You are on page 1of 11

International Journal of Machine Tools & Manufacture 41 (2001) 22332243

Simulation and analysis of assembly processes considering


compliant, non-ideal parts and tooling variations
Min Hu
a

a,*

, Zhongqin Lin a, Xinmin Lai a, Jun Ni

Institute of Mechanics and Engineering, Shanghai Jiaotong University, Shanghai, PR China


b
Department of Mechanical Engineering, The University of Michigan, Michigan, USA
Received 2 December 2000; accepted 2 April 2001

Abstract
Variations in parts and tooling are a major problem in automobile body assembly processes. Those
discrepancies adversely affect body assembly quality, functionality, cost and time-to-market. Variation
simulation analysis has been used in the design stage to predict such uncertainties, but variation analysis
based on rigid body assumptions usually yields over-estimates of the assembly variations. This paper
presents a new numerical simulation method for the assembly process incorporating compliant non-ideal
parts. This method considers the interaction and interference between compliant parts due to part variation,
assembly tooling variation, welding distortion, and spring back effects. A butt-to-butt joint assembly
example is used to illustrate the effects of various variation causes. A three-parts assembly example examines the assembly and welding sequences. Results from these two examples demonstrate that the proposed
method is theoretically sound and practically useful. 2001 Published by Elsevier Science Ltd.
Keywords: Variation; Spot welding; Sheet metal; Assembly

1. Introduction
With increased competition in the automobile market, more attention has been given to managing variations in automobile body assembly processes. Dimensional variation affects fit quality
and functionality. For example, variations in a body-in-white (BIW) can ultimately cause poor
sealing, undue effort required for door closing, water leaks, excessive wind noise, prolonged
time-to-market and added manufacturing costs. Typically, the automobile body assembly process
comprises numerous steps, utilizing 300500 compliant sheet metal parts, 50120 assembly sta* Corresponding author.
E-mail address: hm70702@mail1.stju.edu.cn (M. Hu).

0890-6955/01/$ - see front matter 2001 Published by Elsevier Science Ltd.


PII: S 0 8 9 0 - 6 9 5 5 ( 0 1 ) 0 0 0 4 4 - X

2234

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

tions and 30006000 spot welds. Each step in the process is capable of contributing a degree of
variation. Those variations in turn act on one another to compound distortion in the final BIW.
The complexity of this interaction places severe demands on the existing methods of simulation,
which currently fall far short of satisfaction.
Traditional variation analysis focuses on either the worst case, or the statistical analysis, or the
Monte Carlo simulation. The worst case analysis is primarily used for one-dimensional assembly
that assumes all parts take on their extreme values, but in fact the chances that all components
will take their extreme values simultaneously are remote. As a result, tight tolerances to meet asdesigned specifications are the norm. That strict standard inevitably inflates manufacturing costs
needlessly [1].
Statistical analysis specifies variations in parts as statistical distributions. Calculating design
function distribution is the critical task. If that is accurate, the method is superior to the worstcase approach in modeling the interchangeability of mass production and more realistic in estimating assembly tolerances [2].
Monte Carlo simulation samples points from the distribution being evaluated. The method selects points randomly and analyzes them to determine the design function values. Accuracy therefore depends on the number of samples and can require an extensive series of calculations. VSA
and VSM are two commonly used packages for predicting assembly variation [3].
All three of those methods are inadequate for automobile body assembly. They proceed from
the assumption that parts are rigid and never deformed during assembly. Yet, almost all BIW
parts are made from sheet metal, which is compliant and subject to deformation during placing,
clamping, locating and welding. When current simulations ignore this fact, they invite large-scale
inaccuracies. From his 1980 statistical study of the automotive body panel, Takezawa [4] concluded that the conventional addition theorem of variance is no longer valid for determining the
permissible limits for the auto-body assembly. The success of the 2 mm Program in the American automobile industry [5] gave further support to the principle that deformation of compliant
parts plays an important role in automobile body final assembly variation.
In Liu et al. [6], a cantilever beam model and mechanical variation simulation were used to
predict deformable-part assemblies by combining engineering structural models with statistical
analysis. Their research revealed important concepts and fundamental properties central to the
assembly of deformable parts. In the same year, Liu and Hu [7] proposed a concept the offset
beam element. This was a revision of the stiffness matrix that took into consideration force
behaviors at spot welding points. Subsequently, Liu et al. applied these two methods to the study
of variation propagating factors with such parameters as joint styles, part thickness, part and tool
variations. Their models confine spot welding to joined parts at beam-ends.
Later, Chang and Gossard [8] constructed a CAD model of compliant parts to predict variations
in the final product, but excluded from their models were the consequences of deformationinduced interaction and the effects of force. Moreover, most of their models were created using
a self-developed code.
Proposed in this paper is a general numerical method of simulating assembly processes involving compliant parts and dimensional variations. This method combines elastic and contact analyses
by using a popular FEM code ANSYS [9].
This paper is organized as follows. Section 2 examines an assembly process in details and
presents a simulation method based on contact and search analysis. Section 3 applies the method

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

2235

to a butt welding joining assembly process, obtains some results to better understand this method
and prove its suitability. Section 4 applies the proposed method to analyze a practical assembly
of compliant sheet metals and draws some conclusions.
2. Simulation method
2.1. Causes of variation
It is necessary when building a model to make clear root causes of variation in the auto body
assembly process. Any factor which can affect the manufacturing quality of part and tooling,
should be included. Figure 1 illustrates the causes in details.
2.2. Assembly and simulation cycles
To incorporate part compliance in the simulation of automobile body assembly processes, the
stiffness matrix of each part and each step is required. In a typical automobile body assembly
station, parts or subassemblies are assembled in the following four steps:
1. Positioning parts are placed on work-holding fixtures. Part variation is a natural phenomenon
in the sheet metal manufacturing and assembly process. Part variation {d} occurs in stamping,
transportation or prior subassemblies. Therefore, the initial matching gap between the parts is
an inevitable problem;
2. Clamping parts are clamped to the fixtures, which can usually be assumed to be rigid relative
to the compliance of the part. The initial gaps at the clamping positions will be overcome by
clamping forces. Variations in clamps and locators can also cause parts to deform and interact.
It can be described as {Fu}=[Ku]{du}, in which {du} is the closed gap vector at the clamping
positions, {Fu} is the clamping force vector and [Ku] is the stiffness matrix;
3. Welding two parts are joined with point-to-point connection at each welding spot. Deformation occurring as the gap between the parts is closed. Welding guns often add further defor-

Fig. 1.

Root causes of assembly variation.

2236

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

mation to the parts by pulling or pushing them. Frequently, the actual welding spot strays from
the designed spot;
4. Releasing clamps are released and the assembled part springs back to minimize the strain
energy stored during the preceding operations. Subassembly contact status varies. Spring-back
deformation is calculated by removing displacement boundaries at clamping points to simulate
clamp release. It can be calculated by the equation: {Fw}=[Kw]{dw}, where {dw} is the springback deformation vector, {Fw} is equal to the negative reaction forces at the clamping positions
and [Kw] is the stiffness matrix of the welding assembly structure.
Moreover, the compliance of sheet metal makes it sensitive to exterior loading (from clamping
and welding) and interior loading (such as from residual stresses), both with the potential to
produce deformation.

Fig. 2.

Flowchart of the NVAM simulation based on contact and search analysis.

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

2237

2.3. Simulation method


The four steps described above constitute an elementary assembly cycle called PCWR. A new
simulation procedure for automobile body assembly variation analysis is shown in Fig. 2, which
is named NVAM (new variation analysis method). It combines contact and interaction analysis
between parts an approach more in line with the actual PCWR process and more readily
understood. Thus, the entire automobile body assembly process can be formulated as mPCnWR,
in which m equals the total number of assembly cycles and n is the number of welding points in
one PCWR. For simplicity, the process is assumed to be friction free and linear.
In addition to the part variations, clamping processes often introduce variation due to variations
in clamps and locators. Thus, displacement boundaries can be applied to simulate the variations
in clamping and locating points. The deformation due to clamping and weld squeezing can be
calculated in the displacement-based finite element formulation: {F}=[K]{d}. The minimum force
is dependent on the material properties and the dimensions of the parts.
For welding assembly with an initial gap and variations, it is very important to understand the
contact and interaction between parts under clamping and squeezing forces. Under the effect of
forces, gaps at the positions with forces working on are closed, geometries of parts change and
contact will occur between parts. The contact will have a large effect on deformation in the further
assembly. There will be an interactive force between parts.
In order to evaluate assembly design and quality, a set of standards must first be established.
Usually, there are some key characteristics on parts which could significantly affect the target
value of controlled variation, the performance of part function and customer satisfaction. Therefore, the CPs (critical points or surfaces) related to those key characteristics are defined as controlled parameters to ensure manufacturing and assembly. In fact, the determination of CP should
take the geometry of parts, the practical assembly process, the performance of the part or subassembly function and the variation requirement into consideration.
According to the flowchart described in Fig. 2, a FEM model can be generated with ANSYS5.5.
The model was then map-meshed with structural solid elements and surface contact elements,
as in the following examples. Note the fine mesh at the matching interfaces, where parts will be
jointed into a whole.
3. Verification example butt joint welding
In order to verify the proposed simulation method, a simple example is selected for which an
analytical solution is available. We will compare the results obtained from the proposed numerical
simulation with the analytical solution.

Fig. 3. An illustration of butt joint assembly example.

2238

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

Butt joints are widely used to weld two parts in automobile body assembly. In a butt-to-butt
joint, the flanges of two or more parts are welded together to form an assembly. An assembly
with a butt joint is shown in Fig. 3, and is used as an example to analyze the final assembly
variation. In order to simplify, we assume only part variation exists. Then, the butt joint assembly
process is as shown in Fig. 4.

Fig. 4. The process of two parts butt joint assembly with part variation du. (a) Part variation du from nominal position;
(b) clamping and welding gun closing; (c) butt joint welding; (d) releasing clamps; (e) dimensions of assembly structure.

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

2239

Fig. 5. Meshes and constraints in FEM analysis of butt joint example.

Both parts shown in Fig. 4 are made of mild sheet metal with Youngs modulus
E=207,000 N/mm2 and Poissons ratio n=0.3. The thickness of both parts is 1.2 mm. Assume
du=0.5 mm, L=100 mm, c=10 mm and b=8.533 mm.
3.1. Simulation in FEM
According to ANSYS code and NVAM proposed in Section 2.3, the finite element mesh,
including contact analysis for the assembly model, is shown in Fig. 5 with 94 elements and 120
nodes, including 18 pairs of contact elements at the matching surface with the butt joint. Each
welding spot is simulated in the style of coupled nodes. The small triangles in Fig. 5 represent
the constraints provided by fixtures.
3.2. Analytical solution
According to Hookes law, the theoretic variation of the above assembly structure can be calculated using Eq. (1):
uy

3a2
d
(4a+b)b u

(1)

3.3. Comparison
The same parameters are used in simulative and analytical solutions. A comparison of results
is summarized in Table 1. The simulation results are in close agreement with the results calculated
by the analytical solution using Eq. (1).
It is clear from Table 1 that the NVAM that integrates contact, search and couple analysis is
sufficiently accurate. In the following section, this method will be applied to examine a practical assembly.
Table 1
Comparison of results calculated by analytical solution and simulation method
Variation
ux
uy

Theoretical result

Simulation result
0.5
0.83168

Error
0.497027 0.59%
0.834663 0.36%

2240

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

Fig. 6. Show of parts and constraints.

4. Application example: a dash panel cover assembly problem


In this section, the NVAM proposed simulation method will be utilized to investigate a practical
assembly, which is the assembly of dash panel cover and dash panel as shown in Fig. 6. Figure
6 also shows the finite element model for the subassembly with constraints from locators, clamps
and welds connecting the instrument panel with the front-end body. As the original status, there
is a gap between two matching surfaces shown in Fig. 7, which will be assembled by butt-joint

Fig. 7. The original status.

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

2241

Fig. 8. Welding points.

welds. These two panels are welded together at 26 points (see Fig. 8) after the panels are located
and clamped. Excessive deformation in the subassembly of these two panels affects the assembly
quality of the front window glass and instrument panel. The deformation is calculated by simulation and measured by experiment at the critical points as shown in Fig. 9. Figure 10 shows the
constructed fixture and measuring method in the experiment.
According to the symmetric characteristic of parts, two welding gun is considered. Thus, there
are three commonly used welding sequences. They are, respectively: the current adopted sequence,
from the middle to outwards, and from the outwards to middle, which can be referred to as
sequences 1, 2 and 3 as the following.
1. Sequence 1: 125357911131517192123 and 226468101214161820
2224;
2. Sequence 2: 2624222018161412108642 and 252321191715131197
531; and
3. Sequence 3: 135791113151719212325 and 246810121416182022
2426.
Then, two parts are assembled in terms of different assembly sequences. The results CP1, CP2
from simulation and MP1, MP2 from experiment are shown in Fig. 11. In fact, it is very difficult

2242

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

Fig. 9. Critical points.

Fig. 10.

Experiment for assembly analysis.

to make the measurement point and critical point fully coincident. According to Fig. 11(a)(d),
the conclusion can be drawn that the results from the NVAM simulation and experiment are fairly
close and the NVAM proposed above is very useful.
5. Conclusions
Outlined in this paper is the NVAM method and procedure for simulating and analyzing automobile body assembly variations, considering both various tooling variations and compliant, nonideal parts. Founded on contact, search and coupled node analysis, the method is a powerful tool
for comprehending and calculating the final assembly variation. Through a simple butt joint
example, the simulation is tested and verified against the theoretical solutions. From there, it has
been extended to variation simulation of more complex practical compliant sheet metal assemblies.

M. Hu et al. / International Journal of Machine Tools & Manufacture 41 (2001) 22332243

2243

Fig. 11. Comparisons of assembly deformation in different sequences. (a) Assembly in three different sequences; (b)
assembly in sequence 1; (c) assembly in sequence b; (d) assembly in sequence 3.

References
[1] E.I. Fortini, Dimensioning for Interchangeable Manufacturing, Industrial Press, New York, 1967.
[2] C. Uroy, R. Liu, T.C. Woo, Review of dimensioning and tolerancing: representation and processing, ComputerAided Design 23 (7) (1991) 466483.
[3] V.J. Skowronski, J.U. Turner, Using Monte-Carlo variance reduction in statistical tolerance synthesis, ComputerAided Design 29 (1) (1997) 6369.
[4] N. Takezawa, An improved method for establishing the process wise quality standard, Reports of Statistical and
Applied Research, JUSE 27 (3) (1980) 6376.
[5] W. Cai, S.J. Hu, J. Yuan, Deformable sheet metal fixturing: principles, algorithms and simulation, Transaction of
the ASME, Journal of the Manufacturing Science and Engineering 118 (3) (1996) 318324.
[6] S.C. Liu, H.W. Lee, S.J. Hu, Variation simulation for deformable sheet metal assemblies using mechanistic models,
Transactions of NAMRI/SAE 23 (5) (1995) 235240.
[7] S.C. Liu, S.J. Hu, An offset finite element model and its applications in predicting sheet metal assembly variation,
International Journal of Machine Tools and Manufacturing 35 (11) (1995) 15451557.
[8] M.H. Chang, D.C. Gossard, Modeling the assembly of compliant, non-ideal parts, Computer-Aided Design 29 (10)
(1997) 701708.
[9] ANSYS, Users Manual, ANSYS Inc, USA, 1994.

You might also like