Вы находитесь на странице: 1из 11

Int J Adv Manuf Technol (2011) 57:453463

DOI 10.1007/s00170-011-3306-6

ORIGINAL ARTICLE

Design for manufacturing applied to turbomachine


components
Julien Chaves-Jacob & Grard Poulachon &
Emmanuel Duc & Christian Geffroy

Received: 12 October 2010 / Accepted: 28 March 2011 / Published online: 13 April 2011
# Springer-Verlag London Limited 2011

Abstract This paper proposes a new design method for


complex workpieces applied to turbomachine components.
The proposed method is integrated into a simultaneous
engineering process considering design for manufacturing
aspects; it is based on the definition of indicators which
quantify the capability of a workpiece computer-aided
design (CAD) model to check functional requirements.
This paper especially develops the definition of manufacturability indicators for turbomachine components; hydraulic and mechanical indicators are not presented. The
proposed design method is based on the definition of a
parametrical CAD model. Then a large randomized number
of parametrical combinations are tested and the generated
CAD models are classified using indicators. The machinability indicator is obtained using the key issues of current
machining problems.
Keywords Impeller . 5-Axis machining . Design for
manufacturing . Manufacturability indicators . Ruled surface
J. Chaves-Jacob (*) : G. Poulachon
Arts et Metiers ParisTech, LaBoMaP,
Rue porte de Paris,
71250 Cluny, France
e-mail: julien_chaves@hotmail.com
G. Poulachon
e-mail: gerard.poulachon@ensam.eu
E. Duc
Laboratoire de Mcanique et Ingnieries, LaMI,
IFMA UBP, Campus des Czeaux,
BP265, 63175 Aubire cedex, France
e-mail: emmanuel.duc@ifma.fr
C. Geffroy
Division Moteurs Spatiaux, Snecma Groupe SAFRAN,
Foret de Vernon, BP 802, 27208 Vernon cedex, France
e-mail: christian.geffroy@snecma.fr

1 Introduction
This paper concerns the manufacturing of turbomachine
components formed by a set of blades, usually named
impellers. The function of an impeller is to transfer energy
between a flow and a mechanical system. The performance
of this transfer is directly associated to the geometry of the
flow passage between two successive blades. Thus, the
design and the manufacturing of the blades are particularly
important. Figure 1 illustrates the impeller design and
manufacturing process. During the design stage, the
functional needs are converted into geometrical specifications. Then the computer-aided manufacturing (CAM) and
manufacturing process will generate the real geometry.
Thus, the manufactured blade must be as close as possible
to the designed model to reduce the errors between the
computer-aided design (CAD) model and the real workpiece to ensure that the functional requirements are met.
One of the aims of this paper is to develop and increase the
interactions between design and manufacturing to limit and
control the generated errors.
Nowadays, impellers are manufactured using 5-axis
machining, which enables a high degree of accessibility.
During the machining stage, roughing and finishing involve
certain difficulties. The roughing of the flow passage is the
longest stage in the machining process and significant
economies can be made during this step [1]. Roughing
induces two main problems: the poor machinability of the
impeller materials [2] and the optimization of the tool path
[1, 3, 4]. Furthermore, the finishing strategy must produce
the required geometry. Blade finishing may be performed
using point- and flank milling strategies. Point milling
produces scallops between each path [5]. Flank milling is
more productive but it generally produces interferences
between the tool and the blade [6, 7]. These interference

454

Int J Adv Manuf Technol (2011) 57:453463

Fig. 2 Constituent parts of an impeller

at the junction of the blade and the hub. Figure 2 illustrates


also the tip edge of the blade; these edges ensure a dynamic
seal between the impeller and the static part. The leading
edge and the trailing edge have opposite requirements:
Fig. 1 Design and manufacturing process for an impeller

errors may be reduced using three approaches: tool path


optimization [6, 814], tool geometry optimization [7], and
the approximation of the nominal geometry by surfaces
which do not generate interferences in flank milling [15].
Thus, the errors generated during the production process
necessarily induce a geometrical deviation between the
modeled and the manufactured surfaces.
The aim of this paper is to propose a design method to
reduce this geometrical deviation between designed and
machined blades with a controlled production cost. This
design method is based on the definition of numerical
indicators to express the capacity of an impeller CAD
model to verify the chosen requirements. In this paper, only
manufacturability indicators are presented in the case of
impeller workpieces. Nowadays, the lack of numerical
indicators does not enable good communication between
impeller designers, thus limiting an overall optimization.
Impellers are the key parts of pumps and turbines. The
constitutive surfaces of an impeller are illustrated in Fig. 2.
A flow passage is delimited by two consecutive blades;
hence, the importance of the intrados and the extrados
surfaces, which must thus conform to many functional
requirements defined during the design of the impeller.
The blades are thin walls, whose thickness depends on
the stress exerted by the flow and the centrifugal forces.
They are connected to the hub by the linking surfaces called
blade shanks (see Fig. 2), which are used to connect
functional surfaces (the intrados and the extrados) to the
hub and to withstand the great mechanical stresses present

The leading edge must be as thin as possible to limit


the penetration resistance of the blade in the flow;
The trailing edge is thickened to withstand the
mechanical forces generated by the turbulence present
in this area.

All these functional requirements are expressed by the


geometrical definition of the impeller. Figure 3 illustrates as
an example the three morphologies of workpieces which
may be classified in the impeller families; these families
are:

Impellers are used to compress or pump a flow. The


particularity of their intrados and extrados is a smooth
part in the leading and trailing edges and a very twisted
part in the middle of the intrados and extrados surfaces.
Inducers are used in pumps to protect the impeller from
cavitation, but they work in cavitating flows. Frequently, inducers are composed of three or four blades with
twisted intrados and extrados surfaces. Contrary to
impellers, the twist is low and uniformly spread over
the blade, and the height of the blade is greater.
Fans or blisks (fans made from a single part) are
used to increase flow speed or to recover flow
energy. Their intrados and extrados are modeled by
complex free-form surfaces, inducing lot of machining problems.

Figure 3 illustrates an example of a turbo pump used to


compress liquid hydrogen in an aerospace engine: on the
left is the pump part with an inducer and two impellers, and
on the right is the turbine part with two fans; this turbo
pump has the following characteristics:

Int J Adv Manuf Technol (2011) 57:453463

455

do not have the same needs and language, but they must
communicate accurately to reduce the impeller development time and production cost and to increase the quality of
the turbo pump. In view of this, a new method and tools to
improve the design of impellers are presented here.
2.1 Mathematical modeling of blades
Commonly, the intrados and extrados of impellers (especially for inducer and impeller workpieces) are modeled by
ruled surfaces. This approach enables the required surface
to be designed while remaining easy to machine.
A ruled surface, S(u,v), is a parametric surface which
depends on two parameters u and v. Figure 4 illustrates the
notations used. This type of surface is generated by the
displacement of a segment [P0(u) P1(u)] along two directrix
curves C0(u) and C1(u) (see Eqs. 1 and 2).

P0 u S u; 0
1
P1 u S u; 1


C0 u S u; 0
8u 2 0; 1
C1 u S u; 1

Fig. 3 Cross-section of an industrial turbo pump

rotating speed: 36,000 rpm;


power of the turbine: 14 MW;
exit pressure: 185 bars;
temperature: 20 K for hydrogen and 960 K in the
turbine.

Subsequently, this example will be used to illustrate the


methodology proposed in this paper. In the next section, the
three main impeller families are referred to as impellers, for
the sake of simplicity.
The first section of this paper presents the impeller
design method. Next, the proposed new method is
developed. This method is based on the definition of
indicators to quantify the level of conformance to a
functional requirement. This paper develops only the
manufacturability indicators. Impeller production problems
are detailed and then indicators are proposed using the new
design method. These indicators are employed to classify
the manufacturability of five industrial impellers.

Peternell [16] proposed the following equation to define


a ruled surface (Eq. 3).
S u;v 1v  C0 u v  C1 uu; v 2 0; 12

Two characteristic parameters are defined for a rule:


h(u) is the length of the rule [P0(u) P1(u)];
(u)=| (N0(u), N1(u)) | is the twist of the surface along
the rule [P0(u) P1(u)], where (N0(u) and N1(u)) are the
unit vectors normal to the surface, at P0(u) and P1(u),
respectively.

2 Impeller design and manufacturing


The impeller design process is long and complex because
many specialists are required. Furthermore, these specialists

Fig. 4 Definition of a ruled surface

456

A ruled surface is developable if all the values of the


twist are equal to zero along the surface; otherwise, the
ruled surface is twisted.
2.2 Two different approaches of an impeller design
The design of an impeller starts with specifications, the
most important of which is flow efficiency. Next, the
impeller geometry and the workpiece material are defined
to withstand the mechanical stresses exerted during operation. Historically, therefore, impeller design involves flow
performance specialists and mechanical structure resistance
specialists, as presented by Young [17].
The sequential design method starts with the definition
by the specialists of the flow passage variations necessary
to obtain the required performance, generating a 3D
geometrical model of the impeller. This model is used to
carry out mechanical computations and to check if the
proposed geometry is able to endure the functional
mechanical stresses. This step gives rise to geometrical
modifications and the new model is then used to compute
the flow performance. Finally, the geometrical impeller
model is obtained after many loops between flow and
mechanics specialists. When the geometrical model is
defined, it is sent to the manufacturing department. This
system has been criticized by Boothroyd [18], who explained
that it creates barriers between impeller designers, inducing
only a local optimization of the geometry. This method
induces local decisions at each step, and often these choices
are made arbitrarily. For example, the corner radius of the
blade shank is frequently defined by flow specialists,
although it does not have a great influence on flow performances but induces a large number of machining problems.
The sequential design method does not consider machining problems, so it induces lot of machining errors
Fig. 5 Influences of machining
errors

Int J Adv Manuf Technol (2011) 57:453463

during the production of workpieces. Figure 5 illustrates


that machining errors can significantly reduce the real
performances of a model compared to the theory. It is
important to note that the required flow performance is
specified for the real workpiece and not the model.
It was to consider machining errors that design for
manufacturing (DFM) was created. Al-Zubaidy [19] applied these concepts to the design of an impeller and
proposed to involve sequentially the flow, the manufacturing, and the mechanics. This method is still sequential with
many loops, but the machining problem is considered
during the design phase. This approach is a first step
towards simultaneous engineering, which consists of
simultaneously starting all the production stage processes.
This method allows a reduction in production delays and a
synchronization of the decisions with all the specialists,
enabling an overall optimization of choices.
Simultaneous engineering uses DFM concepts which need
to model production problems. A first step in the DFM of
milled freeform surfaces was proposed by Duc [20]. His idea
is to define several types of surfaces: functional, linking, and
connecting surfaces. The author proposes that the functional
surfaces can be completely defined by functional requirements. The linking surfaces are defined in accordance with
the machining requirements, and the connecting surfaces
(frequently corner radii) are mainly defined by the machining
operation. The application of these concepts to impeller
design leads to two geometrical consequences. The first is the
definition of simple free forms for the hub and the second is
the definition of the blade shank by the machining problems.
Simultaneous engineering requires good communication
between all the designers, which necessitates the use of a
common support. Nowadays, this is a numerical model,
generated using CAD software, called the CAD model.
Furthermore, each specialist commonly has specific software

Int J Adv Manuf Technol (2011) 57:453463

to accurately optimize the impeller geometry (flow software,


mechanical software and CAM system) because integrated
CAD software does not propose a sufficiently high level of
optimization for specific applications. Thus, each specialist
must have two software applications (CAD plus specific
software), which increases the risk of information loss.

3 Proposed design method

457

mechanical trade rule is: blade thickness can never be equal


to zero. Then, these models are classified using the indicator
values. Finally, the impellers which obtain the best results are
studied in detail to define the best parameter solution.
The main difficulty of this method is the definition of an
appropriate parametric CAD model; this model must be
defined via meetings between all the specialists involved in
the design, production, and use of the impeller.
Therefore, it is necessary to quickly quantify the aptitude
of a model to be manufactured.

3.1 Design method


3.2 Application example
The proposed design method is based on the definition of
numerical indicators to quantify the aptitude of an impeller
CAD model to ensure functional requirements. This method
is based on the parametric CAD model and the use of the
optimization features of CAD software. Figure 6 illustrates
the proposed design approach. First, the functional requirements are defined and quantified. Then a parametric CAD
model of the desired impeller is created. This model is
constructed using knowledge and feedback from industry.
The number of parameters must be high in order to express
all the functional requirements of all the designers (flow,
mechanics, and machining). The boundaries of each
parameter must be defined to ensure the stability of the
algorithm. Next, an algorithm, presented in Fig. 7 and
detailed subsequently, is used to optimize the parameter
values. The algorithm finds the optimal solution within a
set of tested parameter combinations.
Thus, the parameter optimization method uses CAD
software features. Central to this method are the indicators
and the parametric CAD model. This algorithm defines a large
set of randomized parameter combinations. Each combination
generates a defined CAD model. The values of the functional
indicators are computed for each model and used to select the
adequate models using trade rules. An example of a

To illustrate the method presented in Section 3.1, the process


is performed for a centrifugal pump. For a pump, the main
functional requirements are: flow efficiency (flow rate, input,
and output pressure), rotations per minute and overall
dimensions. Figure 8 illustrates an example of the method
used to obtain the parametric CAD model. First, the flow
passage 2D profile is defined using a set of control points.
The coordinates of these control points are placed in the
parameters of the CAD model. Using this 2D profile, the hub
and the tip surfaces are built. Next, the intrados and extrados
surfaces are defined, based on the two directrix curves and a
set of rules. The extremities of the rules and the coordinates
of the points which determine the directrix curves are present
in the parameters. At this stage, one blade has been defined
completely; a circular repetition is then performed to define
the other blades. The number of blades is added to the CAD
model parameters. To conclude, the final details (corner
radius, etc.) are added to the CAD model.
Now the parametric CAD model is defined and the
optimization algorithm (see Fig. 7) can be used to obtain
the best solution.

4 Modeling of impeller manufacture


Fig. 6 New proposed design
method

To define the manufacturability indicators of a geometrical


model, the machining problems are studied. The main
process stages for an impeller are:

Stage
Stage
Stage
Stage
Stage

10,
20,
30,
40,
50,

rough cutting;
impeller shape turning;
blade milling;
grinding of impeller seating;
dynamic balancing of the impeller.

Stages 10 and 20 are controlled steps; their main


problems are the definition of the cutting parameter,
because impeller material commonly has poor machinability as has been presented by Ezugwu [2]. This problem
does not have an influence on impeller geometry. Thus,
only stage 30 is considered; it can be divided into:

458

Int J Adv Manuf Technol (2011) 57:453463

Fig. 7 Optimization method of


the CAD model parameters

Flow passage roughing,


Blade semi-finishing,
Hub semi-finishing,
Hub finishing,
Blade finishing.

Young [1] explained that flow passage roughing is the


longest step in the machining process and the step where
the maximum economy can be made. However, the final
geometry is created by blade and hub finishing. Dugas [21]
proposed a full simulation of the machining operation. The
author explains that the main machining errors come from:
the kinematical error of the milling machine, the geometrical error of the tool path, and flexions of the tool and the
workpiece. In this paper, milling machine kinematical
errors are not considered when defining the manufacturability indicators of the impeller, because this problem is
mainly linked to the tool path of the milling tools, unknown
at the design stage.
4.1 Geometrical tool path errors

Fig. 8 Definition of the parametric CAD model

This problem relates to the finishing stages, especially the


finishing of the blades. The problem is interference between
the tool and the nominal CAD model. In the following, a
positive interference, called overcut, has a positive value
and an undercut has a negative value.
Blade finishing may be carried out using point and flank
milling strategies. These two strategies have been presented
and compared by Tnshoff [22] in a European project:
FLAMINGO (FLAnk MIlliNG Optimization). The author
explains that flank milling allows reductions in both
machining time and production costs. For these two

Int J Adv Manuf Technol (2011) 57:453463

reasons, flank milling is commonly preferred over point


milling for blade machining. Rehsteiner [6] explained that
only developable surfaces can be machined without
interferences in flank milling. Many studies [614] have
dealt with the problem of the flank milling of twisted ruled
surfaces. All these studies explain that this interference
problem is linked to the surface twist. Thus, the geometrical
problem is mainly linked to the surface twist.
4.2 Flexions of the tool and the workpiece
During machining, forces are generated which induce
flexions. Many authors have studied tool flexion; Lpez
de Lacalle [23] optimized the two angles of tilt to limit
geometrical errors in point milling; Landon [24] and Larue
[25] modeled the errors generated by tool flexion; Ferry
[26, 27] also modeled these flexions and adapted the tool
feed rate to limit the generated geometrical errors; Suh [28]
and Dpinc [29, 30] modified the tool path to correct tool
flexion. Furthermore, in the case of impellers the flexions
are amplified, because:

The blade geometry is a thin-wall structure;


Access to the machined surface is difficult because the
space between two consecutive blades is reduced;
The machinability of impeller materials is poor.
Ezugwu [2] studied in detail the reasons for this low
machinability.

Flexions of the tool and the blade are therefore major


problems in impeller machining.
To conclude, the two key issues in impeller machining
are the twist of the intrados/extrados surfaces and blade/tool
flexions.

5 Proposed impeller manufacturability indicators


The new method, proposed in Section 3, is based on trade
rules and indicators. The trade rules and indicators are,
respectively, used to eliminate CAD models which could
not be accepted by a functional requirement and to classify
the tested CAD models. In the next section, trade rules and
indicators associated with the machining requirements are
presented; those associated with the hydraulic and mechanical requirements are not presented. The previous section
highlighted the fact that the two key issues of impeller
manufacturing are linked to blade/tool flexion and to the
intrados/extrados twist.

459

which reduce this value, and the mechanical requirements


which tend to increase it. Thus, the machining requirement
is the same as the mechanical requirement, and the greater
blade thickness requirement will be expressed by a
mechanical indicator and cover mechanical and machining
needs.
Furthermore, blade machining imposes the use of long
tools with a reduced diameter to access the surfaces
between two blades. The defined indicator therefore
expresses the minimum tool overhang necessary to machine
the blade surfaces. Eq. 4 illustrates this computation.
Overhang

Lmin
Dmax

With Lmin the minimum tool length and Dmax the maximum
tool diameter.
Dmax is computed by determining the minimum distance
between an intrados point (Sin(uin, vin)) and an extrados point
(Sex(uex, vex)), for consecutive intrados and extrados Eq. 5.
Dmax
n
o
min Sin uin ; vin ; Sex uex ; vex 8uin ; vin 2 0; 12 8uex ; vex 2 0; 12

5
Qi [4] proposed in his studies to define the minimum
tool length, Lmin, as the minimum surface rule length
between the intrados and extrados (Eq. 6).
n
o
Lmin max hin uin ; hex uex 8uin ; vin 2 0; 12 8uex ; vex 2 0; 12

6
Thus, the overhang indicator expresses the tool flexion
problem; although this computation is for a cylindrical tool,
it covers all flexion problems (cylindrical and conical
tools). Overhang is also used as a trade rule, limiting its
value to 4, for example. Thus, all tested CAD models are
eliminated if the overhang value is greater than 4.
5.2 Geometrical problem indicator
Section 4.1 explains that the geometrical machining
problem is linked to the value of the surface twist. The
twist is mathematically defined for ruled surfaces only. The
following therefore presents the definition of manufacturability indicators for an impeller with functional surfaces
modeled by ruled surfaces. Then, an equivalent twist is
defined for surfaces which are not ruled in order to compute
manufacturability indicators.

5.1 Flexion problem indicator

5.2.1 Ruled surfaces

Blade flexion is mainly linked to blade thickness. This


thickness is a compromise between the flow requirements

The next indicators express the geometrical problem linked


to the twist of the impeller surfaces. Two indicators are

460

Int J Adv Manuf Technol (2011) 57:453463

defined, one to define the maximum twist value (Eq. 7) and


other the average twist value (Eq. 8).
Twist maxfain u; aex u8u 2 0; 1g
R1
Twist

ain u  du

R1

aex u  du

Table 1 expresses all the combinations of these two


indicators. The indicators are complementary, Twist quantifying the amplitude value of the interferences and Twist
the generalization of this problem, although they are linked:
twist is necessarily greater than Twist.
These indicators may also be used to define trade rules
and to select the finishing strategy for impeller blades.
Figure 9 illustrates the cases presented in Table 1 and the
limits of the use of a flank milling strategy. These limits are
proposed for a simple tool path algorithm and common
accepted interferences. Thus, if a company chooses, for
economic reasons, to employ only a flank milling strategy,
this will induce the following two trade rules: the maximum
acceptable value of indicator twist is 25 and that of Twist
is 20 (these limits are defined for usual cases).
5.2.2 Non-ruled surfaces
The definitions presented subsequently may only be used
for surfaces which are not ruled but close to being ruled;
then numerical elements are proposed to define the limit.
Firstly, the apparent twist is defined; this value is
computed using Eq. 9. N0(u) and N1(u) are respectively
the orthogonal projection of N0(u) and N1(u) (see Section 2.1)
on a plane perpendicular to segment [P0(u) P1(u)].
0

aapparent u jN 0 u; N 1 uj

Equation 11 is inverted to obtain the value of the twist,


gap, which creates a maximum interference value equal to
the gap. A sign is allocated to the gap value; Fig. 10
illustrates the notation used: a positive value for a convex
surface and a negative value for a concave surface. The sign
of gap gives rise to two cases:

Then, Eq. 10 is used to compute the maximum gap


between segment [P0(u)P1(u)] and the isoparametric curve
of the surface which is near to a rule. The definition of an
equivalent rule may therefore be only used if all surface
gap values are lower than 0.5 mm.
gapu maxfdistanceS u; v; P0 uP1 ug8v 2 0; 1

Subsequently, this gap is converted into a gap twist,


gap. A large number of papers deal with the relation
between interferences and the twist [6, 8, 9]. To convert the
value of gap into a gap twist value, it is necessary to
select a method to position the tool on the surface. We
selected single point offset (SPO) positioning, presented by
Liu [8], because it is simple and it may easily be obtained
with many CAM systems. This positioning places the tool
tangent to the surface in the middle of the rule and the tool
axis collinear to the rule. Equation 11 is obtained using this
positioning and the hypotheses presented by Chaves-Jacob
[7] which are: the tool is a perfect mathematical cylinder,
the directrix curves are approximated by their tangents
computed at the extremity of the considered rule, and the
twist variation is linear along the rule. This equation
expresses the value of the maximum interference generated
by the tool, , depending on the twist value of the surface,
, in the considered point. The value of the tool radius, r, is
selected as equal to one-half of Dmax (Eq. 5).

 a 
" r  1  cos
11
2

10

If gap is positive, the tool is lifted, reducing the overcut


problem generated by the SPO positioning at points
P0(u) and P1(u) [8]. Thus, in Eq. 12, the sign of gap is
negative to obtain a reduction of equivalent compared to
apparent, in Eq. 13.
If gap is negative, the interference problem generated
by the SPO positioning is amplified by gap, hence the
positive sign in Eq. 12.

gap
"
 arc cos 1 
agap 2 
12
r
jgapj

Equation 13 therefore defines the equivalent twist for a


non-ruled surface. Then, as for a ruled surface, the values of
the manufacturability indicators twist and Twist are

Table 1 Twist indicator configurations and generated interference problems

Twist
Twist
Impact on interference
problems

Case 1

Case 2

Case 3

Low
Low
Few interference
problems

High
Low
Large interference problem in amplitude,
but localized

High
High
Large generalized interference
problem

Int J Adv Manuf Technol (2011) 57:453463

461

needs to be condensed into one indicator: Imanufacturability.


This indicator is a linear summation of the indicators
presented, considering that the following three factors are
equivalent for a flank milling strategy:

An overhang of 4;
A twist of 25;
A Twist of 20.

Furthermore, if the values of twist and Twist impose the


use of point milling (see Fig. 9) then the surface twist is not
a problem for machining. Considering that point milling is
less productive than flank milling [22], its manufacturability indicator is penalized as presented in Eqs. 14 and 15,
which define the Imanufacturability indicator. The lower the
value of Imanufacturability is, the easier impeller machining is.
Fig. 9 Point milling and flank milling zones

Imanufacturability
defined. Note that an absolute value is used to compute the
equivalent twist, because if gap is positive, gap may
overcompensate for the overcut problem caused by the SPO
positioning, inducing an equally problematic undercut
problem.


aequivalent aapparent agap 
13

6 Impeller manufacturability classification


6.1 Manufacturability indicator for a turbomachine
component
To use the presented design methodology (Section 3) it is
necessary to quantify manufacturability using only one
indicator. Thus the three indicators presented in Section 5

14

With:
8
8
< Twist
< twist

if
twist

25
if twist  20
and T
T
25
20
:
:
1 if twist > 25
1 if twist > 20

15
Thereby, in Eq. 15 if the point milling strategy is used,
the values of T and T are limited to one because point
milling is indifferent to the surface twist.
6.2 Classification examples
To illustrate the usefulness of the manufacturability
indicator, it is used to classify five turbomachine components, representative of the three impeller families (see
Fig. 3). The five turbomachine components are presented in
Fig. 11:

Fig. 10 Influence of gap sign

Overhang
T T
4

Inducer versions 1 and 2 are used at the entrance of two


variants of the turbo pump presented in introduction.
Turbo pump impeller. This impeller is used in the same
turbo pump as the new inducer version.
Pump impeller. This is used to pump a flow in a turbo
pump.
Blisk.

All these workpieces have intrados and extrados modeled by ruled surfaces, except the blisk. Indeed, for the
blisk, the definition of equivalent twist, presented in
Section 5.2, is used. The more twisted rule of this surface
has an apparent twist equal to 34.2 and a maximum gap
of 0.08 mm, inducing an equivalent twist of 54.7 using
Section 5.2. Values of Table 2 are computed using the
impeller CAD models and Section 5.

462

Int J Adv Manuf Technol (2011) 57:453463

Fig. 11 Examples of manufacturability classification of


impeller

This paper develops a new method for workpiece design,


applying the presented methodology to turbomachine
components. This method uses parametrical CAD models
to generate a large randomized number of geometries
whose performances are tested. For each tested CAD
model, flow-based, mechanical and manufacturability performances are evaluated and quantified by numerical
indicators which express functional requirements. This
paper presents only the requirements and the manufacturability indicator. These requirements are the definition of an
easily manufacturable geometry while reduce machining
errors and production costs. The key issues of impeller

manufacturability are therefore tool flexion during machining and interferences between the tool and the nominal
CAD model. Thus, three indicators are developed to
express these problems and are synthesized in a single
manufacturability indicator. This indicator is defined for all
types of turbomachine components (inducers, impellers and
fans) although their morphologies and the mathematical
modeling of their blades are different (ruled surfaces and
free-form surfaces). Indicators expressing flow-based and
mechanical requirements are defined (they are not presented in this paper); all these indicators are used to classify
the CAD model geometries, ascribing equal importance to
manufacturing, flow, and mechanics. The presented method
consequently defines an impeller geometry which verifies
the flow performances and mechanical strength, with
controlled manufacturing errors and production costs.
Finally, to illustrate the usefulness of the defined
manufacturability indicator, it is used to classify the
manufacturability difficulties of five industrial impellers
(representative of the three impeller families).
Future perspectives following on from this study include
the generalization of the presented design method to other
types of surfaces, in order to define a range of manufacturability indicators applicable to all types of manufacturing
problems.

Table 2 Manufacturability indicators of five impellers

Lmin (mm)

Dmax (mm)

Overhang

Twist ()

Twist ()

Imanufacturability

48
50
20
25.5
62.5

14
14
12
5.14
10

3.43
3.57
1.67
4.96
6.25

8.70
10.50
14.80
32.30
54.73

6.70
7.08
7.84
9.09
26.90

1.54
1.67
1.40
2.69
3.56

Values of Table 2 are used to classify the five impellers


presented in Fig. 11. Thus, the defined indicators allow a
numerical comparison of the manufacturability between
two inducers: for example, inducer version 2 is 8.4% less
machinable than version 1, and this is caused by an increase
in tool flexion and interference. Furthermore, these indicators enable a manufacturability comparison between several
impeller families: for example the presented blisk is 2.5
times less machinable than the impeller.

7 Conclusions

Inducer version 1
Inducer version 2
Impeller
Pump impeller
Blisk

Int J Adv Manuf Technol (2011) 57:453463

References
1. Young HT, Chuang LC, Gerschwiler K, Kamps S (2004) A five-axis
rough machining approach for a centrifugal impeller. Int J Adv
Manuf Technol 23(34):233239. doi:10.1007/s00170-003-1677-z
2. Ezugwu EO, Bonney J, Yamane Y (2003) An overview of the
machinability of aeroengine alloys. J Mater Process Technol 134
(2):233253. doi:10.1016/S0924-0136(02)01042-7
3. Chuang LC, Young HT (2007) Integrated rough machining methodology for centrifugal impeller manufacturing. Int J Adv Manuf
Technol 34(1112):10621071. doi:10.1007/s00170-006-0675-3
4. Qi R, Liu W, Bian H, Li L (2009) Five-axis rough machining for
impellers. Front Mech Eng Chin 4(1):7176. doi:10.1007/s11465009-0010-4
5. Bedi S, Ismail F, Mahjoob MJ, Chen Y (1997) Toroidal versus
ball nose and flat bottom end mills. Int J Adv Manuf Technol 13
(5):326332. doi:10.1007/BF01178252
6. Rehsteiner F (1993) Collision free five-axis milling of twisted
ruled surfaces. Ann CIRP 42(1):457461. doi:10.1016/S00078506(07)62485-8
7. Chaves-Jacob J, Poulachon G, Duc E (2009) New approach to 5axis flank milling of free-form surfaces: computation of adapted
tool shape. Comput-Aided Des 41(12):918929. doi:10.1016/j.
cad.2009.06.009
8. Liu XW (1995) Five-axis NC cylindrical milling of sculptured
surfaces. Comput-Aided Des 27(12):887894. doi:10.1016/00104485(95)00005-4
9. Rubio W, Lagarrigue P, Dessein G, Pastor F (1998) Calculation of
tool paths for a torus mill on free-form surfaces on five-axis
machines with detection and elimination of interference. Int J Adv
Manuf Technol 14(1):1320. doi:10.1007/BF01179412
10. Redonnet JM, Rubio W, Dessein G (1998) Side milling of ruled
surfacesoptimum positioning of the milling cutter and calculation of interference. Int J Adv Manuf Technol 14(7):459465.
doi:10.1007/BF01351391
11. Gong H, Cao LX, Liu J (2005) Improved positioning of
cylindrical cutter for flank milling ruled surfaces. CAD Comput
Aided Des 37(12):12051213. doi:10.1016/j.cad.2004.11.006
12. Pechard PY, Tournier C, Lartigue C, Lugarini JP (2009)
Geometrical deviations versus smoothness in 5-axis high-speed
flank milling. Int J Mach Tools Manuf 49(6):454461.
doi:10.1016/j.ijmachtools.2009.01.005
13. Menzel C, Bedi S, Mann S (2004) Triple tangent flank milling of
ruled surfaces. Comput-Aided Des 36(3):289296. doi:10.1016/
S0010-4485(03)00118-0
14. Senatore J, Landon Y, Rubio W (2008) Analytical estimation of
error in flank milling of ruled surfaces. Comput-Aided Des 40
(5):595603. doi:10.1016/j.cad.2008.02.007
15. Chu CH, Chen JT (2006) Tool path planning for five-axis flank
milling with developable surface approximation. Int J Adv Manuf
Technol 29:707713. doi:10.1007/s00170-005-2564-6

463
16. Peternell M, Pottmann H, Ravani B (1999) On the computational
geometry of ruled surfaces. Comput-Aided Des 31(1):1732.
doi:10.1016/S0010-4485(98)00077-3
17. Young HT, Chuang LC (2003) An integrated machining approach
for a centrifugal impeller. Int J Adv Manuf Technol 21(8):556
563. doi:10.1007/s00170-002-1382-3
18. Boothroyd G (1994) Product design for manufacture and
assembly. Comput-Aided Des 26(7):505520. doi:10.1016/00104485(94)90082-5
19. Al-Zubaidy SN (1995) A proposed design package for centrifugal
impellers. Comput Struct 55(2):347356. doi:10.1016/0045-7949
(93)E0050-X
20. Duc E, Lartigue C, Tournier C, Bourdet P (1999) New concept for
the design and the manufacturing of free-form surfaces: the
machining surface. CIRP Ann Manuf Technol 48(1):103106.
doi:10.1016/S0007-8506(07)63141-2
21. Dugas A, Lee JJ, Hascot JY (2002) An enhanced machining
simulator with tool deflection error analysis. J Manuf Syst 21
(6):451463. doi:10.1016/S0278-6125(02)80051-6
22. Tnshoff HK, Gey C, Rackow N (2001) Flank milling optimizationthe FLAMINGO project. Air Space Eur 3(34):6063.
doi:10.1016/S1290-0958(01)90058-9
23. de Lacalle LN Lopez, Lamikiz A, Sanchez JA, Salgado MA (2007)
Toolpath selection based on the minimum deflection cutting forces in
the programming of complex surfaces milling. Int J Mach Tools
Manuf 47(2):388400. doi:10.1016/j.ijmachtools.2006.03.010
24. Landon Y, Segonds S, Lascoumes P, Lagarrigue P (2004) Tool
positioning error (TPE) characterisation in milling. Int J Mach Tools
Manuf 44(5):457464. doi:10.1016/j.ijmachtools.2003.12.001
25. Larue A, Anselmetti B (2003) Deviation of a machined surface in
flank milling. Int J Mach Tools Manuf 43(2):129138.
doi:10.1016/S0890-6955(02)00189-X
26. Ferry W, Altintas Y (2008) Virtual five-axis flank milling of
jet engine impellersPart I: mechanics of five-axis flank
milling. ASME International Mechanical Engineering Congress
and Exposition. Proceedings 3: 339353. ISBN: 978-0-79184297-3
27. Ferry W, Altintas Y (2008) Virtual five-axis flank milling of jet
engine impellerspart II: feed rate optimization of five-axis flank
milling. ASME International Mechanical Engineering Congress
and Exposition. Proceedings 3: 355369. ISBN: 978-0-79184297-3
28. Suh S, Cho J, Hascoet JY (1996) Incorporation of tool deflection
in tool path computation: simulation and analysis. J Manuf Syst
15(3):190199. doi:10.1016/0278-6125(96)89571-9
29. Dpinc P, Hascot JY (2005) Active integration of tool deflection
effects in end milling. Part 1. Prediction of milled surfaces. Int J Mach
Tools Manuf 46(9):937944. doi:10.1016/j.ijmachtools.2005.08.005
30. Dpinc P, Hascot JY (2005) Active integration of tool
deflection effects in end milling. Part 2. Compensation of tool
deflection. Int J Mach Tools Manuf 46(9):945956.
doi:10.1016/j.ijmachtools.2005.08.014

Вам также может понравиться