You are on page 1of 9

Structural Optimization 18, 193 201 @ Springer-Verlag 1999

Finite element analysis and optimization in fixture design


S. K a s h y a p *
Department of Mechanical Engineering, Rennselaer Polytechnic Institute, Troy, NY, USA
kashyap@mach.pr.luceut.com

W . R . DeVries~
Department of Mechanical Engineering, Iowa State University, Ames, IA, USA

Abstract Fixturing locating point synthesis considers the optimization techniques and a comparative study of differ-
workpiece and the fixturing elements to be rigid, but however ent optimization techniques were not employed in fixturing
they are elastic and deformable. To ensure sustained quality of analysis. In this paper finite element analysis will be used
manufacture to meet the design tolerances, fixture design must be for assessing the deformation in workpiece and fixturing ele-
predictably repeatable. This paper is concerned with minimizing
ments. Secondly, optimization procedures will be adopted to
deformation of the workpiece due to machining loads about fixtur-
ing support positions, especially in thin castings. Finite element minimize these deflections, in addition to determining if the
analysis is used in simulating the deformation of the workpiece physical properties and the cross-section of the locating tool
at selected points. An optimization algorithm is developed to points are adequate to resist the machining forces.
minimize deflections at these selected nodal points by considering Specifically, the goals for fixture design analysis and op-
the support and tool locations as design variables. The result- timization are as follows.
ing support locations and tool point designs ensure part support,
kinematic closure and minimal workpiece deflections during ma-
chining. Describe the FEM methods used in computing the de-
flections in the workpiece and the fixturing elements, and
formulate a finite element model for fixturing analysis.

Treat the fixture design as an optimal design problem by


1 Introduction
formulating an objective function to determine the loca-
Research in fixture design considered the workpiece and the tion of fixturing supports.
fixturing elements to be rigid for kinematic analysis. But,
in fact both the workpiece and the fixture are elastic and Describe the closed form solutions used to determine the
deformable. In this paper, the elastic deformation of the appropriate cross-section and physical properties of the
workpieee, primarily normal to the primary reference plane locating tool points.
is of interest. In particular, optimization techniques to min-
imize deflections of the workpiece are outlined that select
positions of the three supports which are in the primary ref- 2 P r e v i o u s r e s e a r c h i n f i x t u r i n g f i n i t e e l e m e n t anal-
erence plane. In addition, it is determined whether the phys- ysls a n d o p t i m i z a t i o n
ical properties and cross-section of the locating tool points
are adequate (or rigid) to resist machining forces. One of the requirements of fixturing is that it should limit
The machining forces and the clamping forces imposed on elastic deformation of the workpiece. It is important to con-
the workpiece and the fixturing elements bring about some sider cutting forces as well as clamping forces on workpiece.
deformation in them. Quality of manufacturing is high only A finite element model of the workpiece is created and cut-
when the workpiece is accurately and rigidly positioned or ting and clamping forces can be identified and quantified and
it is recursively sensed and compensated for. With these their effect on the behaviour of the modelled workpiece de-
deformations in the workpiece and fixturing elements, the termined.
quality of manufacture is compromised. Lee and Haynes (1987) used FEA to minimize fixturing
In the past, little analysis work was devoted to the design force/workpiece deflection. They were able to determine the
of fixtures. With the recent development of new techniques, deflection of a part for a specific fixture assembly. Menassa
software, and mathematical theories, much effort can be ex- and DeVries (1991) further extended this work using fixtur-
pended beyond design synthesis for locating points. Past re- ing elements as an elastic body. DeMeter (1996) and Hock-
search in optimization in fixtures could not handle supports enberger et al. (1995) formulated micro-displacements at the
not necessarily on a single plane. Use of better and faster workpiece and fixturing element interface with optimization
* Presently at Lucent Technologies/BellLaboratories Princeton, NJ possibilities. More recently, Cai et al (1996) have approached
t Formerly at RensselaerPolytechnic Institute Troy, NY this optimization problem for sheet metal fixtures.
194

3 F o r m u l a t i o n of t h e o b j e c t i v e f u n c t i o n only at a candidate set of nodes to our interest. This leads


to the following conclusions.
One of the most important technical criteria in machining
fixture design is deflection of the workpiece due to either ma- 9 The effect of arbitrarily positioned load points on a work-
chining, and/or clamping (Lee and Haynes 1987). Minimizing piece causes critical deflections only at the corner nodes,
deflections, thus becomes a key point for the fixture design and the mid-edge nodes.
process to ensure repeated highest quality manufacture. It
becomes necessary to formulate an objective function to take 9 Maximum deflections occur at the nodes at which the
loads are applied.
into effect the worst possible applied manufacturing and as-
sembly/clamping loads on the workpiece deflections.
This is true in many metal working, or assembly opera- 3.2 Objective function for deflection minimization
tions. In fabrication, especially in drilling or riveting, deflec-
The objective function for minimizing the deflections of a
tions are important. And also for the accuracy in achieving
three-dimensional workpiece can formulated from the obser-
acceptable tolerances in the manufactured part, and increas-
vations in the last section. The design problem is to optimally
ing the repeatability of the fixtures workpiece-fixture deflec-
position the supports, Si, i = 1 , . . . , N S. T h e point of appli-
tions are important design constraints. Because of these prac-
cation of the load has the large deflection as observed from
tical considerations, the analysis in the following sections is
the previous section. So the objective function would be to
aimed at determining the positions of the fixture supports
minimize deflections at these locations ~:i for NP loads, along
and minimizing the deflection of certain key points on the
with deflections at N c corners denoted by Ci, and at the N rn
workpiece, given the workpiece geometry, material and an
mid edge nodes denoted by Mi, where D denotes the deflec-
initial fixture layout as determined from the initial fixture
tion vector. The objective function is given below (Kashyap
design.
1997; Kashyap and DeVries 1997, 1998)

Nc
3. I Finite element mesh of lhe workpiece
Fi(D, S) -_ E [ D T ( c i ) W D ( C i ) ] +
Mostly prismatic workpieces with some simple curved sur- i=1
faces are chosen for this analysis. The mesh is a tetrahedral
mesh with isotropic material properties. The mesh consists N -~ Np
of N e octant elements and with N p nodal points, as shown in E(DT[Mi)WD(Mi)] + E[DT(Li)WD(Li)] , (2)
Fig. 1. There are six tetrahedral elements embedded within i=1 i=1
each of the N e octant elements. In many assembly and ma-
Nc N~
chining processes (like drilling, fastening, etc.), the part is
supported on the bottom surface of the workpiece. The 3-2-1 Fi(D , s ) = ~ 1115(Ci)[I + Z III5(M~)tE+
locating principle has three supports that form the primary i=1 i=1
plane, since the three points define a plane. Considering a Np
primary supporting plane that rests on the three flxturing
E IIb(Li)N (3)
supports S i (PNT0, PNT1, and PNT2 as shown in Fig. 2).
i=1
A point load i~ applied normal to the surface and
translated vertically down is denoted by the vector = sum of corner deflections + sum of mid-point deflections

+ sum of load point deflections,


P l = (Pzl, P y l , P z l ) T = ( 0 . , 0 . , P z l ) T . (1)
where [ } ( C i ) , f ) ( g i ) , D ( L i ) in (3) (Menassa and DeVries
When considering an arbitrary edge, numerical results 1991) are the weighted displacement vectors at the corner,
have shown the following observations. Maximum deflections mid-edge, and the load point node respectively. And W is
occur at corner nodes Ci and at a node in-between the the the matrix that weighs the importance of deflections in the
two 'corner nodes. When calculating these deflections with three orthogonal directions. The objective function is essen-
the FEM, a refined mesh is desired to provide accurate deflec- tially a quadratic expression.
tions, but is uneconomical while considering the large number The objective function in (3) is set up to handle sev
of tetrahedral elements. Therefore determination of the ex- eral load cases simultaneously, which constitutes the prac-
act location of the maximum deflection at the middle node tical cases that might arise such as gang-drilling operations.
between the two corner nodes Ci, along each edge is not prac- Process plans require that the part be manufactured by a
tical. From FEM simulation results it is observed that when sequence of manufacturing processes requiring many fixture
a load is far from an arbitrary edge, maximum deflection oc- setups. This is a costly practice as from a practical point of
curs near the mid-edge node and when the load is close to the view since fixtures have to be separately designed, fabricated,
arbitrary edge the deflections shifts along the edge towards and stored. Also, workpiece have to be refixtured for each op-
the load. Thus it is reasonable to minimize the deflections at eration. This leads to a loss of accuracy in positioning and
the corner nodes, mid-edge nodes and nodes at the point of increases the lead time in manufacturing if the fixture has to
load application, making it possible to minimize deflections be fabricated or reconfigured.
*Figures r e p r e s e n t the default P r o / E N G I N E E R c o o r d i n a t e axes The solution is to reformulate (3) to optimize N g loads
where the Y-axis is t h e d e p t h simultaneously, in order to generate a single robust location
195

T y

Tetrahedron < ~
element int

jz

"-4
Octant Element

Fig. 1. Tetrahedral mesh

T freedom and flexibility to choose a choice of nodal deflections,


load cases, and other technical parameters of interest into the
objective function at the user's will. The exact implementa-
tion of this problem will be discussed later in this paper,

II[Fi(D, S)] = E { D ( C i ) , D ( M i ) , D(Li) } . (4)


i=1

Alternatively, the deflection minimization can be defined


as a set of optimum support locations O(si) which results in
a minimum of the objective function value II[Fi(D, S)] as in
(5),
VSk E G k, O(Sk) = rain II[Fi(D,S)], (5)
s~ co(sk)
where G k defines the geometrically feasible solution set as
defined by the bounds on supports S i.
r

Fig. 2. Support positions, corner and mid-edge nodes on the work- 3.2.1 Weighting function selection. Assembly forces like riv-
piece eting are mostly applied normal to the surface of the work-
piece. Under these circumstances the deflections normal to
the surface are most important, while deflections in the work-
design for supports. The new objective function which takes piece plane are neglected, as shown by the matrix in (6),
into effect all of the load cases can be formulated as in equa-
tion (4). In the exact implementation, the user can select O 0 0
nodal deflection visually from the FEM model into the ob-
jective function, or set a threshold on the minimum deforma-
tion needed and choose the nodal deflection exceeding this
minimum deflection into the objective function. The imple-
W=
I0 0
0 0
0
1

Most machining processes, except perhaps for drilling,


(6)

mentation of this problem is such that it gives the user enough even applied vertically on the top surface of the part are
196

usually three-dimensional. As a result the weighting matrix may have to be supported on different regions not necessar-
in this case would be a unity matrix as (7). While the choice ily co-planar. To facilitate these requirements, the user can
of the weights depend on the application and estimated as- sketch area/volume bounds on the support location solution
sembly and machining loads. The objective function in (3) set. These geometric area/volume bounds termed Geomet-
can be used to determine how to position the supports, ric Constraint Regions (GCR) can be sketched at the fixture

W=
[ 100 ]
0 1 0 . (7)
design stage on Pro/ENGINEER or at the analysis stage on
Patran 1.4.
In circumstances where there have to be more than two
0 0 1 supports existing within the same geometric constraint re-
gion, the constraints can be treated as inequality constraints
to keep the supports away from each other. Therefore in-
3.2.2 Remeshing. The fixture support locations are design equality constraints are needed to keep at least one degree of
variables for the optimization problem and also constraints freedom for supports S 2 and Sa. Both supports in this case
to the FEM model. The optimtzation algorithm changes the are allowed to have the same x-coordinate but not the same
values of these support locations, the nature of the FEM z-coordinate. In general the constraints can be designed as
problem also changes to become nonlinear. The optimiza- follows:
tion algorithm will not match the already existing nodal val-
ues with the optimum locations. Hence there is a need for xsi=~ij.Xsj, i,j=1,2, ..,N s , i e j , (8)
remeshing the FEM model, but this remeshing can be done
locally such that the FEM node close to the optimized solu- Ysi =13ij "Ysj, i , j = 1,2, . . , N s , i C j (9)
tion is moved to replace the optimally generated node. This
remeshing should not be applied to a rather coarse mesh, be- Zsi=Sij.Zsj, i , j = 1,2, . , N s , i 7~ j (10)
cause as the tetrahedral elements become distorted and the
error in the evaluation of the nodal deflection increases. A where cqj,/3ij , 6ij are termed as support location feasibility
very fine mesh on the other hand is also not suitable since coefficients (SLFC) between zero to one. In practice the lower
it requires excessive CPU time to solve the mesh. For the and the upper bounds on the support locations are not de-
sired since a value of zero will force the supports to exist on
parts studied with surfaces to be machined with a tolerance
of 0.01 inch, we have about three thousand degrees of free- the boundaries of the GCR and a value of one may overlay
dom for good results. We were limited in increasing the num- one support on another.
ber of nodes on the FEM model because of the insufficient The above observation leads to the conclusion that, if
RAM available for processing and solving this FEM problem G k ( S i ) , k = 1,2, 3 defines the feasible space for the support
on a IBM RS6000 work-station using Pro/ENGINEER and locations S k as defined by the optimal locations O ( S k ) , then
Patran 1.4. In cases of machining dies for thin castings with G k are collectively disjoint,
surfaces and designing fixturing for metal stamped parts with
surfaces. It is very critical to maintain dimensional and form VSk e G k ; if O(Sk) N G k ( S i ) # { O } ; then
tolerances. A more accurate representation of the surface can
only be maintained by a finer mesh with increased number G 1 N G 2 r ] G a = {1~}, (11)
of nodes.
In summary, the FEM computes the deflections D j, based
on the loads and constraints from the (3) which can be evalu-
3.3 Geometric constraints
ated for N l load cases. Then the design problem for locating
The objective function consists of the terms given in (3) which supports can be posed as the following optimization problem.
comprise the deflections at the selected nodes expressed as a Given
sum. The objective function needs constraints on the sup-
ports Si, i = 1,... N s, otherwise no deflection will occur as a S i = (Xsi,Ysi,Zsi) T i = 1, . . , N s,
result of placement of supports directly under the load. This
is however undesirable practically, for example in drilling a Ti = ( x t i , Y t i , z t i ) T i = 1, .,N t
through hole because it may result in a hole being made in
the fixture support. An over-hung problem will arise espe-
Pi = (Xpi,Ypi,Zpi) T i = 1, . . , N p,
cially in an airplane panels if the supports 8 i are placed close minimize
to the loads. So therefore this leads to a geometrically feasi- (Sk, k = 1 , . . . , N s) YI(Fi[:DJ' sk)] =
ble region where the supports can be placed providing part
support with no interference with the machining tool. An Nl
optimal solution for the support location can be determined minimize
(12)
for minimum deflection in the workpiece. ( S k , k = 1 , . . . , N s) X - ~ ' 'Z--'ritDj'Sk)
i=1
The constraints to this minimization problem are geomet-
ric in nature. Usually in fixturing assemblies, it becomes nec- Subjected to
essary to support the workpiece with temporary weldment.
As a constraint to this minimization problem, the support Xsi <_c~ij.xsj z,j= 1,2...N S i~k j
locations must not interfere with these weldment or other
features on the fixture/workpiece. In reality the workpiece Ysi <-~Sij'Ysj i,j= 1,2...N S i#j
197

zsi <_6ij.Zsj i,j=l,2...N S i~j


Pro/ENGINEER
xsi(MIN ) <_ Zsi <_ x s i ( M A X ) , i
Ysi(MIN) < Ysi < Ysi(MAX), | Perform [_ ID o fixturing~l
~anufammb- |
zsi(MIN) <_ zsi < zsi(MdX ) ,
where T i are the tool point locations, xsi(M[N, MAX),
ysi(MIN, MAX), and zsi(MIN, M A X ) are the lower and [ Create ~II P.e.rfo.rrn. 1! InteractivelY~lI Applyloads&~
upper bounds as defined by the GCR. IGe~ & | I mtmmmng | defineboundsll
And finally,
Y
o(~+i, vii, z+i)T e B(W),
PATRAN
where B ( W ) is the part boundary of the workpiece.
Alternative formulation. In reality, fixturing beds are ma-
chined with slots, or made up of peg holes for assembly of 1 + +
r so,+,+ +
fixturing components. These pre-defined locations on fixtur- ,-~ FEM problemII ,o:n,%n+++
ing beds can help pose the fixturing optimization problem
as a discrete optimization problem. This simplifies the fix-
turing optimization problem by directly linking the support
locations to the geometric model of the workpiece/fixture,
eliminating the need for performing a FEA. But since there /Optimize by[_ interactively~l IInteractivelyL-
are no standard fixturing beds of the same type, and since I define I / derma I
the research motivation is to develop a CAE tool to monitor
and diagnose deformations during machining. A continuous
objective function in conjunction with finite element analy- Fig. 3. Schematic of the framework
sis is deemed appropriate for this deflection minimization in
general.
a fairly simple example of a thin part considered for gang-
drilling operation where four holes have to be drilled simulta-
4 O p t i m i z a t i o n m e t h o d s , c o d e , a n d implementation neously. This example demonstrates how excessive deflection
can be reduced by optimization, this is particularly useful
Optimization methods implemented are by the penalty func- when thin castings have to be machined. The second ex-
tion methods, sequential programming, and simulated anneal- ample is a comprehensive one in that a design of fixture is
ing (Rekaltis et al. 1983). A solution to the example in sought to handle machining of several holes in a sequence and
this paper was obtained by using the penalty function ap- milling of a slot. The second example shows how the fixture
proach. These optimization methods were implemented as design methodology is robust to handle several operations in
C / C + + classes referring to these optimization methods from a single fixture setup.
the books by Rekaltis et aL (1983), Press et al. (1992), and
Vanderplats (1990). Most of the line searches and matrix
utilities were copied from the book by Press et al. (1992). 5.1 Example 1: drilling four holes simultaneously
These optimization routines are interfaced with Pa- This example is a fairly simple problem where a fixture is
tran 1.4 through Patran's Command Language (McNeal- designed to drill four holes simultaneously in a thin rectan-
Schwendler Corporation 1995) facilitating design automation
gular Aluminum workpiece (i.e. 4 inches • 3 inches x 0.25
in performing this fixturing optimization. P r o / D E V E L O P
inches). The fixture design process is first to determine the
API for P r o / E N G I N E E R was used for developing the fix- positions of the locating tool points and the reference data to
turing design tool. The suite of optimization methods that machine the workpiece. The positions of these locating tool
are accessible on the main Patran 1.4 menu as a result of this points along with machining loads pose as boundary condi-
design automation effort are Interior Penalty Function, Exte-
tions to the finite element model. The positions of the centres
rior Penalty Function, Method of Multipliers, and Sequential of the four holes and the drilling loads are 0.5 inch from their
Quadratic Programming. A schematic of the entire fixturing respective edges as shown in Fig. 4. The support positions
design and optimization integration is illustrated in Fig. 3. are optimized to mmimize workpiece deflection based on the
A range of values for SLFC's from { 0 , . . . , 1} in increments FEM simulation and the optimization procedure developed
of 0.1 are used in performing the post-optimality analysis on
in this paper. The constraints on this FEM model being three
the optimal solution for each support location.
fixed displacement constraints on the support locations, the
three tool point locations, and the four load cases.
5 F E A a n d d e f l e c t i o n minimization The elastic properties along with the coordinates of the
sample nodes and the magnitudes of the loads are displayed
To illustrate the optimal design methodology for fixture de- in Table 1. In addition Table 1 identifies the constraints for
sign analysis, consider two examples using the developed soft- the optimization process such as GCR and SLFC that are
ware on P r o / E N G I N E E R and Patran. The first example is imposed.
198

L3 Table 2. Results of Example 1, simultaneous drilling

Obj fun / / = 6.63 E-03 in


New obj fun f/* = 0.0
C4 ~ C
Supports N s = 3 S 1 = ( 1.5 in, 1 in, 0 in)
S 2 = ( l i n , 2.5in, 0 i n )
S 3 = ( 3 i n , 2in, 0in)

New supports S~ = ( 0 in, 0.5 in, 0 in)


S~ = ( 0.75 in, 1.5 in, 0 in)
8~ = ( 3.25 in, 3.0 in, 0 in)
Fig. 4. Tool, load and support positions on Example 1
Clamping force fc =175 lbs
Table 1. Values used in optimization for Example 1, simultaneous
drilling
mid-edge nodes ( M i , i = 1 , . . . ,4), and the four load point
nodes (Li, i = 1 , . . . , 4). The values for the displacements at
Workpieee dimensions 4 in x 3 in x 0.25 in
these twelve nodes before optimization and after optimization
are shown in Table 3. A maximum deflection of 0.003 inch
Elastic properties of A1 E = 1.0E7 Psi and u =0.3
was noted at C 1.
Corner nodes, N c - 4 Mid-edge node N m = 4 Table 3. Deflection minimization at sample nodes for Example 1

C1 = (0in, 0in, 0in) MI= (2 in, 0in, 0in)


C2 = ( 4 in, 0 in, 0 in) M2 = (4 in, 1.5 in, 0 in) Deflection 73z Deflection Minimization
C3= (4in, 3in, 0in) M3= (2 in, 3in, 0in)
C4 = ( 0 in, 3 in, 0 in) M4 = (0 in, 1.5 in, 0 in) Sample nodes Before opt After opt

Tool points N t T1 = (0.5 in, 3 in , 0.125 in) C1 -0.28E-2 0.0


T2 = (3.5 in, 3 in, 0.125 in)
T3 = (4 in, 0.5 in, 0.125 in) C2 -0.39E-2 -0.132

Load Nodes & Loads L1 = (0.5 in, 0.5 in, 0.25 in) C3 -0.104E-2 -0.7E-2
L2 = (3.5 in, 0.5 in, 0.25 in)
L3 = (3.5 in, 2.5 in, 0.25 in) C4 -0.53E-2 -0.6E-2
L4 = (0.5 in, 2.5 in, 0.25 in)
P1 = (0, 0,-300 lbf) M1 -0.43E-2 -0.6E-2
P4= P3 = P2 = P1
M2 0.023 0.022
GCR x = 0.75 in, y = 0 i n , z = 0 m
x = 3.25 in, y = 3 i n , z = 0 i n M3 -0.1E-2 -0.151

SLFC c~12 = c~13 = 0.8,/323 = 0.8 M4 -0.0 -0.25E-2

L1 -0.194E-2 -0.349E-3

5.1.1 Simultaneous drilling. Four loads are applied simulta- L2 -0.33E-3 -0.244E-3
neously to both ends of the workpiece and given the locating
tool point locations, the problem is to find the optimal loca- L3 -0.682E-3 -0.357E-3
tions for the supports. Figure 5 displays the vector plot of the
results from the FEM analysis. It can be inferred from the L4 -0.65E-3 -0.259E-3
Fig. 5 that the four simultaneous forces bends the part about
the support locations. Out of the all other penalty function
approachs used, optimization of this particular problem was
most efficient with the method of multipliers. The end results
are shown in Table 2.
5.1.2 Example 2: sequential drilling and milling. This ex-
The objective function for this example is a function of ample is concerned with drilling three holes sequentially, and
twelve nodes, the four corner nodes (Ci, i = 1 , . . . , 4), the four milling a slot in the workpieee. The CAD model with the
199

.003405

.003178

.002951

.(X)2724

.002497

.002270

.002043

.001816

.001589

.001362

.001135

.0009080

.0OO6810

.0004540

.0002270

Fig. 5. Vector plot of the deformation results for Example 1

forces and constraints attached onto the workpiece are shown forces for the tangential and the radiai components. The in-
in Fig. 6. The workpieee is made of aluminum with the elas- formation about the cutting conditions for machining the slot
tic modulus and a Poisson's ratio of E = 1.0 x 107 ibf/inc 2 are given Table 4. Because of these assumptions about the
and u = 0.3. physical problems, like loads applied to the arbitrary posi-
tions to the surface, the weighting matrix in (7) is used in
calculating the objective function (2). Since the loads that
0 Fx, - 2 0 0 ~ are assumed to be the result of drilling holes through the
Aluminum, the supports should not be directly under them.
0 Fx, For this reason, the GCR are defined to be a rectangular
area where the supports are located, the values of which are
0 Fx, -200 Fy, shown in Table 4. In this study two of the supports 81, S 2
can share the same value for the x coordinate, but must be
separate in their y coordinates. For this reason, inequality
constraints are used to keep supports '-'r and S 2 away from
each other. The SLFC's Ctl2 = c~13 = 0.8 will limit support
81 in the X-direction to 80 % of the X-coordinate of $2 and
83, fl2a = 0.8 will limit support 82 in the y-direction to 80
% of the distance of support 83, since all the supports are in
the same plane 6i = 1.0. These values are all shown in Ta-
ble 4. The selection of these constraints depend on practical
Fig. 6. Force constraints model for FEA and optimization, considerations, for example, the size of a support will limit
Example 2: sequential drilling and slot milling created on how close the supports can be to each other.
Pro/ENGINEER

In order to demonstrate the robustness of this optimiza-


With the reference data and locating tooi points deter- tion scheme consider drilling of three holes on the part and
mined by the initial fixture design, the constraints for solving milling of a slot on the workpiece one after another. The
the FEM model and the optimization for deflection are ap- objective is to minimize deflection considering all these ma-
plied. It will be shown that an optimal configuration can be chining loads by finding an optimal location for the supports
achieved that further minimizes the objective function. The for all of these machining loads considered simultaneously.
FEM mesh for the workpiece in example two is as shown in Table 5 summarizes the numerical values used to represent
Fig. 1. The milling force is applied to the entire face of the these loads and their locations along the initial support po-
slot floor after calculating the average force for the resultant sitions.
200

Table 4. Example 2: sequential drilling and slot milling values Table 5. Numerical values for drilling and milling considered su-
optimization multaneously

Workpiece envelope (10 in x 10 in x 5 in) Load point nodes & load L1 = (1.0 in, 0 in, 1.0 in)
L2 =( 1.0 in ,0 in, 9.0 in)
Elastic properties of A1 E = 1.0E7 Psi & u = 0.3 L3 = (5.0 in, -2.0 in, 5.0 in)

Corner nodes, N c = 4 Mid-edge nodes, N m = 4 Drilling force P1 = ( 0, -200 lbf, 0 )


P2 = P3 = P1
Cutter specification Diameter = 0.25 in
Number of teeth N t = 4 Milling force (203 lbf, -128 lbf, -358 lbf)
Helix angle = 0 degrees
Objective function II o = 2.303E-0.3 in
Cutting conditions f = 30 ipm New objective function H* = 1.624E-03
N = 550 rpm
b = 0.25 in Supports, N S = 3 301 = (2.5 in, -5 in, 5 in)
S~ = (2.5 in,-5 in, 7.5 in)
C1 = (0 in, 0 in, 0in) M1 = (0 in, 0 in, 5 in) s~, = ( 7 . 5 in, -5 in, 7.5 in)
C2 - (0 in, 0 in, 10in) M2 = (4 i n , - 2 in, 10 in) New supports, N S = 3 S~ = (0 in,-5 in, 2 in)
C3 = (10in, 0in, 10in) M3= (10in, 0in, 5 i n ) S~ = (6 in,-5 in, 0 in)
C4 = (10 in, 0 in, 0in) M4 = (4 i n , - 2 in, 0 in) 8~ = (6 in,-5 in, 6 in)
Tool points, N T =3 T1 = (10 in, -2.5 in, 7.5 in) Clamping force fc = 600 lbs
T2 = (10 in, -2.5 in, 2.5 in)
T3 = (2.5 in, -2.5 in, 0 in)

GCR Sl(X ) = [6, 10] in, Sl(Z ) =[6,10] in nonlinear optimization techniques to be used in minimizing
S2(x ) = [6, 10] in, S2(z ) =[0,4] in the deflection in the workpiece. Also, to ensure part stability,
S3(x ) = [0, 4] in, S3(z ) =[2,6] in and fixturing tool points and machining tool noninterference
geometric constraint regions for each support were developed.
SLFC's a12 = ~13 = 0.8
The tool points were designed to withstand about 20% of the
323 = 0.8,51[i = 1,2,3] = 1.0 maximum machining load possible in the three coordinate di-
rections. The examples demonstrated the robustness of the
procedure by considering cases of simultaneous drilling and
The effect of the drilling, milling, and clamping forces on milling loads that would occur in a practical situation. In
the workpiece about the support locations causes the work- cases tested the optimization reduced the workpiece deflec-
piece to deform as shown by the vector plot in Fig. 7. This tion by the order of 30% from the initial deflection. The al-
effect of deformation is reduced by finding the new locations gorithm was implemented by customizing Patran 1.4 to per-
for the supports by optimization. In this case, an optimal form fixture analysis and optimization. The initial fixture
solution is found that minimizes the objective function by design was performed by customizing Pro/ENGINEER. It is
30%. There are situations in solving this minimization prob- the belief of the authors that the developed work can aid
lem where multiple manufacturing loads are applied simul- in reducing lead time in design of fixtures ensuring higher
taneously. It has not been possible to minimize all the de- quality of manufactured parts.
flections although the objective function value has decreased.
Table 5 shows a case where the compromise solution has a
References
large objective function value. This problem could be re-
solved in two ways. Firstly, if automatic fixtures are used
Cai, W.; Hu; S.J.; Yuan, J.X. 1996: Deformable sheet metal fix-
and thus fixturing components should be able to move to dif-
turing: principles, algorithms, and simulations. J. Manufacturing
ferent configurations while machining. Secondly, perhaps ad- Sci. Engrg. 118, 318-324
ditional supports or supports with different geometries could
be used. The latter is an area for future research. DeMeter, E. 1996: The development of meta-models for the analy-
sis and design optimization of machining fixtures. Manufacturing
ReV.
6 Conclusions
Hockenberger, M.J.; DeMeter, E.C. 1995: Effect of machining
This paper described a method of analysing and optimizing fixture design parameters on workpiece displacement. Manufac-
a fixture design for minimal workpiece deflection during ma- turing Rev. 8, 22-32
chining. Given the nature of machining loads and the com- Kashyap, S. 1997: An integrated methodology for fixture design,
plexities on the geometry, both the finite element analysis and analysis, and optimization. Ph.D. Thesis, Rensselaer Polytechnic
optimization procedures are nonlinear; necessitating efficient Institute, Troy, NY
201

.C

.[
-.000000,

Fig. 7. Vector plot for deformation for simultaneous drilling and milling

Kashyap, S.; DeVries, W.R. 1997: Kinematic constraint Menassa, R.; DeVries, W. 1991: Optimization methods applied to
of workpiece during machining: A feature-based approach. selecting support positions in fixture design. J. Engrg. Ind. 113,
Trans.NAMRI/SME, pp. 141-146 412-418

Kashyap, S.; DeVfies, W.R. 1998: A concurrent engineering Press, W.H.; Teukolsky, S.A.; Vellerling, W.T.; Flannery, B.P.
framework for manufacturability evaluation, fixture design, anal- 1992: NUMERICAL RECIPES in C the art of scientific comput-
ysis, and optimization in virtual prototyping. Proc. 31-st CIRP ing, 2nd edition. Cambridge University Press
Conf. in Manufacturing Systems, pp. 517-522 Reklaitis, G.V.; Ravidran, A.; Ragsdell, K.M. 1983: Engineering
Lee, J.; Haynes, L. 1987: Finite-element analysis of flexible fixtur- optimization methods and applications. Wiley-Eastern Publica-
ing system. J. Engrg. Ind., Trans. A S M E 109, 134-139 tion
Vanderplaats, G.N. 1990: Numerical optimization techniques for
McNeM-Schwendler Corporation 1995: Patran FEM modeling,
engineering design: with applications. McGraw Hill Series in Me-
customization, and analysis users guide
chanical Engineering

Received August 2~,, 1998


Revised manuscript received December 14, 1998
Communicated by J. Sobieski

You might also like