Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 21 (2014) 159 – 164

24th CIRP Design Conference

Design and Evaluation of In-Line Product Repair Strategies for Defect

Reduction in the Production of Electric Drives
M. Colledania*, D. Coupekb, A. Verlb, J. Aichelec, A. Yemanea
a
Politecnico di Milano, Mechanical Engineering Department, Via la Masa, 1, 20156, Milano, Italy
b
Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW), Seidenstraße 36, 70174 Stuttgart, Germany
c
Corporate Research Robert Bosch GmbH, Postfach 300240, 70442 Stuttgart, Germany
* Corresponding author. Tel.: +39-02-23998587; fax: 39-02-23998585. E-mail address: [email protected]

Abstract

Manufacturing companies are continuously facing the challenge of operating their manufacturing processes and systems in order
to deliver the required production rates of high quality products of increasing complexity, with limited use and waste of
resources. This aspect is particularly critical in emerging sectors, such as the e-mobility industry, where state of the art quality
and process control technologies show strong limitations. This paper proposes new solutions for implementing in-line product
repair strategies in the production of electric drives for the automotive industry. Moreover, it develops an innovative quantitative
tool to estimate the impact of the proposed strategies on the overall process-chain performance. The benefits of the approach are
validated within a real industrial context.
Published
© 2014 The by Published
Authors. Elsevier B.V. This is an
by Elsevier open access article under the CC BY-NC-ND license
B.V.
Selection and peer-review under responsibility of the International Scientific Committee of “24th CIRP Design Conference” in the person of
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
the Conference
Selection Chairs Giovanni
and peer-review Moroni and Tullio
under responsibility of theTolio.
International Scientific Committee of “24th CIRP Design Conference” in the person of the
Conference Chairs Giovanni Moroni and Tullio Tolio
Keywords: Product Repair, Electric Drives Assembly, Zero-defect Manufacturing;

1. Introduction, Motivation and Objectives process level the creation of a non-conformity on the
manufactured workpiece, and the defect propagation phase,
"Zero Defect Manufacturing" is an emerging paradigm i.e. at process-chain level, the transmission of a defect
aiming at going beyond traditional six-sigma approaches in throughout the stages of the manufacturing system. The work
highly technology intensive and emerging strategic reported in this paper is focused on this second issue.
manufacturing sectors through knowledge-based approaches.
Traditional six-sigma techniques show strong limitations in 1.1. Industrial context of reference
highly turbulent, variable and changeable production contexts,
characterized by small batch productions and in-line product Cars are with about 12 million manufactured units (EU27)
inspections. Innovative and integrated product, process and one of the most important products in Europe. Due to the
system design, management and control methods as well as threatening lack of petrol and efforts for a healthy
advanced technological enablers have a key role to achieve the environment the change from the combustion engine towards
overall "Zero Defect Manufacturing" objective. electrical drives in cars is on going. For example, suppliers
The work proposed in this paper is part of the activity of and car manufacturers in Germany are working together to
the European funded project “MuProD - Innovative proactive reach the ambitious goal of 1 million sold electrical vehicles
quality control system for in-process multi-stage defect in 2020. Therefore, the current manufacturing processes for
reduction”. The zero defect manufacturing paradigm proposed producing electrical drives must be improved to support the
in MuProD is based on methodological and technological achievement of this goal. As a result of the complete different
advances tackling both the defect generation phase, i.e. at construction technique of electrical motors, the perfected

2212-8271 © 2014 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/3.0/).
Selection and peer-review under responsibility of the International Scientific Committee of “24th CIRP Design Conference” in the person of the
Conference Chairs Giovanni Moroni and Tullio Tolio
doi:10.1016/j.procir.2014.03.186
160 M. Colledani et al. / Procedia CIRP 21 (2014) 159 – 164

methods of manufacturing and quality control of combustion 1.3. Objectives of the paper
engines cannot be directly transferred to electrical drives.
In the production of automotive electric drives, the state of As shown in the previous section, an approach for
the art quality control is the so called “End Of Line” (EOL) considering several different defect management actions in
testing, as the major final product functional and quality test the same framework and evaluate their impact on the overall
and as approval test for the customers. This testing method is integrated quality and production logistics performance of the
executed after all manufacturing steps have been completed system has never been proposed. Moreover, advanced
and can therefore been classified as off-line inspection. If a technologies for in-line repair of electric drives are currently
defect occurs in one of the upstream production stages, it will not industrial state of the art solutions. Issues like “What is
not be detected in-line. Consequently, value adding processes the impact of product rework at a given process stage on the
will still be applied on an already defective product. output throughput of conforming items?” or “What are the
Following the EOL testing approach, there is no possibility of overall benefits of selective assembly at system level?” still
applying in-process quality control techniques, since process remain unsolved. The objective of this paper is to develop a
data are not available at the relevant process stages. general methodology and quantitative tools to design on-line
To overcome this drawback, a new inline quality defect management policies in manufacturing systems. The
inspection strategy, involving a new device and a new effectiveness of the proposed approach is demonstrated within
methodology, is developed within the EU funded research the electric drive production system at Bosch. The paper is
project MuProD. In addition to EOL testing, inspections are structured as follows: in Section 2, the current process-chain
also shifted to upstream process stages. By doing so, more in the industrial case is described. In Section 3, the defect
detailed information about the product quality features can be management strategies are described in detail. In Section 4,
gathered. For example, the currently applied total magnetic the developed system level model is explained and applied to
flux measurement of the laminated steel stacks can be the real system in Section 5. Results are shown in Section 6.
replaced by a space-resolved measurement of the flux. This
permits the identification and allocation of deviations in the 2. Description of the Rotor Assembly Line
magnetic field caused by specific defective or weak magnets.
With this new inspection technology it is possible to support The current production process for electric drives is
the development and implementation of advanced in-line represented in Fig. 1, where squares represent processing and
strategies for profitably managing non-conforming stacks, inspection stages (Mi) and circles represent buffers (Bi,j) for
thus avoiding the propagation of defects throughout the storing inventory between Mi and Mj. The line produces a
process stages. However, in order to select the most proper number T of different rotors, t=1,..,T. A rotor t is composed of
defect management strategy, the analysis of their impact on St laminated stacks, which can be seen as the size of the batch
the overall integrated quality and production logistics of stacks to be assembled. Each stack has Ns magnets. The
performance of the process chain needs to be carried out. line is composed of two main branches, respectively dedicated
to the assembly of the rotor and to the production of the stator.
1.2. Literature Review The focus of this study is the rotor line. This line is composed
of seven main stages, dedicated to the following operations:
The design and development of methodologies and • M1: loading of the stacks on the pallet.
technologies for defect management practices in the • M2,1,M2,2: assembly of the magnets on the stacks. The
automotive industry, including in-line product repair, scrap, station is composed of a pick and place system for
and rework have been recently addressed in the scientific positioning the magnets in their locations.
literature. In [1] the problem of designing in-line rework • M3: stack magnetization and total flux measurement.
practices in automotive paint shops, jointly considering • M4: heating station. A rotating table moves the stacks into
quality and productivity implications, is addressed. Product a heating chamber.
repair and defect propagation in multi-stage systems has also • M5: assembly machine. The required number of stacks is
been addressed in battery manufacturing [2]. The application taken and a pile of stacks in the z direction of the machine
of selective assembly strategies in the automotive industry has is formed by mounting each stack on the central shaft.
also been investigated. However, selective assembly is mainly • M6: rotor balancing station.
being proposed for traditional mechanical problems, such as • M7: rotor marking station.
the sleeve-and-shaft type assembly. In [3] a General Selective After assembling the rotor and the stator, the completed
Assembly approach is presented, which extends the classical motor undergoes the EOL inspection. At this stage, motor
approach of selective assembly. Repair strategies are characteristics as well as customer requirements such as
identified as critical aspects in electric drive production. A torque, speed, etc. are tested. Since defects in the magnetic
prototype system for disassembly of internal magnets in a circle have a considerable effect on the performance of the
rotor stack for product repair is proposed in [4]. In [5] two whole electric car, 100% EOL testing is needed.
possible assembly strategies were presented for the product Each stage in the system is subject to breakdowns,
repair in the production of electric drives. These strategies characterized by a failure rate p, which is the inverse of the
influence the performance of the overall production system mean time to failure, and a repair rate r, which is the inverse
and this aspect is generally neglected. of the mean time to repair. The company collects estimates of
these parameters. Their value is not provided for
 M. Colledani et al. / Procedia CIRP 21 (2014) 159 – 164 161

confidentiality reasons. Moreover, each stage is characterized stage in the order in which they are produced in the previous
by a specific processing rate (parts/time) that is also omitted. stages. A priori, there is no knowledge available about the
stacks and the magnetization of the single magnets. Therefore,
Magnets
no sorting policy is applied on the stacks before entering the
B0,1
assembly station. Once the stacks are piled up, the resulting
B1,1 M 2,1 rotor is inspected after M5 by a space resolved measurement
M1 B2 M3 B3 M4 B4 M5 M6 M7 ASSEMBLY
Conforming
parts
of the magnetic field. Stacks containing magnets with
B1,2 M 2,2 Mg M K −1 MK deviations from the nominal value can be rotated for
STATOR LINE
compensation if there is a second peak in a different stack.
B0,2 Scrap
Mx Mz Mk Mj or rework Then one stack can be rotated so that the weak magnet of the
Magnets
first stack is located next to the strong magnet of the second
Process Stages Inspection Stage
stack. If the deviation in the magnetic field exceeds the
M1: laminated stack from M4: heating station MK: EOL inspection
suppliers loading
M2: magnets assembly
M5: rotor assembly tolerances and cannot be compensated (e.g. broken magnet),
M6: rotor balancing
M3: stack magnetization M7: marking then the defective stack has to be replaced by a stack from a
buffer. In both cases, the output of the assembly process is a
Fig. 1. Current production line for electric drives with EOL inspection. fully assembled rotor with a magnetic field that is within the
desired tolerances.
In the current production line, several defect sources have
been identified and classified. Each defect is coded as (i,j)
where i indicates the stage where it is generated and j is a
progressive index. These defects are described in Table 1.
After applying the MuProD solution to the production
system, new measurement devices will be included at stages
M3 and M5 for detection of deviations in the magnetic field of
a single rotor stack and the complete rotor, respectively. In
details, these actions will consist in:
• The implementation of a sensor for the space resolved
measurement of the magnetic flux of each stack. This
results in an inspection point located after machine M3.
• The development of a new multi-sensor system distributed
in the z axis of the rotor, for measuring the field of each
stack in the assembled rotor and check for uniformity of
the overall rotor magnetic flux after stage M5.
Fig. 2. Defect management actions: proposed reference framework. Red lines
3. Characterization of Defect Management Strategies represent costs for the company.

A reference framework that comprehensively describes all 3.2. Downstream Repair

possible defect management actions is shown in Fig. 2. This is
the basis for the selection of those technically feasible actions, As deviations in the magnetic field of single rotor stacks
among which the optimal solution to be implemented for each are generated at stage M3, the goal is to measure them after M3
specific defect type can be selected. The main classes of and to compensate these deviations by applying an optimal
defect management strategies include: (1) No action, (2) strategy ʌopt in the downstream assembly stage M5, where a
Scrap and (3) Repair. This last class includes different options number St of stacks is assembled to form one rotor. Two
that are further detailed and described, relating to the possible downstream repair methods are investigated, namely
industrial reference case, in the following. sequential assembly and selective assembly.
Sequential Assembly. A batch of St stacks is produced,
3.1. In-line Rework stored in a buffer and inspected, so that St magnetic profiles
Bs(n), n=1,..,Ns, are available, where n indicates the specific
In-line rework involves product inspection and repair in the magnet. The space resolved measurement of stacks yields the
same process stage. In this way, the defective part can be profiles matrix B.
corrected without re-clamping. The workpiece repair can be
ª B1 ( n) º ª b1,1 " b1, N1 º
performed by one of the following possible controller actions. « » « »
Firstly, machine or process parameters can be adjusted; B=« # »=« # # » (1)
secondly, the actuators can be directly set, either by switching « BS ( n) » «bS ,1 " bS , N »
them on/off (e.g. heaters, valves) or by setting new continuous
¬ t ¼ ¬ t t s ¼

or discrete set-points; thirdly new code-segments to the The aim of this approach is to change the stack order and
numeric control of the machine can be generated and loaded. to impose an angular misalignment Į between the St stacks
In the production of electric drives, rotor assembly at stage with respect to a reference axis in order to gain uniformity and
M5 can be seen as a process with St stacks as input and exactly reduce variability of the output field intensity. This yields the
one complete rotor as output. The stacks enter the assembly optimal sequential assembly strategy ʌopt for a specific batch
162 M. Colledani et al. / Procedia CIRP 21 (2014) 159 – 164

of measured St stacks. The entity of the misalignment, namely Non-uniform

magnetic field
the elements of the vector Į, has to be computed by an (5,1) intensity of the Geometric M5 M5
1, 2.1, 2.2,
optimization algorithm. In order to find the global optimum a 3.1, 3.2.1
stacks in the
brute-force method is used as it searches for the minimum in rotor.
the complete solution space. The number of possible
permutations ʌall, as well as the computational time, grow 4. Production Line Model
exponentially with respect to the number of stacks St.
St −1 A generalized production system model for the joint
§ NS ·
π all = ¨ ¸ (2) analysis of quality and production logistics performance under
© 2¹ application of the defect repair strategies is developed to study
To reduce the computational effort and consequently the their influence at system level. The proposed model analyzes
negative influence on the production cycle time, the value of a general manufacturing system that is composed of multiple
ʌall has to be decreased, with the risk of deteriorating the processing stages (blue squares) and inspection stages (red
quality of the proposed assembly strategy ʌopt. Two strategies squares) defined as Mk, k=1,..,K, (Fig. 3).
Defective
are investigated for reducing the number of combinations. The products
first approach is to consider only stacks with magnets out of
tolerances in order to decrease the exponent of ʌall. The M0 M1 B1,2 M2 B2,3 M3 B3,4 M4 M5 B5,6 M6
second approach aims at reducing ʌall by consideration of the
Scrap Conforming
most relevant magnets of the stacks within one batch. Downstream repair Rework
products
Selective Assembly. In industry, selective assembly is
applied to produce high precision assemblies from low Fig. 3. Modelling formalism for a generic multi-stage production line.
precision components [6]. Selective assembly consists in
measuring the key quality characteristics of each sub- Individual stages are connected by material transportation
component and sorting the components into bins according to systems or inter-operational buffers (yellow circles), Bi,j
the measurement outcome. Depending on the space resolved storing work in progress between stage Mi and Mj. For
magnetic field measurement Bs(n) of each stack, clusters of example, in Fig. 3, processing stages M1, M2 and M4 perform
stacks Ci are formed and the stacks are temporarily stored in manufacturing transformation processes on the incoming
class-dependent buffers. The binning strategy ʌbinning fixes the workpieces. Inspection stations M0, M3 and M5 measure key
number of clusters and the sorting policy. For improving the quality features of parts processed at upstream manufacturing
product quality, the assembly station is allowed to assemble stages. Based on the information collected by inspection
components only according to he assigned bins’ matching stations, defect management actions can be taken at inspection
policy ʌmatching. This strategy determines how many stacks machines and these actions are indicated in blue arrows.
have to be taken from each class Ci to form one rotor. The The behaviour of each stage is modelled as a continuous time-
assembly policy ʌassembly finally defines the loading policy discrete state Markov chain of general complexity. The
from different classes. underlying transition rate matrix is λǤ This framework allows
to model machines having multiple operational and failure
3.2.1. Selection of Technically Feasible Actions states, connected by means of an arbitrarily complex
Markovian structures. When the machine is in an operational
For each defect type, the technically feasible defect state o, it processes parts at a rate of μo parts per minute. A
management actions, selected in the framework of Fig.2, are breakdown state is simply characterized by μ=0. These
associated (Table 1). In order to select among the feasible processing rates [parts/t.u.] are collected in the quantity
actions the most suitable solution for each defect type a model reward vector μ. For each operational state a statistical
will be presented that considers, at system level, the impact of distribution of the processed quality characteristic y is
the defect management action on the production logistics assumed, namely fo(y). According to the Specification Limits
performance (Work in Progress, throughput) as well as on the imposed by design on the processed feature, the yield is
quality performance (yield). defined for every state o, namely Yo; these elements are
collected in the quality reward vector Y. The total fraction of
Table 1: Typical defects and corresponding potential management strategy. defects generated by the stage is denoted as γ. The
performance measures of interest are the following:
Defect Defect Defect Stage Stage Feasible
Code Description Category Generated Inspected Actions
• Average total production rate of the system, ETot, including
1, 2.1, 2.2, both conforming and defective parts, observed in output.
(2,1) Missing magnet. Binary M2 M3 3.1, 3.2.1, • Average effective production rate, EEff, of conforming
3.2.2 parts, observed in output.
Magnet with low 1, 2.1, 2.2,
(3,1) magnetic field Dimensiona M3 M3 3.1, 3.2.1, • System yield, Ysystem, that is the fraction of conforming
intensity. 3.2.2 parts produced by the system (EEff / ETot ).
Non-uniform • WIP, which is the total average inventory of the system.
magnetic field
intensity of the
1, 2.1, 2.2, Having derived the characteristic parameters (λi, μi Yi) for
(3,2)
magnets in the
Geometric M3 M3 3.1, 3.2.1, each stage, the steady-state probability vector πi of the
3.2.2
stack.
 M. Colledani et al. / Procedia CIRP 21 (2014) 159 – 164 163

Markov chain and the performance of the stage in isolation, μ=μ U psq μ=μ Sq
pre μ=μ re
i.e. not integrated in the production line, can be computed: μ=μ U U Sq
U Re rsq
π i λi = 0 rre
p r
EiEff (3) p r p
EiTot = π i ⋅ μiT EiEff = π i ⋅ diag(μi ) ⋅Yi T =
Mi
Y p
Tot
E i D
D μ=0 (b)
The above formalism does not consider the application of the (a)
μ=0
defect management actions. Since the defect management
policies affect the material flow and the behavior of stages, Fig. 5. State transition diagram of stages performing rework (a) and
this impact has to be included in the stage models. sequential assembly (b).

5. Modeling the Effect of Defect Repair Strategies at The modelling of this policy requires knowledge on the time
Process-chain Level required for the rework operation i.e. RT. The adjusted
transition rates and processing rates of the machine in the
5.1. Model of a stage performing scrap reworking state (Re) can be evaluated with equations (5).
pre = μU ⋅ γ
( )
Scrapping can be an appropriate policy on defective parts
if repairing the non-conformity is not economical or is rre = μU ⋅ 1− γ
technically infeasible. There are different actions that can be μU (5)
considered as post treatment for the scrapped parts that might μre =
1 + RT ⋅ μU
entail further decisions, as sketched in the defect classification
framework of Fig. 2. However, in this analysis, these further
decisions are neglected. The state transition diagram of a 5.3. Model of a stage performing downstream repair policies
machine with single failure mode without (a) and with (b) the
scrapping policy is shown in Fig. 4. The transition rates can Sequential assembly: under the sequential assembly policy
be evaluated using equation (4). State U is the operational the machine performs measurement, optimization and
state and D is the down state. Since produced parts do not assembly of components, according to the optimized angle.
proceed to the next stages when scrapping is activated, the Since additional time is needed for these activities, the
processing rate of the scrapping state Sc is set to 0. production logistics behavior of this stage needs to be
psc = μU ⋅ γ adjusted accordingly. Consequently, this policy can be
modeled with the state transition diagram in Fig. 4, with the
rsc = μU ⋅ 1− γ ( ) (4) specific set of transition and processing rates given in
equation (6). In this case, the times related to assembly (Tasbl),
μ=μ U μ=μ U psc μ=0 inspection (Tinsp), optimization (Topt), and additional
U U Sc operations, i.e. part positioning (Tadd) must be considered.
rsc
p p r
psq = μU ⋅ γ
r
p
rsq = μU ⋅ 1− γ( )
D D 1
= Tasbl + Tinsp ;
1
= Tasbl + Tinsp + Tadd + Topt
(6)
μ=0 μ=0
(a) (b) μU μ sq

Fig. 4. State transition diagram of stages without (a) and with (b) scrapping. Selective assembly: By employing selective assembly,
high precision products can be assembled from low precision
5.2. Model of a stage performing inline rework components, at the cost of increasing the complexity of the
system management. In the reference industrial case, selective
The main logistics consequence of inline rework is related assembly is applied as follows. Each stack is sorted in two
to the need to reprocess the fraction γ of defective parts classes depending on the measured total magnetic flux
generated by the stage. After the reprocessing, it is assumed intensity. The buffer size for the two classes is identical and
that the reworked parts continue the flow in the remaining equal to half the size of the buffer in the current configuration.
portion of the line. In Fig. 5(a) the state transition diagram of Then, the assembly machine only couples stacks with high
a stage performing rework is presented. flux with stacks with low flux intensity. Due to the
complexity in the system management, modeling selective
assembly requires more technical mathematical derivations
than previous defect management policies. Due to space
limitations, we omit this derivation. It can be found in [7].
164 M. Colledani et al. / Procedia CIRP 21 (2014) 159 – 164

5.4. System Performance Evaluation a quantitative methodology to support the design of the best
possible strategy by estimating the impact of the actions on
The proposed analytical method is based on a recent idea the overall system performance. The benefits of the approach
of decomposition approach that applies to Markovian are demonstrated within a real industrial process-chain,
machines, characterized by transition rate matrix λ and dedicated to the production of electric drives.
processing rate vector μ, that was recently proposed in [8]. Some guidelines for the implementation of the tools
Therefore, it applies to the stage models described in the supporting this approach are given. Although it is a system
previous section. The idea of the decomposition approach is level approach, some aspects are not directly taken into
to decompose the K-machine system into a set of K-1 two- account. The implementation of selective and sequential
machine one-buffer sub-systems l(k), i.e. one for each buffer assembly strategies in the production system of electric drives
in the original system. The performance of each sub-system requires additional component handling devices. In addition,
can be evaluated with the exact analytical method developed by using an optimal matching policy, the inspection of the
in [9]. The decomposition equations for such general system entire rotor could be removed. Furthermore, the algorithms
settings are provided in [8]. This method proved to be for solving the sequential assembly optimization problems in
accurate in estimating the system performance, showing real-time require high calculation capacity, which must be
errors against simulation below 3%. provided by adequate computers or machine controls. These
implications should be considered before implementation.
6. Numerical Results and Strategy Comparison The proposed approach is general and applicable to
systems in several industries, thus paving the way to the
By combining the stage level defect management policies implementation of the zero-defect manufacturing paradigm in
described in the previous sections and associated to each industry.
defect type at production system level, six system level defect
management scenarios have been generated for the electric Acknowledgements
drive production system under analysis. Considering the
existing defect management policy (no action) as the baseline The research leading to these results has received funding
case, five additional scenarios are evaluated and compared in from the European Union Seventh Framework Program
terms of system performance. The five scenarios are: (FP7/2007-2011) under grant agreement n° 285075.
• Scenario 1: scrap at M3 and rework at M5.
• Scenario 2: no action at M3 and rework at M5. References
• Scenario 3: scrap at M3 and sequential assembly at M5.
• Scenario 4: no action at M3 and sequential assembly at M5. [1] Li, J., Blumenfeld, D.E., Marin, S.P. Manufacturing system design to
• Scenario 5: no action at M3 and selective assembly at M5. improve quality buy rate: an automotive paint shop application study. IEEE
The results are summarized in Table 2. As shown, the Trans Autom Sci Eng 2007; 4 (1), 75-79.
application of the best scenario (scenario 4) can yield an [2] Ju, F., Li, J., Xiao, G., Huang, N., Biller, S. A quality flow model in
improvement of 16.55% in the production rate of conforming battery manufacturing systems for electric vehicles. To appear in IEEE Trans
Autom Sci Eng, 2013.
parts of the system (the time unit is hidden for confidentiality [3] Tan, M., Wu, C. Generalized selective assembly. IIE Transactions 2012;
reasons). The proposed approach for the quantitative analysis 44(1), 27-42.
of defect management policies at system level suitably [4] Klier, T., Risch, F., Franke, J. Disassembly strategies for recovering
supports the strategy design in industrial settings. valuable magnet material of electric drives. Proceedings of the 3rd Int. EDPC,
IEEE Catalog Number CFP1385P-PRT, 2013, pp. 12-14.
Table 2: Comparison of scenarios; electric drive production system. [5] Coupek, D., Verl, A., Aichele, J., Colledani, M. Proactive Quality
Control System for Defect Reduction in the Production of Electric Drives.
EEff ETot Ysystem ¨% EEff vs. Proceedings of the 3rd Int. EDPC, IEEE Catalog Number CFP1385P-PRT,
Strategy 2013, pp. 440-445.
[parts/t.u] [parts/t.u] Baseline
Base line 0.5752 0.6729 0.85 - [6] Iyama, T., Mizuno, M., McKay, K. N., Yoshihara, N., Nishikawa, N.
Scenario 1 0.6478 0.6478 1.00 +12.62% Optimal strategies for corrective assembly approach applied to a high-quality
Scenario 2 0.5693 0.5693 1.00 -1.04% relay production system. Comput Ind 2013; 64, 556–564.
Scenario 3 0.6613 0.6613 1.00 +14.96% [7] Colledani, M., Ebrahimi, D. Performance Evaluation of Selective and
Scenario 4 0.6704 0.6704 1.00 +16.55% Adaptive Assembly Systems. SMMSO 2013.
Scenario 5 0.6261 0.6726 0.93 +8.85% [8] Colledani, M.., Gershwin, S.B. A Decomposition Method for
Approximate Evaluation of Continuous Flow Multi-stage Lines with General
7. Conclusions and Guidelines for Implementation Markovian Machines, Ann Oper Res 2012; 209(1), 5-40.
[9] Gershwin, S., Tan, B. Analysis of a general Markovian two-stage
continuous flow production system with a finite buffer. Int J Prod Econ 2009;
The paper proposes several technical solutions for avoiding 120(2), 327-339.
the propagation of defects in multi-stage production lines and