Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 27 (2015) 241 – 246

13th CIRP conference on Computer Aided Tolerancing

Assembly model representation for variation analysis

N. Caia, L. Qiaoa*, N. Anwerb
a
Advanced Manufacturing Technology and Systems Research Center, Department of Industrial and Manufacturing Systems Engineering, Beihang University, 37,
Xueyuan Road, Haidian District, Beijing, 100191, China
b
LURPA, ENS Cachan, 61, Avenue du Président Wilson, Cachan Cedex, 94235, France
* Corresponding author. Tel.: +86-010-82317708; fax: +86-010-82317735. E-mail address: lhqiao@buaa.edu.cn

Abstract

Assembly modeling is a very important activity for computerized analysis of product assembly, and it has been a subject of intense research
over the past years. However, most of the assembly models focus on the assembly connective relationships between parts rather than assembly
spatial relationships. Variation analysis is becoming a critical issue when specifying and verifying the geometric and dimensional requirements
of assembly. In order to support dimensional variation analysis, a mathematical assembly model representation considering assembly spatial
relationships is proposed in this paper. Four types of coordinate systems are defined. The nominal and variational spatial relationships among
assembly elements are represented as homogeneous transformation matrix among different coordinate systems. The variations of assembly
elements, mating interface and locating interface are represented as variation transformation matrix among corresponding coordinate systems.
The variation of the final assembly is modeled as stack up of the differential motion vectors of coordinate systems of all the parts with respect
to global coordinate system, and it can be calculated by homogeneous transformation and variation transformation. The different situations for
two types of assemblies are distinguished in this novel assembly model representation.
© 2015
© 2015 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier B.V
B.V.This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of 13th CIRP conference on Computer Aided Tolerancing.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of 13th CIRP conference on Computer Aided Tolerancing
Keywords: Type your keywords here, separated by semicolons ;

1. Introduction Variation analysis is quite critical for dimensional control

Assembly is a process of putting a set of components and
parts together to form the final product. It is of great
importance in the life cycle of a product because it is
intensively related to the development time, cost and
dimensional quality of the final product. The design and
analysis of assembly is more and more supported by IT
technologies and computers. Assembly models and
representations should be therefore developed. Assembly
modeling has been a subject of intense research over the past
years. Many researches mainly focus on the assembly
connective relationships between parts, such as the liaison
graph by Bourjault [1], the non-directional blocking graph by
Wilson [2], the relational model by Homem de Mello [3], the
connector-based relational model by Yin [4] and the multi-
level assembly model by Chen [5], the k-piece graph by Wang
[6], the constraint transformation-based assembly model by
Jiang [7] and the multi-level and distributed assembly model
by Gao [8].
and quality improvement, and it has also been the subject of
considerable amount of research [9]. Mantripragada and
Whitney [10] proposed a state transition model to propagate
and control variation in mechanical assemblies. For more
generality and applicability, Huang et al. [11, 12] built a 3D
single-station variation model and a 3D multi-station variation
model based on Stream-Of-Variation Analysis (SOVA).
Cheng et al. [13] presented an analysis method for variation
propagation of Aeronautical Thin-Walled Structures (ATWS)
with automated riveting by dividing the whole assembly
process into two stages and eight states. Franciosa et al. [14]
presented a FEM-based computer tool for statistically
analyzing variations occurring in assembly processes of
flexible parts.
In order to support dimensional variation analysis of
assemblies, assembly spatial relationships, besides assembly
connective relationships should also be considered in
assembly model. Therefore, a mathematical assembly model
representation based on Coordinate Systems (CSs) definition

2212-8271 © 2015 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 13th CIRP conference on Computer Aided Tolerancing
doi:10.1016/j.procir.2015.04.072
242 N. Cai et al. / Procedia CIRP 27 (2015) 241 – 246
and Homogeneous Transformation Matrix (HTM) is proposed
in this paper. Four types of CSs are defined to locate parts,
features and fixtures. The nominal and variational spatial
relationships among assembly elements are represented by the
relationships among different CSs. The different situations for
type-1 assemblies and type-2 assemblies are also considered
in the assembly model representation. Through the proposed
assembly model representation, variation analysis can be
realized by homogeneous transformation and variation
transformation.
The remainder of this paper is organized as follows:
Section 2 presents the mathematical representation of
assembly geometry. In section 3, the explicit assembly model
representation for type-1 assemblies is developed. And in
section 4, the explicit assembly model representation for type-
2 assemblies is developed. Section 5 draws the conclusions.
2. Mathematical representation of assembly geometry
In general, assemblies can be classified into two types,
type-1 assemblies and type-2 assemblies [15]. Type-1
assemblies are composed of parts with fully predefined
mating features. The assembly process is putting parts
together at their prefabricated mating features, and parts are
always properly constrained by the mating features. The final
assembly variations for type-1 assemblies are determined
parts are classified into two categories according to their
functions: datum features and mating features. Datum features
compose the locating interface between part and fixture and
locate a part on a fixture. Mating features compose the mating
interface between parts and joint two parts together.
Fixture Coordinate System (FiCS) is a coordinate system
associated with a fixture that locates a part. FiCS is usually set
according to the locating scheme. The origin of FiCS is set at
the intersection of the locating datum surfaces or their
extension of the located part. The Z-axis is normal to the
primary datum surface. The X-axis is normal to the secondary
datum surface and the Y-axis is normal to the tertiary datum
surface.
For simplicity, the X, Y and Z axes of all the CSs are set
parallel to each other, respectively.
2.2. Mathematical representation of assembly spatial
relationships
Based on the definitions of CSs, the spatial relationships
among assembly elements can be described by relationships
among different CSs. The linear transformation between these
CSs is described using Homogeneous Transformation Matrix
(HTM). For example, CS1 can be defined with respect to (wrt)
CS2 as
completely by the variation induced by each part in the ª t 2  T
T

assembly. Type-2 assemblies are composed of parts whose r12

«¬ 1 T 
2 T
1
º
»¼
(1)
assembly features and relative locations are determined during
the assembly process. Parts are always under-constrained, so T

fixtures are needed to locate parts in the assembly process. As where t12 ª¬t12x t12y t12z º¼ is a location vector and
a result, the final assembly variations for type-2 assemblies T12
T
ª¬T12x T12y T12z º¼ is a rotation vector. r12 is a coordinate
are crucially dependent on the locating scheme. In order to
support variation analysis, it is then necessary to represent vector. The HTM of r12 can be written as
assembly geometry mathematically.
ª R12 t12 º
2.1. Coordinate systems T12 T  r12  « » (2)
¬0 1¼
Parts, features and fixtures are the three assembly elements
involving in an assembly process. Describing their relative where R12 is a rotational matrix of the orientation vector as
positions and geometrical deviations is quite essential for shown in (3) at the bottom of the page.
variation analysis. Therefore, four types of Coordinate Meanwhile, the variations of assembly elements can be
Systems (CSs) are defined to locate parts, features and described by the deviations of their CSs. For example, the
fixtures in space. deviation of CS1 can be defined wrt CS2 as
Global Coordinate System (GCS) is a coordinate system in
which all elements in the assembly are located and oriented. It T
D12 ª d 2  T (δ12 )T º» (4)
is the same as the world coordinate system in a CAD or CAE ¬« 1 ¼
system. In this paper, the GCS is assumed to be error free and
unchanged during the entire assembly process. T
Part Coordinate System (PCS) is a coordinate system where d12 ª¬d 12x d 12y d 12z º¼ is a small translational
associated with a part. T
T
deviation and δT 2
ª¬GT 2
GT 2
GT12z º¼ is a small rotational
Feature Coordinate System (FeCS) is a coordinate system 1 1x 1y

associated with a feature on a specific part. The features of

ªcos T12x cos T12y cos T12z  sin T12x sin T12z  cosT12x cosT12y sin T12z  sin T12x cosT12z cosT12x sin T12y º
« »
R 2
1 R T
2
1  «sin T1x cos T1y cos T1z  cos T1x sin T1z
2 2 2 2 2
 sin T12x cosT12y sin T12z  cosT12x cosT12z sin T12x sin T12y » (3)
«  sin T1y cos T1z
2 2
sin T12y sin T12z cos T12y »
¬ ¼
 N. Cai et al. / Procedia CIRP 27 (2015) 241 – 246 243

deviation. D12 is a Differential Motion Vector (DMV). The

HTM of D12 can be written as (a) (b)

P2
ª R(δT12 ) d12 º
δT1
2
TD 2
1  « »
P2

¬ 0 1¼ MFe2

ª 1 GT12z GT12y d 12x º

« 2 » (5) MFe1 MFe2

« GT1z 1 GT12x d 12y » MFe1 P1

P1
« GT12y GT12x 1 d 12z » GCS GCS
« »
«¬ 0 0 0 1 »¼ Fig. 1. Example of type-1 assembly (a) before assembly; (b) after
assembly.
δT12 is defined as Variation Transformation Matrix (VTM).
From equation (5), we can obtain a linear relationship
between D12 and δT12 . Then, once δT12 is known, D12 can be P1 MFe1 P2 Assembly
MFe2
obtained by a linear mapping function:
Fig. 2. Liaison graph.
D12 M (δT12 ) (6)
n
n P1
TMFe
n MFe1
TMFe TPMFe 2
n
The actual HTM and the nominal HTM between CS1 and
n
TPG1
1
n G
2
n G
2 n
TPG2 TPP21
TMFe1
TMFe 2

CS2 can be related by VTM δT12 as

Fig. 3. Nominal spatial relationship graph.
T12 n
T12 ˜ δT12 (7)
From Fig.3, it can be shown that:
And the VTM δT12 can be expressed as The nominal positions and orientations of assembly
elements are represented as { n TPG1 , n TMFe
G
1,
n G
TMFe 2 , n TPG2 }.

 T12  ˜ T12 The spatial relationships between assembly elements are

1
δT12 n
(8)
represented as { n TMFe
P1
1,
n MFe 2
TP 2 }.

3. Assembly model representation for type-1 assemblies The mating relationship is represented as { n TMFe
MFe 1
2 }.

The nominal assembly of P1 and P2 is modeled as

For type-1 assemblies, parts are constrained completely by
1 ˜ TMFe 2 ˜ TP 2
n
the mating features between them. Therefore, the overall TPP21 n P1
TMFe n MFe 1 n MFe 2
(9)
variation of the assembly is determined by the parts and
features only, irrelated with the assembly process. The 3.2. Variational assembly model representation
assembly elements only involve parts and features for type-1
assemblies. Due to manufacturing errors, the features may be deviated
from their nominal positions and orientations. Meanwhile, due
3.1. Nominal assembly model representation to assembly error, the mating interfaces may also be deviated.
All the deviations will be transmitted along assembly process
An example shown in Fig.1 is used to illustrate the chain and cause the variation of the final assembly.
proposed method for type-1 assemblies. Part 1 (denoted as P1) Based on the nominal spatial relationship graph, the
and part 2 (denoted as P2) are assembled together through variational spatial relationship graph can be acquired as
mating features of P1 and P2 (denoted as MFe1 and MFe2). shown in Fig.4. The actual positions and orientations of
The corresponding liaison graph is shown in Fig.2. Based on assembly elements are represented as the HTMs of their
the liaison graph, the nominal spatial relationship graph can actual CSs wrt GCS. The variations of features are
be acquired as shown in Fig.3. represented as the VTMs of FeCSs wrt the corresponding
The nominal positions and orientations of assembly PCSs. The variations of the mating interfaces are represented
elements are represented as the HTMs of their nominal CSs as the VTMs between the FeCSs of the mating features.
wrt GCS. The nominal spatial relationships between assembly From Fig.4, it can be shown that:
elements are represented as the HTMs between their nominal The actual positions and orientations of assembly elements
CSs. Moreover, the mating relationships are represented as
the HTMs between the nominal CSs of the mating features.
Fe 2 ˜ GTMFe 2
n
n P1
TMFe1 ˜ GTMFe1
P1
TMMFe 1 MFe1
TP 2 ˜ GTPMFe
n MFe 2 2

Finally, the nominal assembly is modeled as a chain of HTMs TPG1 G

TMFe1
G
TMFe 2
2
TPG2 TPP21
from one nominal CS to another thus from part to part.
Fig. 4. Variational spatial relationship graph.
244 N. Cai et al. / Procedia CIRP 27 (2015) 241 – 246

G
are represented as { TPG1 , TMFe G G
1 , TMFe 2 , TP 2 }.
P1
Fi1 Fi2
The variations of features are represented as { δTMFe 1 ,

δTPMFe
2
2
}.
The variation of the mating interface is represented as
MFe 1 DFe1 DFe2
{ δTMFe 2 }.

The actual assembly of P1 and P2 with variation is

modeled as P1 P2 Assembly
MFe1 MFe2

TPP21 1 ˜ δTMFe 1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2 ˜ δTPMFe

n P1 P1 n MFe 1 MFe 1 n MFe 2 2
TMFe 2 (10)
Fig. 5. Liaison graph.

The variation of the final assembly is modeled as stack up n

TFiG1 n
TFiG 2
of the DMVs of PCSs of all the parts wrt GCS, which can be
obtained by the actual and nominal HTMs of the PCSs wrt n
TFi1
DFe1
n
TFi 2
DFe 2

GCS.
In this example, it is assumed that there is no deviation of n G
TDFe n G
TDFe
1 2
PCS of P1 wrt GCS, then DGP 1 can be calculated by equations n
TPDFe1 n
TPDFe 2
1 2
(11), (12) and (13). n P1 n MFe1
TMFe
n
TPMFe 2
n
TMFe1 2 2 n
TPP21
TPG1 n G
TMFe
n G
TMFe
n
TPG2
1 2

TPG1 n
TPG1 (11)
Fig. 6. Nominal spatial relationship graph.
δT G
P1  n
T P1 
G 1
˜T G
P1  n
T G 1
P1  ˜ T n G
P1 I (12)
totally located by fixture 1 (denoted as Fi1) through datum
DGP 1 M (δTPG1 ) M (I) 0 (13) feature 1 (denoted as DFe1). Part 2 is underconstrained by the
mating features, so it is additionally located by fixture 2
(denoted as Fi2) through datum feature 2 (denoted as DFe2).
Similarly, DGP 2 can be calculated by equations (14), (15) The corresponding liaison graph is shown in Fig.5. Based on
and (16) shown at the bottom of the page. the liaison graph, the nominal spatial relationship graph can
Then the variation of the final assembly is be acquired as shown in Fig.6.
The parts are located by fixtures, and the nominal fixture
T
V ª DG T D  G T
º (17) locating schemes are represented as the HTMs of PCSs wrt
¬« P 1 P2
¼» the corresponding FiCSs.
From Fig.6, it can be obtained that:
4. Assembly model representation for type-2 assemblies The nominal positions and orientations of assembly
elements are represented as { n TFiG 1 , n TDFe
G
1 ,
n G
TP 1 , n TMFe
G
1 ,
For type-2 assemblies, parts are underconstrained by their n G n G
TP 2 , n TDFe
G n G
TMFe 2 , 2 , TFi 2 }.
mating features and fixtures are needed to locate parts and in-
The spatial relationships between assembly elements are
process adjustment can be incorporated during the assembly
process. Therefore, the overall variation of the assembly is represented as { n TDFe Fi 1
1 ,
n DFe 1
TP 1 , n TMFe
P1
1 ,
n MFe 2
TP 2 , n TPDFe
2
2
,
n Fi 2
determined not only by the parts but also by the fixtures and TDFe 2 }.
the assembly process. The assembly elements involve parts, The mating relationship is represented as { n TMFe
MFe 1
2 }.
features as well as fixtures for type-2 assemblies. The nominal fixture locating scheme of P1 is modeled as
4.1. Nominal assembly model representation
e 1 ˜ TP 1
n
TPFi11 n Fi 1
TDF n DFe 1
(18)
A type-2 assembly composed of two parts is used to
illustrate the proposed method. Part 1 (denoted as P1) and part The nominal fixture locating scheme of P2 is modeled as
2 (denoted as P2) are assembled together through mating equation (19).
features of P1 and P2 (denoted as MFe1 and MFe2). Part 1 is

TPG2 TPG1 ˜ TMFe

P1
1 ˜ TMFe 2 ˜ TP 2
MFe1 MFe 2 n
TPG1 ˜ n TMPFe
1
1 ˜ δTMFe1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2
P1 n MFe1 MFe1 n MFe 2
˜ δTPMF
2
e2
(14)

 TPG2  ˜ TPG2  TPG2  ˜ n TPG1 ˜ n TMFe

1 1
δTPG2 1 ˜ δTMFe1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2 ˜ δTPMFe
n n P1 P1 n MFe1 MFe1 n MFe 2 2
2 (15)

DGP 2 M  δTPG2  M  T n G 1
P2  ˜ n TPG1 ˜ n TMFe
P1
1 ˜ δTMFe1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2
P1 n MFe1 MFe1 n MFe 2
˜ δTPMFe
2
2
 (16)
 N. Cai et al. / Procedia CIRP 27 (2015) 241 – 246 245

e 2 ˜ TP 2
n
TPFi22 n Fi 2
TDF n DFe 2
(19)
TFiG1 n
TFiG1 ˜ GTFGi1 TFiG 2 n
TFiG 2 ˜ GTFGi 2
n
TFi1
˜ GT
Fi1 n
T Fi 2
˜ GTFi 2

The nominal assembly of P1 and P2 is modeled as DFe1 DFe1 DFe 2 DFe 2

G G
TDFe TDFe
1 ˜ TMFe 2 ˜ TP 2
n
TPP21 n P1
TMFe n MFe 1 n MFe 2
(20) 1 2

n
TDFe1
P1 ˜ GT DFe1
P1
n DFe 2
T
P2 ˜ GTDFe 2
P2

Fe 2 ˜ GTMFe 2
n
n P1
˜ GT P1
TMMFe 1 MFe1 n MFe 2
˜ GT MFe 2
4.2. Variational assembly model representation TPG1
T
MFe1 MFe1
G G
TMFe
T
P2 P2
TPG2 TPP21
TMFe1 2

Based on the nominal spatial relationship graph, the Fig. 7. Variational spatial relationship graph.
variational spatial relationship graph can be acquired as
shown in Fig.7. Since the fixtures are used to locate parts, two
more variations including variations of fixtures and variations modeled as shown in equation (23).
of locating interfaces are introduced in the assembly process. The variation of the final assembly is modeled as stack up
The variations of fixtures are described by the VTMs of their of the DMVs of PCSs of all the parts wrt GCS, which can be
FiCSs wrt GCS. The variations of the locating interfaces are obtained by the actual and nominal HTMs of the PCSs wrt
represented as the VTMs between FeCSs of the datum GCS.
features and FiCSs of the corresponding locating fixtures. In this example, P1 is located by Fi1 through DFe1, so the
From Fig.7, it can be obtained that: actual HTM of P1 wrt GCS can be calculated by equation (24),
The actual positions and orientations of assembly elements and the DMV of PCS of P1 wrt GCS can be obtained by
are represented as { TFiG 1 , TDFe
G G G G G equations (25) and (26).
1 , TP 1 , TMFe 1 , TMFe 2 , TP 2 ,
G G
P2 is located by Fi2 through DFe2 and by P1 through
TDFe 2 , TFi 2 }. MFe2 simultaneously. So the actual HTM of P2 wrt GCS is
The variations of features are represented as { δTPDFe
1
1
, determined by both Fi2 and P1, and it can be calculated by
P1 MFe 2
, δTPDFe 2 equations (27), (28) and (29). In (27), Fi 2 TPG2 represents actual
δTMFe 1 , δTP 2 2 }.
The variations of fixtures are represented as { δTFiG 1 , δTFiG 2 }. HTM of P2 wrt GCS determined by Fi2. And in (28), P 1 TPG2
The variation of the mating interface is represented as represents actual HTM of P2 wrt GCS determined by P1. In
MFe 1
{ δTMFe (29), F is the combination function determined by the
2 }.
specific locating scheme of P2.
The variations of locating interfaces are represented as The DMV of PCS of P2 wrt GCS can be obtained by
Fi 1 Fi 2
{ δTDFe 1 , δTDFe 2 }. equations (30) and (31).
The actual fixture locating scheme of P1 is modeled as
 TPG2  ˜ TPG2  TPG2  ˜ F  Fi 2 TPG2 , P 1 TPG2 
1 1
δTPG2 n n
(30)
TPFi11 1 ˜ δTDFe 1 ˜ TP 1 ˜ δTPDFe
n Fi 1 Fi 1 n DFe 1 1
TDFe 1 (21)

The actual fixture locating scheme of P2 is modeled as DGP 2 M (δTPG2 ) M  Tn G 1

P2  ˜ F  Fi 2 TPG2 , P 1 TPG2   (31)

TPFi22 2 ˜ δTDFe 2 ˜ TP 2 ˜ δTPDFe

n Fi 2 Fi 2 n DFe 2 2
TDFe (22)
2
Then the variation of the final assembly is shown as
equation (32).
The actual assembly of P1 and P2 with variation is

TPP21 1 ˜ δTMFe1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2 ˜ δTPMFe

n P1 P1 n MFe1 MFe1 n MFe 2 2
TMFe 2 (23)

TPG1 n
TFiG1 ˜ GTFGi1 ˜ n TDFe1 ˜ GTDFe1 ˜ TP1
Fi1 Fi1 n DFe1
˜ GTPDFe
1
1
(24)

 TPG1  ˜ TPG1  TPG1  ˜ n TFiG1 ˜ GTFiG1 ˜ n TDFe

1 1
δTPG1 1 ˜ GTDFe1 ˜ TP1 ˜ GTPDFe
n n Fi1 Fi1 n DFe1 1
1 (25)

DGP1 M (δTPG1 ) M  T
n
P1
G 1
˜ n TFiG1 ˜ GTFiG1 ˜ n TDFe
Fi1
1 ˜ GTDFe1 ˜ TP1
Fi1 n DFe1
˜ GTPDF
1
e1
 (26)

Fi 2
TPG2 n
TFGi 2 ˜ GTFiG 2 ˜ n TDFe
Fi 2
2 ˜ GTDFe 2 ˜ TP 2
Fi 2 n DFe 2
˜ GTPDFe
2
2
(27)

P1
TPG2 TPG1 ˜ n TMFe
P1
1 ˜ δTMFe1 ˜ TMFe 2 ˜ δTMFe 2 ˜ TP 2
P1 n MFe1 MFe1 n MFe 2
˜ δTPMFe
2
2
(28)

TPG2 F  Fi 2 TPG2 , P1 TPG2  (29)

246 N. Cai et al. / Procedia CIRP 27 (2015) 241 – 246

T [3] Homem de Mello LS, Sanderson AC. A correct and complete algorithm
V ª DG T D 
G T
º (32)
¬« P 1 P2
¼»
5. Conclusion
In order to support variation analysis, a mathematical
assembly model representation based on coordinate system
definition and homogeneous transformation matrix is
proposed in this paper. With the proposed method, the
nominal and variational spatial relationships among fixtures,
parts and features are defined. The variation of the final
assembly is modeled as stack up of the differential motion
vectors of part coordinate systems of all the parts with respect
to global coordinate system, and it can be obtained by the
actual and nominal homogeneous transformation matrices of
the part coordinate systems with respect to global coordinate
system. The relationship between the input variation sources
and the output assembly variation can be established
mathematically. The different situations for two types of
assemblies are also distinguished. The proposed assembly
model representation provides the basis for variation analysis
so as to better evaluate the dimensional quality of assemblies.
Acknowledgements
This work is supported by advanced research project,
research foundation (A0320131002) and Beijing Municipal
Special Funding of Building a Project. The authors would like
to thank those agencies.
References
[1] Bourjault A, Henrioud JM. Computer aided mechanical assembly
planning, Boston: Kluwer Academic Publishers 1991.
[2] Wilson RH. On geometric assembly planning, Ph.D. thesis, USA:
Stanford University (1992).
for the generation of mechanical assembly sequences. IEEE Transactions on
Robotics and Automation 1991; 7(2): 228-240.
[4] Yin ZP, Ding H, Li HX, Xiong YL. A connector-based hierarchical
approach to assembly sequence planning for mechanical assemblies.
Computer Aided Design 2003; 35: 37-56.
[5] Chen X, Gao S, Yang Y, Zhang S. Multi-level assembly model for top-
down design of mechanical products. Computer Aided Design 2012; 44:
1033-1048.
[6] Wang H, Ceglarek D. Representation, generation, and analysis of
mechanical assembly sequences with k-ary operations. Journal of Computing
and Information Science in Engineering 2012; 12(1): 011001--1-011001--12.
[7] Jiang D, Na G, Ji K. Research on assembly model based on assembly
constraints transformation. 2012 International Workshop on Information and
Electronics Engineering (IWIEE) Procedia Engineering 2012; 29:770-774.
[8] Gao S, Zhang S, Chen X, Yang Y. A framework for collaborative top-
down assembly design. Computers in Industry 2013; 64(8): 967-983.
[9] Ceglarek D, Huang W, Zhou S, Ding Y, Kumar R, Zhou Y. Time-based
competition in multistage manufacturing: stream-of-variation analysis
(SOVA) methodology—review. The International Journal of Flexible
Manufacturing Systems 2004; 16(1): 11-44.
[10] Mantripragada R, Whitney DE. Modeling and controlling variation
propagation in mechanical assemblies using state transition models. IEEE
Transactions on Robotics and Automation 1999; 15(1): 124-140.
[11] Huang W, Lin J, Bezdecny M, Kong Z, Ceglarek D. Stream-of-variation
modeling I: a generic 3D variation model for rigid-body assembly in single
station assembly processes. Transctions of the ASME - Journal of
Manufacturing Science and Engineering 2007; 129(4): 821-831.
[12] Huang W, Lin J, Kong Z and Ceglarek D. Stream-of-variation (SOVA)
modeling II: a generic 3D variation model for rigid-body assembly in
multistation assembly processes. Transctions of the ASME - Journal of
Manufacturing Science and Engineering 2007; 129(4): 832-842.
[13] Cheng H, Wang RX, Li Y and Zhang KF. Modeling and analyzing of
variation propagation in aeronautical thin-walled structures automated
riveting. Assembly Automation 2012; 32(1): 25-37.
[14] Franciosa P, Gerbino S and Patalano S. A computer–aided tool to
quickly analyse variabilities in flexible assemblies in different design
scenarios. International Journal of Product Development 2013; 18(2): 112-
133.
[15] Whitney DE. Mechanical assemblies: their design, manufacture, and role
in product development, New York: Oxford University Press 2004.