, stiffness, load capacity and error motions. All specifications are for an air pressure of 1.0 N/m2 (150 psig) but the load capacities and stiffness are approximately linear with air pressure.

24

Figure 2.1: Portable master axis 3R.

Table 2.1: Master axis specifications.


RADIAL WORKING LOAD CAPACITY STIFFNESS ERROR MOTION TOTAL WEIGHT 140 N (32 lbs.) 88 N/m (0.5 lbs./in.) < 25 nm (< 1in.) AXIAL 380 N (85 lbs.) 260 N/m (1.5 lbs./in.) < 25 nm (< 1in.) 58 N (13 lbs.) TILT 7.8 N-m (70 lbs.-in.) 0.18 N-m/rad (1.6 lbs./rad) < 0.1rad

The master axis has advantages for use as a datum compared to a master sphere. The master axis has high stiffness and damping in conjunction with error motions less than 25 nm. This error motion is better than the form error of most lapped artifacts. The high stiffness of the bearing also ensures that the master axis rigid body motion reliably duplicates the error motions of the spindle under test. Dynamically, the first resonance of the bearing occurs at approximately 1100 Hz. Even without considering the high damping of the bearing, this means that the magnification of error motions at the stator

25

due to a radial excitation of the rotor is less than 5% for frequencies less than 250 Hz (15,000 RPM). Figure 2.2 is a schematic showing the installation of the master axis on a vertical spindle. The rotor of the master axis is mounted to the spindle under test with appropriate mounting hardware, in this case a flange mount. The stator of the master axis is constrained from rotation with the use of an anti-rotation arm. Essentially, the master axis stator is allowed to move in the all degrees of freedom except for rotation about the z-axis. Displacement probes target the stator of the master axis for making measurements

SPINDLE UNDER TEST

Z AXIS

MASTER AXIS ROTOR

ANTI-ROTATION ARM

DISPLACEMENT SENSORS

MASTER AXIS STATOR

OUTPUT FROM INTEGRATED ENCODER

Figure 2.2: Vertical spindle implementation of master axis.

In Figure 2.2, the measurement probes are targeted radially and axially on the stator. Hence, the measurement probes measure the rigid body motion of the master axis as it

26

undergoes the same error motions as the spindle under test at a specific radial and axial location. Measurement of axial error is expedited with the master axis. For example, because the bottom of the master axis is flat rather than curved like the top of a ball, the axial measurement is less sensitive to misalignment of the axial probe. This is important because analysis of axial error motions do not allow for removal of first order errors which result from misalignment of a displacement sensor to the top of a master ball. Not only can measurements be made with probes oriented radially and axially on the master axis, but sensors can also be positioned to measure face or tilt errors as well. Face or tilt error can be measured directly at a radial or axial location in relation to the spindle under test because the gaging surface is non-rotating. Gaging bars are hardmounted to the master axis stator and allow for measurements to be made in different planes in relation to the spindle. Figure 2.3 shows the measurement of radial error at the location of the bearings of the spindle under test.

27

SPINDLE UNDER TEST

MA
GAGING BARS

DISPLACEMENT SENSORS

Figure 2.3: Measuring tilt and face error motion with the master axis.

All of these tests mentioned above can be performed in addition to standard analyses recommended as outlined by the ANSI Standard B89.3.4M: Axes of Rotation [4].

28

2.2: Master Axis Validation on a Rotary Table A simple experiment is performed to validate the performance of the portable master axis against the result obtained using a master sphere. Both setups measure the radial error motion in a fixed sensitive direction of a rotary table. Figure 2.4 illustrates the basic concept of this experiment. Green designates rotating surfaces while red designates a non-rotating surface. The blue areas represent the gaging area surface. Note that for the master axis, the gaging surface remains stationary.

SPHERE

MASTER AXIS

Figure 2.4: Schematic of master axis validation on a rotary table.

For each test, the master axis or sphere is centered on the axis of rotation of the rotary table to nearly 1-micron. Radial error is recorded in 5 increments for four

complete revolutions of the rotary table using both the master axis and the master sphere. The probe height above the surface of the rotary table is identical for each test. The data is analyzed and plotted using MATLAB, with a program that removes the centering errors and superimposes the result on an arbitrarily sized base circle. Figure 2.5 shows the resulting polar plots of each revolution of the error using each method. It is observed that the error motion does not exactly repeat from revolution to revolution. This results

29

from a combination of the asynchronous error of the table along with thermal drift effects. Figure 2.6 is a comparison of the average error motion using the master axis and the master sphere. The total error of the table between the measurement techniques is repeatable within 5%, or 175nm (7 in.) for this rotary table, which is certainly within the repeatability of the radial error from revolution to revolution and the thermal uncertainty of the experiment. The average radial error measured by each of the techniques also correlates well. This simple experiment shows the master axis can be used as an adequate substitute for a master sphere in a ball bearing spindle measurement.

Master Sphere Result 120 4 Revolutions Average 150

90

6 Microns 60 4 30 2

Master Axis Result 120 4 Revolutions Average 150

90

Microns 60

4 30 2

180

180

210

330

210

330

240 270 Degrees

300

240 270 Degrees

300

(a)

(b)

Figure 2.5: Rotary table validation results: (a) Radial error motion of rotary table using master sphere (b) Radial error motion of rotary table using master axis

30
Master Sphere vs. Master Axis 120 Sphere Average Master Axis Average 150 2 4 30 90 6 Microns 60

180

210

330

240 270 Degrees

300

Figure 2.6: Comparison of average radial error using master axis and master sphere.

31

2.3 Experimental Setup The machine tool used in the remaining experiments is a 3-axis vertical milling machine made by Bridgeport Machines during the late 1960s. This mill is a standard machine tool used for basic milling, boring and drilling operations. The ball bearing spindle of this machine tool will be measured using the master axis to quantify the error motions. Figure 2.7 shows the Bridgeport machine used in the master axis experiments.

Figure 2.7: Bridgeport milling machine used in master axis testing.

A flange adapter allows the master axis to be mounted into the Bridgeport quill using a standard R8 19 mm () collet. The rotational degree of freedom of the master axis stator is constrained using an anti-rotation arm mounted to the quill housing and an elastic band. Lion Precision DMT 10 capacitance gages are aligned in the X and Y direction on the diameter of the master axis for measurement of the error motion and are

32

secured to the table using 2-4-6 blocks and strap clamps. The index pulse output of the encoder of the master axis is recorded simultaneously with the capacitance gage output using a HP Dynamic Signal Analyzer 35670A at sample rates of 25.6 kHz. Efforts are made to properly ground the capacitance gages to the targets, the amplifiers and the data acquisition system to minimize the effect of electrical noise on the data collected. Figures 2.8 and 2.9 show detailed views of the arrangement of the gages and the implementation of the master axis onto the Bridgeport spindle. These setups are used to measure the rotating sensitive direction of the machine.

Figure 2.8: Master axis and gage arrangement with torque arm implementation on milling machine.

33

Figure 2.9: Portable master axis on milling machine showing static load application and data acquisition setup.

Detailed explananations of the four main experiments performed on the milling machine are discussed in Chapter 3.

34

CHAPTER 3 EXPERIMENTAL PROCEDURE

This chapter discusses the four main experiments using the master axis on a vertical milling machine in detail. The details involved with the post processing of the data are also reviewed.

3.1: Testing Procedures Four main experiments using the master axis are considered: measurement of static and dynamic compliance at different spindle speeds, measurement of radial error motions under load, measurement of structural error motions, and the effect of drawbar torque on spindle error motion. The following four sections discuss these experiments in greater detail.

3.1.1: Dynamic Compliance Measurement at Speed Dynamic stiffness or compliance may be tested with the master axis by applying a time varying load to the master axis stator during spindle operation. The dynamic response of the spindle is monitored using a capacitance gage. Figure 3.1 shows the implementation of this setup. Table 3.1 outlines the major parameters for the dynamic stiffness testing. For these tests, a 100 N (25 lbs.) Ling Dynamic Systems shaker is mounted to the milling machine table. A stinger connects the shaker to a 200 N (50-lb.) Kistler 9712A50 load cell mounted to the master axis stator. The load cell measures the time varying

35

force, which for these tests is white noise, applied by the shaker. The RMS force levels are approximately 20 N and 40 N (5 lbs. & 10 lbs.) as measured by the load cell. The displacement response of the master axis is measured using a DMT 10 Lion Precision capacitance gage. The compliance frequency response function of the spindle over a range of speeds is computed using the HP Dynamic Signal Analyzer 35670A.

SPINDLE UNDER TEST

FORCE

DISPLACEMENT SENSOR MA

SHAKER

Figure 3.1: Schematic of dynamic stiffness measurement on milling machine.

Table 3.1: Parameters for dynamic stiffness tests.


Spindle Speeds: LO Range HI Range Load -- White Noise Excitation Frequency Range Load Cell Sensitivity Capacitance Gage Sensitivity 60, 140, 220, 300, 380 and 500 RPM 500, 1400, 2200, and 3200 RPM 20 and 40 Newtons RMS 0 - 400 Hz 22.4 mV/N 40 mV/micron

36

It is important to recognize that energy is being added to the spindle from sources other than the shaker while the machine is in operation. For example, a motor imbalance may add substantial amounts of energy to the spindle at the rotational frequency of the motor, which would corrupt the frequency response function at that frequency. Therefore, care must be taken to ensure that the shaker is the primary source of energy while computing the frequency response function and the dynamic compliance of the spindle. This is accomplished by monitoring the coherence of the resulting frequency response during testing as the spindle speed changes. If the quality of the coherence function decreases at the resonances, the RMS load can be increased until the coherence becomes satisfactory.

3.1.2: Radial Error Motion Measurement Under Load and Static Stiffness Analysis Two gages mounted onto the milling machine table are used to measure the rotating sensitive radial error motion of the vertical milling machine as shown in Figure 3.2. This setup is also used to measure the static stiffness of the spindle at speed. The gages are positioned 10 cm (4 inches) below the nose of the spindle on the master axis diameter. These gages simultaneously monitor the X and Y static deflection across the structural loop as well as the X and Y radial error motion of the spindle. A comprehensive set of tests is performed where both the load and spindle speed is changed. Table 3.2 outlines the parameters for these tests.

37

SPINDLE UNDER TEST

DISPLACEMENT SENSOR MA

LOAD

Figure 3.2: Schematic for static stiffness measurement and radial error motion on milling machine.

Table 3.2: Radial error motion test parameters for milling machine.
Spindle Speeds: LO Range HI Range Spindle Loads Capacitance Sensitivity Capture Length Data Rate 60, 140, 220, 300, 380, and 500 RPM 500, 1400, 2300, and 3200 RPM 0, 18, 36, 54, 72, 90, and 108 N 40 mV/micron ~ 50 Revolutions 1000-3000 Points/Revolution

Weights are placed in the tray to load the spindle in the X-direction, which could be considered a rough approximation of the average load applied to the spindle in a side milling operation. These weights are incremented from 0 to 110 N (0 to 25 lbs) and the deflection is measured with the capacitance gages to obtain a stiffness curve for the

38

spindle at speed. The same procedure is later repeated to measure the Y-direction static stiffness at speed. The error motions of the spindle are also measured and recorded. In total, the various speeds and loads described in Table 3.2, generate 70 experiments in which the radial error motion is measured for a rotating sensitive direction of the operating spindle under a static load in the X direction. It is important to note that this radial error motion is a combination of the spindle error in conjunction with a number of other effects including structural error motion, thermal variations, and other externally induced noise. To better quantify these other types of errors, the structural error motion is measured in a separate experiment.

3.1.3: Structural Error Motion Once the radial error motion of the spindle is measured, the structural error motion of the overall machine is examined. In many cases, the error motion measured during a standard spindle measurement will be largely structural error motion. This may be due mechanical imbalances of drive motors, or compliance in the probing arrangements. This step is often neglected when performing spindle measurements resulting in improper interpretation of the source of error of a machine. Therefore, the structural error motion of the machine is measured using the same test setup as used for the measuring the radial error motion of the spindle. The master axis is mounted in the spindle as before, but the capacitance gages target a non-rotating surface on the quill of the machine. Figure 3.3 illustrates this concept.

39

QUILL

MA

Figure 3.3: Schematic for measurement of structural error motion.

The mounting blocks for the displacement gages are not shown in Figure 3.3 for clarity. Seventy additional tests are performed using the same parameters as described for measuring the radial error motion of the spindle (see Table 3.2).

3.1.4 Variation of Drawbar Torque It is well known that mounting equipment, such as three jaw chucks, may elastically deform parts prior to machining. Even after machining, the part will have some residual error caused from the distortion from the chuck. This effect also occurs in spindles equipped with poorly designed or machined tooling. The elastic deformation caused by the improper mounting of a spindle in a machine may cause the bearing races to become distorted and influence the error motion of the spindle. The drawbar used to hold a collet

40

in place while machining on a vertical mill may also deform the spindle rotor or bearings, affecting the error motion. For a Bridgeport mill, a long threaded rod is used in the hollow section of the spindle to draw collets into the tapered spindle nose, which holds the cutting tool. The proper torque for the tightening of the drawbar is not specified by the users manual. It may be reasoned that the tightening of the drawbar may cause elastic deformations of the spindle shaft and affect the accuracy of the spindle and the machine. The Bridgeport Operators Manual states that the drawbar ...should be tightened with a normal amount of pressure using the wrench furnished with the machine. [22]. No torque specification is given within manual, which makes the normal amount of pressure a subjective input affecting the accuracy of the spindle. Hence, the last series of tests are performed to investigate the relationship of the drawbar torque on the error motion of the spindle. The master axis is used to measure the spindle error as previously described in Section 3.1.2. The test parameters for the drawbar experiments are shown in Table 3.3. Table 3.3: Parameters for variation of drawbar torque measurements.
Drawbar Torque Spindle Speed Spindle Loads Capacitance Sensitivity Capture Length Data Rate 8, 11, 17, 23, 26 Newton-meters 2300 RPM 0, 40, 80 Netwons 40 mV/micron ~ 50 Revolutions 5000 Points/Revolution

The master axis, at first, is held in the collet with the lowest torque (8 Newton-meters) on the drawbar. The error motion is then measured for three different loads. The purpose of

41

the load is to check if error motions become more pronounced or attenuated. The process is repeated for the remaining 14 combinations of torque and load.

42

3.2 Post Processing of Results For analyzing the radial error motion of the spindle, a MATLAB program is written to extract the desired information from the capacitance gages and tachometer pulse from the encoder. A separate MATLAB script is used to plot the results. Figure 3.4 is a diagram showing the general procedure used in analysis.
INPUT
X - DATA Y - DATA
TACH PULSES

PROGRAM
REMOVE D.C. COMPONENTS RECORD START & STOP FOR EACH REVOLUTION CALCULATE & SUBTRACT ONCE AROUND (X & Y) VIA FOURIER ANALYSIS CALCULATE SYNCHRONOUS & ASYNCHRONOUS ERROR MOTION

OUTPUT
PLOT RESULTS

Figure 3.4: Data analysis procedure for error motion analysis.

The data files contain the X and Y error motions as recorded from the capacitance gages and the tach signal from the master axis encoder. First, the dc component of the data channels (X and Y) is removed and recorded. This dc offset, relative to the offset at zero load, corresponds to the static deflection of the spindle in the appropriate direction. The

43

tach signal is then analyzed to find the exact beginning and end of each revolution of the spindle. Fourier analysis is used to calculate the eccentricity of the master axis on the milling machine spindle for both the X and Y data sets. The integration scheme calculates appropriate Fourier coefficients for the eccentricity or once-around. Equations 3.1 and 3.2 are used to find the Fourier coefficients of the eccentricity, a1 and b1, for the X data set,
2nt 2 N xi cos T i t T i =1 2nt 2 N xi sin T i t T i =1

a1 X =

(3.1)

b1 X =

(3.2)

where T is the period of the data sample, N is the total number of points in the sample length, xi is the value of the ith data point of the X data set, n is the number of revolutions in period T, ti is the time corresponding to the ith data point, and t is the time between data samples. The same equations are used to find the Fourier coefficients for the Y data set. These Fourier coefficients are used to reconstruct the once around or fundamental frequency of the gage data, RX or RY, of each data set as shown with Equations 3.3 and 3.4,
2nt 2nt + b1 X sin T T

R X = a1 X cos RY = a1Y cos

(3.3)

2nt 2nt + b1Y sin T T

(3.4)

44

This once-around waveform can be subtracted from the entire data set so that only 2nd order and higher harmonics of the error are included. The analysis to this point results in two sets of data, X and Y, for numerous revolutions of the spindle that do not include any error components at the spindle speed frequency. Thirty-two consecutive of these revolutions are chosen for averaging and plotting. A mathematical routine averages the error motion for each revolution of the X or Y gage. Each revolution consists of a discrete number of data points. Because the spindle speed is not perfectly constant, this number of points per revolution may vary from revolution to revolution. This prevents the direct averaging of each revolution. Consequently, the minimum number of points per revolution is needed for the entire data set, in order to compare each revolution to one another. For example, if the minimum number of points per revolution is 2028 for the both gages, each revolution is sized so that it is 2028 points in length. In some cases, 3-5 data points are lost at the end of a revolution. This allows, however, point 1 of revolution 1 to be averaged with the first points of the remaining 31 revolutions. The averaging is performed for all 2028 data points resulting in a synchronous error motion for the X or Y direction for 32 revolutions of the spindle. This method works well for this application, because of the high sampling rate used to collect data and the spindle speed from one revolution to the next varies only a fraction of a percent. The asynchronous error motion is found after the synchronous component of the error motion is removed. To accomplish this, the synchronous error motion is subtracted from the each revolution of the gage data with the once-around removed. This method results in asynchronous error motion for each revolution of the spindle.

45

A separate program is used to plot the resulting synchronous and asynchronous error motions. Using the X and Y synchronous error motions to calculate the instantaneous sensitive direction as described in Chapter 1 produces the error motion plot for a rotating sensitive direction. Equation 3.5 shows the calculation of r(), the synchronous error motion, r ( ) = ro + x( ) cos + y ( ) sin where ro is the radius of an arbitrarily chosen base circle, x() and y() are the measured error motions and is the instantaneous sensitive direction. For the previous example, consists of values from 0 to 2 at increments of (2/2028) or 0.0031 radians. It should be noted that the base circle size is chosen depending on the peak to peak amplitude error motions. A base circle that is too large will result in a polar plot too close to a perfect circle to resolve important details in the error motion. However, a base circle that is chosen to be too small will distort the results because much of the data will be concentrated at the origin of the polar plot. It has been recommended that the base circle radius be chosen to be on the order of 3 times the peak to peak amplitude of the data [23]. To evaluate the synchronous error motion value, a least squares fit of the data is performed to position the synchronous error motion on a least squares center. The total synchronous radial error is determined to be the maximum value of the synchronous error less the minimum value with respect to the least squares center. The position (a, b) of the least squares center is calculated as recommended by the ANSI Standard B89.3.4M using Equations 3.6 and 3.7,

(3.5)

46

a=

2 xi
i =1

N 2 y i
i =1 N

(3.6)

b=

(3.7)

where a and b correspond to the X and Y location of the least squares circle center, respectively, N is the number of discrete data points, and xi and yi correspond to the value of the ith data point for the X and Y direction, respectively [6]. The asynchronous error motion is presented in the same fashion as for the synchronous error motion, but in the case for asynchronous error motion, more than one complete revolution is shown. The total asynchronous radial error is then the maximum value of the asynchronous error less the minimum value of the asynchronous error. A least squares center is not needed.

47

CHAPTER 4 RESULTS

This chapter presents the test results of the experiments with master axis for spindle and machine measurement. Results concerning the dynamic and static stiffness at speed are presented. Both the structural error motion and spindle error motion are recorded and analyzed using rotating sensitive direction techniques. Finally, drawbar torque is varied on the machine tool spindle and its effect on spindle error is observed.

4.1 Static and Dynamic Machine Stiffness/Compliance Results As stated in Chapter 2, the milling machine is loaded in a single direction (along the Xaxis) while the spindle is operating at a speed. The deflection of the master axis is recorded using both the X and Y direction capacitance gages while the spindle measurement is performed. The deflection appears in the measurement signal as a dc offset in relation to the synchronous and asynchronous components of the raw spindle data. Figure 4.1 shows the stiffness of the machine spindle in the X-direction with the application of a load in the same direction to be 3.95 N/m with the spindle running. Variations in the spindle speed did not affect the results of the static stiffness measurement, so the calculated stiffness is an average for all of the speeds tested at each load.

48
X-Direction Average Stiffness at Speed - Applied X Load 120

100 Average Deflection Linear Fit 80 ) N ( d a o L

60

40

20

10

15 20 Deflection (microns)

25

30

Figure 4.1: X-direction stiffness of milling machine under load and at speed.

The same procedure is repeated for the Y-direction of the machine; a load is applied in the Y-direction and the deflection measured with the Y capacitance gage for various spindle speeds. In the case of the Y-direction, a stiffness of 5.10 N/m is calculated with the spindle running. The stiffness at speed in the Y-direction is shown in Figure 4.2. The machine tool structural loop stiffness is 30% greater in the Y-direction than in the X-direction. This is expected from the C-Frame geometry of this machine. This occurrence of nonsymmetric stiffness may affect the shape of a hole during boring operations.

49
Y-Direction Average Stiffness - Applied Y Load 120

100 Average Deflection Linear Fit 80 ) N ( d a o L

60

40

20

10 15 Deflection (microns)

20

25

Figure 4.2: Y-direction stiffness of milling machine under load and at speed.

Next, a frequency response function (FRF) from the dynamic compliance testing for two different excitation levels is shown in Figure 4.3. First, a 20 N RMS white noise signal excites the rotor at zero speed. The first two peaks occur at modal frequencies of 99 and 118 Hz, which correspond to structural modes of the entire machine. The third mode at 139 Hz corresponds to a mode of the spindle. When the excitation force level is increased to 40 N RMS, a change is observed in the FRF. The peaks corresponding to the structural modes are unaffected, however, the third mode shifts in modal frequency by a few hertz. This suggests a possible non-linear stiffness within the spindle bearing as a result of the point contact of preloaded ball bearings.

50
3 20 N Excitiation 40 N Excitation 2.5

) N / s n o r c i m ( e c n a i l p m o C

1.5

0.5

50

100 150 Frequency (Hz)

200

250

Figure 4.3: Dynamic compliance comparison with different excitation forces.

Finally, the effect of spindle speed on frequency response of the spindle is measured using the master axis. Figure 4.4 shows the FRFs for the milling machine spindle for both on and off conditions. The excitation force for both of the measurements is 40 N RMS. This force level results in adequate coherence even while the spindle is operating. The amplitude of all of three peaks is obviously affected by operation of the spindle. These results suggest that the master axis would work well in testing the dynamic compliance of hydrostatic or hydrodynamic spindles. These types of spindles have characteristics that are highly dependent upon speed and load and can not be practically investigated with traditional spindle measurement methods.

51
3 Spindle On Spindle Off 2.5

) N / s n o r c i m ( e c n a i l p m o C

1.5

0.5

0 0 50 100 150 Frequency (Hz) 200 250

Figure 4.4: Dymanic compliance with 40 N RMS excitation for spindle running and spindle off.

52

4.2 Radial Error Motion Results at Speed and Under Load The following figures, Figure 4.5 and 4.6, are three-dimensional representations of the synchronous radial error motion for 2 different spindle speeds and 4 different loads. Each plot corresponds to one spindle speed while the load is incremented from 0 to 107 Newtons in the X-direction. Thirty-two revolutions of the spindle are used for the average. It is obvious that the shape of the synchronous error motion changes as the speed changes; the error motion is four lobed in Figure 4.5 and three lobed in Figure 4.6. Additionally, the radial load also affects the synchronous error motion shape and magnitude. For instance in Figure 4.5, the well defined four lobed shape of the error motion at zero load becomes rounded as the load increases. In this case, the total synchronous error motion at 107 N of 1.3 microns is nearly 50 % less than the error motion at zero load which is 2.5 microns. At this speed, the error motion is attenuated as the load increases. Figure 4.6 shows that for 3200 RPM the three lobed shape of the synchronous error motion becomes more defined for increased loads. Also, the load has an adverse effect on the error motion magnitude; the synchronous error increases from 1.2 microns to 1.4 microns from the 0 N to 107 N load case. The Figures 4.7 and 4.8 show the same synchronous error as in Figures 4.5 and 4.6 but in the familiar polar plot form with all of the loads tested. Two issues concerning these plots need to be considered during the interpretation process. First, the synchronous radial error motion shown is not motion purely in the radial direction, rather a combination of radial and tilt errors. No attempt is made at this point to separate these errors. Secondly, the error motion does not consist of just spindle error, but a combination of the spindle inaccuracies along with the structural error motion

53

of the machine. The next section discusses the results of the structural error motion tests in greater detail, however, it is important to note at this point that the structural error motion does influence the following results.

Figure 4.5: Synchronous Radial Error Motion for 2300 RPM, 3-D Representation

54

Figure 4.6: Synchronous Radial Error Motion for 3200 RPM, 3-D Representation

Figure 4.7: Synchronous Radial Error Motion for 2300 RPM, Polar Plot

55

Figure 4.8: Synchronous Radial Error Motion for 3200 RPM, Polar Plot

After the removing the synchronous error from the data, the asynchronous radial error motion can also be calculated and plotted. Figures 4.9 and 4.10 show 32 consecutive revolutions of the spindle at a load on 107 N. The asynchronous radial error motion is the difference between the outermost and innermost points on the plot. In these tests, the magnitude of the asynchronous error motion can be nearly 10 times greater than the magnitude of the synchronous error motion, depending on the speed and the load. For Figure 4.9, the magnitude of the asynchronous error is 13 microns, which is 10 times greater than the synchronous error for the same speed and load.

56

Figure 4.9: Asynchronous radial error motion for 2300 PRM, polar plot (32 revs, 107 N load)

Figure 4.10: Asynchronous radial error motion for 3200 RPM, polar plot (32 revs, 107 N load)

57

The trend over all of the test speeds showed that, in general, the error motions increased with increasing spindle speed. As expected, results from intermediate speeds in both the HI and LO range show higher magnitudes of error motion than the higher spindle speeds. This is true for 380 RPM of the LO speed range and 1400 RPM of the high speed range. In the case of 1400 RPM, the synchronous radial error motion values are nearly double the values at 2300 RPM. It could be reasoned that the higher harmonics of this spindle speeds influence the modes of vibration of the machine structure and spindle adversely afftecting the measured error motion of the spindle. For 1400 RPM, the 4th and 5th and 6th harmonics of the spindle speed (93, 117, and 140 Hz) are close to the three resonant frequencies measured during dynamic compliance testing structural mode of the machine of 99, 118 and 139Hz. This result suggests that stuctural error motion could be a large contributor to the increased error motion values at these speeds.

58

4.3 Structural Error Motion Results Figures 4.11 through 4.14 are representative results obtained for the structural error motion tests, which are performed as described in Chapter 3. The magnitudes of these structural error motions are not negligible in comparison to the radial error motion analyzed in the previous section. As suggested, the synchronous structural error motion values range from 25% to 45% of the measured synchronous radial error motion values. Considering that the structural error motion is measured at the end of the quill while the radial error motion is measured 10cm below, it is possible that a significant portion of the measured radial error motion of the machine tool is due to structural error. With this in mind, the structural error motion tests can be repeated to gain more precise information about the machine tool. Using additional hardware, the structural error motion can be measured at the same axial location with respect to the end of the quill as in the radial error motion tests previously described. The setup for this test would be similar to Figure 2.3. Gaging bars can extend down from the quill to the axial location at which data is collected for radial error motion. The values gained from these tests can be subtracted directly from the radial error motion values. This allows for a better understanding of the actual spindle error, without the effect of structural error motion. This type of analysis is recommended for future consideration.

59
Structural Error Motion -- 2300 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure 4.11: Structrual error motion for 2300 RPM, 3-D representation

Figure 4.12: Structrual error motion for 2300 RPM, polar plot

60
Structural Error Motion -- 3200 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure 4.13: Structrual error motion for 3200 RPM, 3-D representation

Figure 4.14: Structrual error motion for 3200 RPM, polar plot

61

4.4 Variation of Drawbar Tension Results The synchronous error motion results for a spindle speed of 2300 RPM without load and multiple drawbar torque is shown in Figures 4.15. As shown by the figures, the error motion shape and magnitude changes appreciably as the drawbar torque is increased more than 300% from its original value of 8 N-m. The magnitude of the error motion is decreases 40% from 1.7 microns at 8 N-m or torque to 1.1 micron at 26 N-m of torque. The shape of the error motion also changes, as shown by Figure 4.15. The four lobes of the synchronous error become less defined as the torque increases. This suggests an improvement in machine performance as the drawbar torque is increased for this machine, even with the large machine errors caused by structural error motion.

Figure 4.15: 3-D synchronous radial error motion for drawbar torque variation (89 N Load, 2300 RPM)

62

A possible explanation for this trend may result from an increased preload of the spindle bearings as the drawbar torque is increased. In any case, the increase of drawbar torque improves the accuracy of the spindle at the spindle speed tested. Future work should include a more detailed investigation error motion as a function of drawbar torque using a more substantial amount of spindle speeds.

63

CHAPTER 5 CONCLUSIONS AND FUTURE WORK

The master axis has shown to be a valuable spindle measurement tool as demonstrated by the results from the 3-axis vertical milling machine spindle tested. The master axis method allows for comprehensive measurements to be taken of a spindle, such as measurement of structural loop stiffness at speed, dynamic stiffness of the spindle at speed, and the effect of spindle loading on error motion at various speeds. Results show the spindle error motion is affected by the speed and applied load; a characteristic that may not be measured practically by traditional spindle measurement methods. The master axis also proves to be useful in measuring the dynamic stiffness of a spindle at speed. Overall, the master axis allows more accurate testing of machine tool spindles in a way that is more representative of the application. Areas of future work should include a rigorous investigation into the possibility of the master axis increasing the accuracy of precision spindle calibration.

64

APPENDIX: SUPPORTING FIGURES


Synchronous Radial Error Motion -- 60 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.1: Synchronous radial error motion for 60 RPM, 3-D and 2-D representation

65
Synchronous Radial Error Motion -- 140 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.2: Synchronous radial error motion for 140 RPM, 3-D and 2-D representation

66
Synchronous Radial Error Motion -- 220 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.3: Synchronous radial error motion for 220 RPM, 3-D and 2-D representation

67
Synchronous Radial Error Motion -- 300 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.4: Synchronous radial error motion for 300 RPM, 3-D and 2-D representation

68
Synchronous Radial Error Motion -- 380 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.5: Synchronous radial error motion for 380 RPM, 3-D and 2-D representation

69
Synchronous Radial Error Motion -- 500 RPM (LO)

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.6: Synchronous radial error motion for 500 RPM (LO), 3-D and 2-D representation

70
Synchronous Radial Error Motion -- 500 RPM (HI)

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.7: Synchronous radial error motion for 500 RPM (HI), 3-D and 2-D representation

71
Synchronous Radial Error Motion -- 1400 RPM

100 80 ) N ( d a o L 60 40 20 0 5 0 0 -5 -5 Microns 5

Figure A.8: Synchronous radial error motion for 1400 RPM, 3-D and 2-D representation

72

REFERENCES

[1] Evans, C. Precision Engineering: an Evolutionary View. Bedford, U.K.: Cranfield Press, 1989. [2] Moore, W.R. Foundations of Mechanical Accuracy. Bridgeport, CT.: The Moore Special Tool Company, 1970. [3] Liebensperger, R. L. So You Think You Know All About Bearing Preload. Machine Design, vol. 44, No. 19, 1972, p.100. [4] Schlesinger, G. Testing Machine Tools: for the use of Machine Tool Makers, Users, Inspectors, and Plant Engineers. 8th Edition. NewYork: Pergamon Press, 1978. [5] Tlusty, J. System and Methods of Testing Machine Tools. Microtechnic, vol. 13, No. 4, 1959, p.162. [6] ANSI/ASME B89.3.4M, Axes of Rotation: Methods for Specifying and Testing. New York: ASME, 1992. [7] Bryan, J.B., Clouser, R.W., and Holland, E. Spindle Accuracy. American Machinist, December 4, 1967, p.149. [8] Donaldson, R.R. A Simple Method for Separating Spindle Error from Test Ball Roundness Error. CIRP Annals, vol.21/1, 1972, p.125. [9] Evans, C.J., Hocken, R.J., Estler, W.T. Self-Calibration: Reversal, Redundancy, Error Separation, and Absolute Testing. CIRP Annals, vol.45/2, 1996. [10] Vanherk, P., Peters, J. Digital Axis of Rotation Measurements. CIRP Annals, vol.22/1, 1973, p.135. [11] Kakino, Y., Yamamoto Y, Ishii, N. New Measuring Method of Rotating Accuracy of Spindle. CIRP Annals, vol.25/1, 1977, p.241. [12] Yao-Sun, S. A New Instrument for Axis of Rotation Metrology. CIRP Annals, vol.34/1, 1985, p.439. [13] Noguchi, S., Tsukada, T., Sakamoto, A. Evaluation Method to Determine Radial Accuracy of High-Precision Rotating Spindle Units. Precision Engineering, vol.17/4, 1995, p.266.

73

[14] Whitehouse, D.J. Some Theoretical Aspects of Error Separation Techniques in Surface Metrology. Journal of Physics E: Scientific Instruments, vol.9, 1976, p.531. [15] Shino, H., Mitsui, K., Tatsue, Y. et al. A New Method for Evaluating Error Motion of Ultra Precision Spindle. CIRP Annals, vol.36/1, 1987, p.381. [16] Zhang, G. X., Wang, R. K. Four-Point Method of Roundness and Spindle Error Measurements. CIRP Annals, vol.42/1, 1993, p.593. [17] Gao, W., Kiyono, S., Nomura, T. A New Multiprobe Method of Roundness Measurements. Precision Engineering, vol.19/1, 1996, p.37. [18] Chapman, P. D. A Capacitance Based Ultra-Precision Spindle Error Analyser. Precision Engineering, vol.7/3, 1985, p.129. [19] Chetwynd, D. G., Siddall, G. J. Improving the Accuracy of Roundness Measurement. Journal of Physics E: Scientific Instruments, vol.9, 1976, p.537. [20] Linxiang, C. Measuring Accuracy of Multistep Method in the Error Separation Technique. Journal of Physics E: Scientific Instruments, vol.22, 1989, p.903. [21] Grejda, R. D., Marsh, E. R., et al. Master Axis of Rotation for Spindle Measurement. Proc. Of ASPE 1998 Annual Meeting, pp. 488. [22] Bridgeport Operators Manual. BridgePort Machines, Bridgeport, Connecticut. 1969. [23] Comments concerning base circle size by Tim Sheridan at the Axes of Rotation B89.3.4M Meeting. St. Louis, MO, October 1998.