Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 43 (2016) 58 – 63

14th CIRP Conference on Computer Aided Tolerancing (CAT)

Assembly Error Calculation with Consideration of Part Deformation

Liu Tingb,Cao Yanlonga,b*,Wang Jingb ,Yang Jiangxinb
a. State Key Laboratory of Fluid Power and Mechatronic Systems, college of Mechanical Engineering, Zhejiang University, Hangzhou, 310027, China
b. Key Laboratory of Advanced Manufacturing Technology of Zhejiang Province, college of Mechanical Engineering, Zhejiang University, Hangzhou , 310027,
China˗
* Corresponding author. Tel.:+86-571-87953198; fax: +86-571-87951145. E-mail address: sdcaoyl@zju.edu.cn.

Abstract

Traditional assembly do not consider mechanical deformation based on the rigid body hypothesis, while assembly precision has close relationship
with part’ deformation in practice. In this paper, temperature, gravity and working load are taken into account to calculate the deflection occurred
in assemblies. Both manufacturing error and deformation error are taken into consideration to establish deviation calculation model based on
Jacobian-torsor model. A tailstock is taken as an example to verify the feasibility of the proposed method.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing
Keywords: deformation; temperature; gravity; working load; assembly error

1. Introduction to its assembly sequence. Besides, the Jacobian [5], the T-

A mechanical assembly consists of two or more
components or subassemblies. Owing to variation in
manufacturing, it is impossible to completely avoid
variations in component dimensional and geometrical
characteristics. Also mechanical products are easily affected
by the working environment in the assembly process. Load,
temperature and gravity will bring about part’ deformation,
increasing the product assembly errors and affecting the
assembly quality. An effective and rapid assembly error
calculation method taking both part machining errors and
deformation errors into consideration will provide a
significant guide for reasonable and economical accuracy in
design.
Many different approaches for calculating assembly
errors have been developed over the years. Bensheng Xu et
al. [1] choose to use the relationship between mechanical
finger’s differential motion and joint’s small displacements
to establish the model of tolerance accumulation; Tang and
Davies [2] present a complete matrix tree chain method for
determining the accumulated tolerances and the working
dimensions. Paul et al. [3] estimate the tolerance by the
calculation of the tolerance sensitivity of critical assembly
features with respect to each source of dimensional variation
in the assembly. C. LU et al. [4] propose a concept called
sensitive tolerance to evaluate the assemblability according
Map model and the deviation domain model [6-7], the direct
linearization method (DLM) [8], the torsor model [9], the
matrix model [10], the vector loop model [11] and the unified
Jacobian-torsor model [12] have been presented successively.
These mentioned tolerance calculation processes assume that
the manufacture parts are rigid without consideration of
deformation in assembly.
In actual working condition, temperature, load and part
gravity may result in changes in dimension, angle and shape,
then influencing the matched space position and final system
reliability. So there is a great need to consider the
deformation in design stage.
So far, several researchers have discussed the tolerance
analysis integrated the part elastic deformation and
displacement. S. Samper et al. [13] present four models in
that allow elastic deformations of mechanisms in tolerancing.
Pierre et al. [14] have integrated thermos-mechanical strains
into tolerance analysis, where they have used finite element
method to determine the strains. G. Jayaprakash [15]
proposes an optimal tolerance design method for mechanical
assembly considering thermal impact, and further takes both
thermal and inertia impact into account. NSGA II and finite
element are used to obtain proper component tolerance
values [16]. Benichou et al consider thermal expansion of
parts integrated within functional tolerancing [17]. As
demonstrated in [18-19] a wider set of operating factors can

2212-8271 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing
doi:10.1016/j.procir.2016.02.007
 Liu Ting et al. / Procedia CIRP 43 (2016) 58 – 63 59

be included in computer-aided tolerance analysis such as

thermal, wear and pressure. Rotating machinery in boring
processing, the static deformation of the spindle shaft is
calculated by finite element method [20]. Tolerance
allocation in assembly with time-variant deviation induced
by mobility and wear is performed to optimize the tolerance
specification [21]. FE simulation as a virtual tool is used to
calculate actual deformations in a designed mechanical part
due to all of its service loads [22]. Weiming Zhang et al
extends and modifies the Jacobian-torsor tolerance model in
actual working condition, quantitatively expresses the
(a)
impact of actual working condition on working performance
and tolerance design [23]. The methods proposed above
either only suitable for deformation in special conditions
without commonality or concerns no relation to accumulated
assembly error, Jianyong Liu [24] puts forward a
computational method for assembly error with consideration
of parts deformation. Nevertheless, analysing dimensional
accuracy in deformation situation only, invalid for the
relevant location and form error.
This paper presents a new method to establish deviation
propagation and calculation model considering
manufacturing error and deformation error simultaneously.
(b)
Deformation error results from the effects of temperature, Fig.1 Heat deformation nephogram. (a) Overall deformation; (b) Axial
gravity and working load on assembly part. Jacobian–torsor deformation.
model is adopted to obtain comprehensive error.
The organization of the article is as follows. The factors
that have an impact on part deformation are discussed in
Section 2. Section 3 proposes the theory and procedures for
assembly error calculation in Jacobian–torsor model. A
three-part assembly is taken as an example to illustrate the
calculative process.

2. Influence factors of part deformation

Fig.2 The actual parts location under the action of gravity

2.1. Temperature 2.3. Load

Most materials change length as temperature is changed. During a practical working situation, the force field
As a result of this change, the dimensions and tolerances of applied to single parts or assembly results in geometric errors
a product become at variance with the design values. and dimensional changes in size, direction and magnitude in
Therefore, thermal impact must be taken into effect during spatial six degrees of freedom. Taking a bourdon tube as
the design process, particularly when a complicated product example, it bears bending load when it works, and the whole
with multiple components and various materials operates bourdon tube tilts to the side of the load. The larger the
under a wide range of temperature. As shown in Fig.1, a bending load, the bigger the bending deflection. Concrete
hollow cylinder in working condition produces deformation displacement can be obtained by finite element analysis.
because of the induced heat.
2.2. Gravity
The gravity leads to the deformation of the component, as
shown in Fig.2. In an ideal 3-dimension model, part locates
in desired position, the centre of the shaft is generally
coincide with hole centre in a hole shaft assembly. However,
in actual situation, part will shift from its axis under the
influence of gravity, offset distance is associated with the
diameter of the hole and shaft, and their geometric tolerance.
The amount of deformation is proportional to the mass of (a) (b)
parts, and can be calculated using finite element analysis.
Then it is incorporated in the tolerance stack up progress,
improving the precision of the error estimation.
60 Liu Ting et al. / Procedia CIRP 43 (2016) 58 – 63

tolerance transport chain.

u v w D E G : lower limits of the tolerance intervals in
six freedom.
u v w D E G : upper limits of the tolerance intervals in
six freedom.
The Jacobian matrix is formulated by:
ª ª Ri º
« ¬ 0 ¼ 3u3
... ª¬Wi n º¼  ª Ri º . ª Ri º º
3u3 ¬ 0 ¼ 3u3 ¬ PTi ¼ 3u3 »

[ J ]FEi [ J ]i0 «    »
(c) « »
Fig.3 Bourdon tube bears bending load. (a) Bourdon tube model; (b) Load « > 0@3u3  i i
ª¬ R0 º¼ . ª¬ RPTi º¼ »
diagram; (c) Strain contours. ¬ 3u3 3u3 ¼ 6u6
(3)
3. Assembly error calculation considering where ª R0i º = >C1i C2i C3i @ is the local orientation changes
¬ ¼
deformation based on Jacobian-torsor of frame i with respect to original frame 0.
> RPTi @ >C1C2C3 @ , C1C2C3 represents the direction
PTi
3.1. The Jacobian–torsor model
change of tolerance with respect to the three coordinate axes
In the assembly process, the final product quality is of frame i.
determined by stack up, coupling and propagation of all ª 0  dzin dyin º
errors. The unified Jacobian-Torsor model introduced by « » is a skew-symmetric
ª¬Wi n º¼ « dzi
n
0  dxin »
Desrochers et al [25] is an innovative tolerance analysis «  dyin dxin 0 »¼
¬
method which uses the small displacement torsor (SDT) [26]
matrix, indicates the position changes on frame n with
for tolerance representation and the Jacobian [27] for
tolerance propagation, The unified Jacobian-torsor model is respect to frame i, in which dxin dxn  dxi ;
built that combine the advantages of these two methods. dyin dyn  dyi ; dzi
n
dzn  dzi . The final [FR] can be
The small displacement torsor can directly represent
calculated based on above equations.
potential variations along and about all three Cartesian axes
in its generic form. Jacobian matrix has been proposed to 3.2. Assembly error calculation taking into account part
map all SDT from their local origin to the point of interest. It deformation
is a matrix whose columns are extracted from the various
The factors referred in Section 2 may changes the
homogeneous transform matrices relating the reference
dimension, shape or position of parts, these errors will be
frames of the functional elements (FE) to that of the
added up and magnified in the assembly process.
functional requirement (FR). Therefore, Jacobian–torsor
Deformation in parts is shown as a translational vectors in
model can be expressed as follows:
the coordinate system and rotational vectors rounding the
[FR] = [J][FE] (1)
coordinate system, covered as follows:
where the functional requirement (play, gap, clearance)
represented by [FR] is the translational and rotational vectors S > 'u 'v 'w 'D 'E 'J
T
@ (4)
of general assembly. [FE] is related to the functional
where 'u , 'v and 'w are respectively three translational
elements for assembly parts mathematically. [J] is a
Jacobian matrix expressing a geometrical relation between a vectors of the origin of the coordinate system on axes x, y,
[FR] vector and corresponding [FE] vector; and z. Likewise, 'D , 'E and 'J are rotational vectors around
Spreading eq. (1) into: the axes.
ª ª[u u] º º The specific value in vector S can be obtained with the
«« »
« «[ v v ] »» » combination of ANSYS and MATLAB. Finite element is
« «[ w w] » » used to get the coordinate information for discrete points in
«« » » characteristics surfaces while MATLAB can associate
« «[D D ]» »
ª[u u] º « «[ E E ]» » corresponding surfaces and extract real derived element, to
«[ v v ] »» «« » »
« « «¬[G G ] »¼ FE1 » gain its position and direction information. The process for
«[ w w] » « »
« » [[ J1 J 2 J 3 J 4 J 5 J 6 ]FE1...[ J1 J 2 J 3 J 4 J 5 J 6 ]FEn ]. « ... » obtaining integral deformation information is shown in Fig.
«[D D ]» « » 4.
«[ E E ]» « ª[u u] º »
« » « «[ v v ] »» »
«¬[G G ] »¼ «« »
« «[ w w] » »
« «[D D ]»
» »
«« »
« «[ E E ]» »
«« » »
«¬ «¬[G G ] »¼ FEn ¼»
(2)
where [‫ܬ‬ଵ ‫ܬ‬ଶ ‫ܬ‬ଷ ‫ܬ‬ଵସ ‫ܬ‬ହ ‫] ଺ܬ‬୊୉୧ is the Jacobian matrix which
describes the relationship between vector [FR] and [FE]. i
represents the number of functional elements on the
 Liu Ting et al. / Procedia CIRP 43 (2016) 58 – 63 61

Three-dimensional
assemblage diagram
dxin' ' dxn '  dxi '  dx n
 'un    dxi  'ui 
FE dx   'un  'ui 
n
i

Discrete feature surface dyin' ' dyin   'un  'ui  ; dzi '
n'
dzin   'wn  'wi 
Temperature
into series of points Ideal Combine Eq. (4-7), we can get:
conditions
Gravity Deformation
Impact factors
ª ª Ri º ª Ri º
« ¬ 0 ¼ 3u3 ¬ PTi ¼ 3u3 3u3

... ¬ªWi n' ' ¼º ¬ª R0i ¼º . ¬ª RPT
3u3
i
º
i ¼ 3u3 »º
i'
[ J ]FEi ' [ J ] «    » (8)
Discreted points 0
« »
Load diplacement information « >0@3u3  ª¬ R0i º¼ . ª¬ RPT
i
º
i ¼ 3u3
»
¬ 3u3 ¼ 6u 6
*36
Real derived feature Besides, [FE] can rewrite from equation following.
association > FE @FEi '
1
> RPTi ' @ >T @ (9)
In consequence, [‫]ܴܨ‬ᇱ
affected by multiple fields is
Real derived feature Ideal derived feature accessible by uniting Eqs.8-9. Assembly error calculation
displacement information diplacement information
process after deformation is shown in Fig. 5.
3DUWV
GHIRUPDWLRQ
Translational Rotational
vector vector

Fig.4 Steps for deformation information obtainment

7UDQVODWRU9HFWRU 5RWDWLRQDO9HFWRU
Step1: Calculate ideal derived feature displacement > 'u 'v 'w@ > 'D 'E 'J @
information including position and direction vector.
Step2: Importing three-dimensional assemblage diagram
into the finite element analysis software, by meshing, adding
constraint and imposing temperature, gravity and load, then
the deformation for each node is obtained.
Step3: Using GPS theory to gained nodes, associating [ J ]FEi ' > FE @FEi '
deprived feature.
Step4: Comparing the vectors acquired in Step1 and Step3 to
get translational vector > 'u 'v 'w@ and rotational vector Fig. 5 Assembly error calculation
> 'D 'E 'J @ . Parameters in Fig. 5 can be gotten by Eqs.5-9, then
Part deformation brings about small change in coordinate Jacobian-torsor is modified for deformed assembly
system which can be calculated in above four steps. In accumulated error calculation. For contrast purposes, the
consequence, functional element FE varies. Variables in Eq. error obtained considering parts deformation is compared to
(3) will be updated as: the case not influenced by multiple fields. Therefore
R0i ' R0i >C xi @ ¬ªC yi ¼º >C zi @ (5) deformation influence can be clearly observed.
1 1 1
RPTi >C xi @ (6) 4. Case study
RPTi ' ¬ªC yi ¼º >C zi @
ª 0  dzin' ' dyin' ' º A simple assembly consist of three parts is taken as an
« » (7)
ª¬Wi 'n ' º¼ n'
« dzi ' 0  dxin' ' » example in this case study. Its shape, dimension and
«  dyin' ' dxi 'n'
0 »¼ tolerance are shown in Figs. 6 and 7. The range of fluctuation
¬
for apex in part 3 is the key issue to ensure the position
where >C xi @ , ªC yi º and >C zi @ are transition matrix which precision because it is always used to localization-clamp.
¬ ¼
rotates 'D , 'E and 'J respectively around the
corresponding axis on frame i. They can be computed as
following.
ª1 0 0 º
[C xi ] «0 cos 'D  sin 'D »»
«
«¬ 0 sin 'D cos 'D »¼
ª cos 'E 0 sin 'E º
[C yi ] « 0 1 0 »»
«
¬«  sin 'E 0 cos 'E ¼» Fig.6 Assembly model
ª cos 'J  sin 'J 0º
[C zi ] « sin 'J cos 'J 0 »»
«
«¬ 0 0 1 »¼
As a result, ݀‫ݔ‬௜௡ , ݀‫ݕ‬௜௡ ܽ݊݀ ݀‫ݖ‬௜௡ update into:
62 Liu Ting et al. / Procedia CIRP 43 (2016) 58 – 63

300

250

200

z/mm
150

100

50

0
-100

-50

0
-100
50 -50
0
50
100
100 150
Fig.7 2-dimension model for assembly y/mm
200
x/mm

(a)

260

240

220

z/mm
200

180

160
50

Fig.8 Connection graph 0

0
-20
-40

Based on the connected relations among the assembly, y/mm -50 -80
-60
x/mm

the connection graph is described in Fig.8. Then Jacobian- (b)

torsor equation of the assembly without consideration of Fig. 10 Contour fitting of the deformation section.
deformation can be established as follows:
FR=FR1+FR2=[[J] 1[J] 2[J] 3[J] 4][[FE] 1 [FE] 2 [FE] 3 [FE] 4] As shown above, Fig.9 (a) and (b) indicates the principal
deformation portion, that is, respectively part3 and its
ª-0.4000 0.4000 º matching part2. Fig. 9 displays contour fitting of the
«-0.3522 0.3522 »»
« deformation section. Real derived feature association results
show that the direction vector of Fig.10 (a) and (b) are
= «-1.6102 1.6102 »
« » separately (1,-2.9506 × 10ିସ , 1.2681 × 10ିଷ ) and (1,-
«-0.0020 0.0020 » 0.0013, 8.1162 × 10ିଷ ) compared to original vector (1, 0, 0)
«-0.0079 0.0079 » and (1, 0, 0).
« » According to Section 3.2, parts deformation results in the
¬-0.0054 0.0054 ¼
changes of Jacobian matrix [J] and torsor matrix [FE],
FR is the final output errors (functional requirements) of FR ' > J @FEi ' > FE @FEi '
the error propagation chains.
In view of part deformation, under the multiple influence ª-0.3991 0.3991 º
of temperature field, gravity field and load field, mainly «-0.3531 0.3531 »
« »
distorted parts are part2 and part3. GPS theory is applied to
= «-1.6111 1.6111 »
gain real derived feature, then variation information is « »
approached to modify Eq. (1). «-0.0021 0.0021 »
«-0.0079 0.0079 »
« »
¬-0.0054 0.0054 ¼
Comparing the result that consider deformation with the
value of FR which ignores deformation, assembly error range
increases in the y-direction and z-direction while value in x-
direction decreases, that is, directly affects the approachable
distribution of apex in final part. Therefore, localization-
clamp precision can’t be ensured.

(a) (b) 5. Conclusions

Fig.9 Principal deformed portion.
This study proposes an accumulated error calculation
method for assembly which considers the deformation
resulting from the temperature, self-weight and load in actual
working condition. Jacobian-torsor model is used to compute
the integrating error in the whole error propagation chains for
assembly. Case result proves that part deformation has make
difference to assembly error calculation. Therefore, taking
deformation error into consideration in design stage can
optimize product performance in advance.
 Liu Ting et al. / Procedia CIRP 43 (2016) 58 – 63
This article only obtains the calculation error results in
extreme dimension, unapproachable to get specific error
value. Influence factors of deformation are not limited to the
referred three. The further exploration will be studied in
future research.
Acknowledgments
This research was supported by the National Nature
Science Foundation of China (No. 51275464 and 51575484)
and Science Fund for Creative Research Groups of National
Natural Science Foundation of China (No.51221004).
6. Reference
[1] Bensheng Xu, Meifa Huang, Yanru Zhong. Research on robotic
kinematics-based modelling of accumulation of three-dimensional
tolerance [J]. Machinary Design & Manufacture, 2007, 01:105-107.
[2] Xiaoqing Tang, B. J. DAVIES. Computer aided dimensional
planning [J]. International Journal of Production Research, 1988,
26:283-297.
[3] Paul J. Faerber. Tolerance analysis of assemblies using
kinematically derived sensitivities [D]. Department of Mechanical
Engineering Brigham Young University.1999.
[4] C. Lu, J. Y. H. Fuh, Y. S. Wong. Evaluation of product
assemblability in different assembly sequences using the
tolerancing approach [J]. International Journal of Production
Research, 2006, 44(23):5037-5063.
[5] Laperrière L, EIMaraghy HA. Tolerance analysis and synthesis
using Jacobian transforms. CIRP Ann—Manuf Technol 2000;
49(1):359–62.
[6] Davidson J K, Shah J J. A new mathematical model for
geometric tolerances as applied to round faces [J]. Journal of
Mechanical Design, 2002, 124(4):609-622.
[7] Giordano, M., Samper, S., and Petit, J.-P., 2007, “Tolerance
Analysis and Synthesis by Means of Deviation Domains, Axi-
Symmetric Cases,” Models for Computer Aided Tolerancing in
Design and Manufacturing, Springer, Netherlands, pp. 85–94.
[8] Chase K W, Gao J, Magleby S P, et al. Including Geometric
Feature Variations in Tolerance Analysis of Mechanical
Assemblies [J]. Iie Transactions, 1996, 28(10):795--807.
[9] Desrochers, A. Modeling three dimensional tolerance zones
using screw parameters. In: Proceedings of the DETC 25th Design
Automation Conference, 12–15 September, 1999, Las Vegas,
Nevada, USA.
[10] Desrochers A, Rivière A. A matrix approach to the
representation of tolerance zones and clearances [J]. International
Journal of Advanced Manufacturing Technology, 1997, 13(9):630-
636.
[11] Jingsong Gao, Kenneth W. Chase, Spencer P. Magleby.
Generalized 3-D tolerance analysis of mechanical assemblies with
63
small kinematic adjustments [J]. Iie Transactions, 1998, 30(4):367-
377.
[12] Ghie W. Statistical tolerance analysis using the unified
Jacobian–Torsor model [J]. International Journal of Production
Research, 2010, 48(15):4609-4630.
[13] Samper S, Giordano M. Taking into account elastic
displacements in 3d tolerancing: Models and application [J].
Journal of Materials Processing Technology, 1998, 78(1): 156-162.
[14] Pierre L, Teissandier D, Nadeau J P. Integration of
thermomechanical strains into tolerancing analysis [J]. International
Journal on Interactive Design & Manufacturing, 2009, 3(4):247-
263.
[15] Jayaprakash G, Thilak M, Sivakumar K. Optimal tolerance
design for mechanical assembly considering thermal impact [J].
International Journal of Advanced Manufacturing Technology,
2014, 73(5-8):859-873.
[16] Govindarajalu J, Karuppan S, Manoharan T. Tolerance design
of mechanical assembly using NSGA II and finite element
analysis[J]. Journal of Mechanical Science & Technology, 2012,
26(10):3261-3268.
[17] Benichou, S. and B. Anselmetti. 2011. Thermal dilatation in
functional tolerancing. Mechanism and Machine Theory, p.1575–
1587.
[18] Baker M and Steinrock G. Dimensional and variation analysis.
In: ADCATS conference, Provo, UT, 15–16 June 2000.
[19] Armillotta A, Semeraro Q. Critical operating conditions for
assemblies with parameter-dependent dimensions [J]. Proceedings
of the Institution of Mechanical Engineers Part B Journal of
Engineering Manufacture, 2013, 227(5):735-744.
[20] Guo J, Hong J, Yang Z, et al. A Tolerance Analysis Method
for Rotating Machinery [J]. Procedia Cirp, 2013, 10:77-83.
[21] Walter M S J, Spruegel T C, Wartzack S. Least Cost Tolerance
Allocation for Systems with Time-variant Deviations [J]. Procedia
Cirp, 2015, 27:1-9.
[22] Barari A. Tolerance Allocation Based on the Minimum
Deformation Zone of Finite Element Structural Frame Analysis [J].
Computer-Aided Design, 2013, 10(4):629-641.
[23] Weimin Zhang, Can Chen, et al. Tolerance modeling in actual
working condition based on Jacobian-Torsor theory [J]. Computer
[24] Jianyong liu. Calculation method for assembly error with
consideration of part deformation [J]. Computer integrated
manufacturing system, 2015, 21(1):94-100.
[25] Desrochers A, Laperrière L, Ghie W. Application of a Unified
Jacobian—Torsor Model for Tolerance Analysis [J]. Journal of
Computing & Information Science in Engineering, 2003, 3(1):2-14.
[26] Bourdet P, Clement A. Controlling a complex surface with a 3
axis Measuring Machine. Annals of the CIRP, Vol. 25,
Manufacturing Technology, pp. 359-361, 1976, Janv.
[27] Laperrière L, Lafond P. Modeling tolerance and dispersions of
mechanical assemblies using virtual joints. In: CD-ROM
proceedings of ASME 25th design automation conference. Fukuoka,
Japan, 1999, pp 330–336