Professional Documents
Culture Documents
International Journal of
Production Research
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/tprs20
Evaluation of Just-In-Time
alternatives in the electric
wire-harness industry
F. Estrada , J.R. Villalobos & L. Roderick
Published online: 15 Nov 2010.
Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any
opinions and views expressed in this publication are the opinions and
views of the authors, and are not the views of or endorsed by Taylor
& Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information.
Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in
connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-
conditions
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
int. j. prod. res., 1997, vol. 35, no. 7, 1993± 2008
This paper investigates some of the theoretical and practical issues for the
introduction of Just-in-Time (JIT) techniques into an automotive wire-harness
assembly line. Some of the issues explored include the number of production
kanbans and the unit load size to be used for every air of subassembly-
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
harnesses of the same family are additional branches or circuits, position of con-
nectors, and di erent connectors or other components. Di erences arise from cars
having more features than others; thus requiring additional wiring. To distinguish
one harness from another within the same family, harnesses are referred to as `levels’
and are identi® ed by a letter. Production requirements for each level vary according
to individual car sales. The harness assembly line must be ¯ exible enough to change
swiftly from one level to another as requirements change. Sometimes more than one
line is required to produce the total volume for a given family. Then, the volume is
divided into the number of lines necessary, but each manufacturing line must be
capable of building all the levels. When considerable di erences between levels exist,
and it is not feasible to build all of them on the same line, families are separated into
sub-families.
The typical medium-size harness family selected to serve as an example through-
out this paper has an average of 120 circuits per unit, ® ve di erent levels, and a total
of 17 subassemblies. Subassemblies considered for the present study include basically
the fabrication process of `splices’. This operation joins two or more circuits by one
end using ultrasonic heat into a single unit called `splice’. The intention of the splice
is to send an electrical signal to more than one place at the same time. The joined
area is covered afterwards by a vinyl tube. Figures 2a and 2b show a diagram of
a splice. The subassemblies are then incorporated into the harness at the assembly
line.
For manufacturing purposes, the 17 subassemblies are grouped in seven work-
stations. Some levels use all the subassemblies while other levels do not. The manu-
facturing time for each subassembly is di erent among levels. Table 1 shows time
values for subassembly fabrication for di erent levels. The time to complete the
subassembly operations was found to follow an exponential distribution pattern.
This table also shows the assembly stations mean times and the typical production
requirements for the di erent harness levels. These production requirements are
customer driven and normally change from one month to another.
Before World Class Manufacturing philosophy gained popularity among harness
manufacturers, most of their plants used a process-oriented layout to set the work
areas. That is, all the machines of one type were grouped in one area, and operated
under a batch production system. Typical unit lot sizes range from two to ten days of
production. Typically a component travels approximately 2500 ft before being incor-
porated into the ® nal assembly. Long manufacturing lead times and large ® nish
product inventories were standard practices. Once exposed to the bene® ts of a
JIT type of system, wire-harness manufacturers started to change their production
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
Assy.
work after Daily
Job Sub-assembly stations stat’ n assy load/line
# 1 2 3 4 5 6 7 avg avg pc %
Plant managers usually make decisions on what system to use based on common-
sense rather than on an analytical study of the alternatives. For instance, since plant
managers usually consider it a waste rather than an investment to replace the current
machinery with the more appropriate smaller capacity equipment, the o -line model
has become a very common approach. The present study tries to ® ll the void of lack
of analytical methods by presenting a method that the decision maker can use to
determine the best con® guration (on-line, o -line) and the corresponding parameters
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015 1998 F. Estrada et al.
(UL S and NPK). The study analyses the two arrangements mentioned above to
compare and contrast their e ectiveness under di erent scenarios.
The fundamental production parameters to be determined for a JIT manufactur-
ing system include:
the capacity of the assembly process ( ¹ 1 ), (given for the present study),
the capacity of the subassembly ( ¹ 2 ), (given for the present study),
the unit load size (UL S ), and lot size ( L S ) for subassemblies, and
the number of production kanbans for subassemblies (NPK).
These parameters constitute the essence of a pull system and the determination of
their values is one of the main objectives of this paper. L S is de® ned as the quantity
of parts produced every time a production kanban is received. UL S is the number of
parts moved as a unit between workstations. For the present study case UL S equals
L S. That is, every time a production kanban is received, a UL S is produced and
moved as a unit, when required.
2. Previous work
Several authors have investigated the theoretical and practical implications of a
JIT system. This section presents an overview of the relevant works in both areas.
Karmarkar et al. (1985 ), analysed the lot sizing e ect on the lead time perfor-
mance in a manufacturing cell. They described the use of a GPSS simulation model
and an analytical model that they called Q-LOTS. These models were used to ana-
lyse the e ect of the lot-size policy on a manufacturing cell at Eastman Kodak’s
Apparatus Division. A comparison of the results obtained from the Q-LOTS and the
simulation models was presented. They concluded that at large lot sizes, similar
results were obtained with both models. In contrast, when small lot sizes were
used the results obtained with both models did not correspond. They suggested
Evaluating JIT in the electric wire-harness industry 1999
that the models have potential uses in devising optimal lot size policies and in
predicting the performance of a facility for a given policy. An equally important
use of the model was to predict the lead times required to produce an order.
Mehra and Inman (1992), analysed the critical elements of JIT implementation.
Enough evidence supported the statement that a relationship between a successful
JIT application and a JIT production strategy existed. There were also enough data
to sustain evidence of a relationship between successful JIT application and a JIT
vendor strategy. As a conclusion, they said that a JIT production strategy is the most
critical factor, with the JIT vendor strategy showing less signi® cance. They recom-
mend that managers seeking to implement a JIT system should direct their e orts
and resources to the elements contained in the JIT production strategy. The present
study used their recommendation to concentrate on determining the design
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
parameters and production strategy that best suit a speci® c real life industrial
environment.
Mejabi and Wasserman (1992), presented a simulation model for JIT environ-
ments. Although the system was designed solely for demonstration purposes, and did
not re¯ ect more complex industrial applications, they made some valid recommen-
dations. The objective of their study was to enhance the ability of companies to
perform detailed analyses and detect the e ects of various JIT con® gurations on
manufacturing operations.
Lisy and LeBeau (1991) analysed the supply chain management for Eastman
Kodak Co. and presented a technique they call the Flowrate. The Flowrate
showed one supply demand relationship at a time and estimated how they ® tted
together. This was accomplished by using standard waste and conversion factors,
that is, the ratio of production divided by sales. Moving-averages of Flowrate ratios
were plotted on a run chart. The result was a time series visually measuring how well
production rates were synchronized with each other. Where rates were not synchro-
nized, this measure allowed one to evaluate alternate, more competitive, production
and inventory strategies.
TrevinÄo-Uribe (1986) explored the possibility of determining the relevant para-
meters for a JIT production system using an analytical or design approach. He
analysed the relationship between two workstations by using a customer± supplier
analogy. In this analogy, he identi® ed the assembly operation as the supplier and the
fabricating operation as the customer. His procedure is based on the elimination of
assembly interruptions due to component stockouts, the achievement of the assem-
bly daily requirements, and the minimization of an annual total cost. In other words,
design combinations that eliminate stockouts and achieve the frozen schedule were
identi® ed and the design combination that minimized the annual total cost function
was selected. His annual total cost function included the inventory carrying cost,
storage cost, production equipment cost, handling cost, operator and handler idling
cost, and stockout cost. Some parameters required by the cost elements of the design
procedure and the probability of stockout were estimated using a continuous-time
Markov chain model and probabilistic and queuing analysis. Simulation was used to
corroborate the validity of the estimate. He used two methods to solve the design
problem. The ® rst method enumerated feasible design combinations, using a series of
steps, and selected the optimal design. The second method was based on a non-linear
programming algorithm that computed the optimal design combination through a
search procedure. As an innovation, he introduced the concept of probability of
stock (PS ), that is, the probability that a workstation would run out of parts. He
2000 F. Estrada et al.
suggested that it could be used to identify those workstations with a high probability
of causing line stoppages.
meeting customer demands, i.e. stopping the car assembly plant, was considered so
high that the penalty associated was considered to be in® nite.
Every pair of assembly-subassembly stations was evaluated according to this
model, as described in the next section, considering the subassembly station as the
supplying process and the assembly station as the consuming process. The bu er
required between stations is determined by the product of the UL S times the NPK.
The determination of parameters for the on-line and o -line scenarios are pre-
sented next.
(2) The corresponding production capacities for the supplying process, ¹ 2 were
determined.
(3) Constraint three was evaluated. The values of ¹ 2 that complied with it were
determined.
(4) For each feasible combination of UL S-¹ 2 the minimum number of produc-
tion kanbans that satis® ed constraint (1) was determined.
MNA
PS = 1 -
å n= 0
Pn
where:
PS= probability of stockout
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
L SAL =
å n= 1
nPn
(1) For the feasible UL S and ¹ 2 , the minimum NPK that satis® ed PSC was
selected.
(2) For the feasible UL S and ¹ h , the minimum NPK that satis® ed PSC was
selected. Where ¹ h was the capacity of the handling process.
(3) The design combination (UL S , ¹ 2 , ¹ h , and NPK) that minimized the PS and
the WIP was selected.
(1986 ). They were evaluated using ® ve di erent UL S values (1, 27, 108, 215 and 430)
to determine the critical design decisions mentioned above. These UL S values repre-
sent the movement of a single piece at a time, one hour, four hours, half say and full
day production, respectively. This evaluation was performed for the on and o -line
strategies. Wire-harness manufacturers consider that the highest cost they can incur
is to shut down the car assembly plant due to a late delivery. This is because of the
high penalties imposed by the car manufacturers. Therefore the ability to comply
with the daily production requirement is the most important evaluation factor to
consider for developing a system of this type. From the results obtained above and
considering PS (which limits compliance to daily production) followed by bu er
requirements (NPK * UL S ) as the two most important factors, the working para-
meters for each workstation were selected. They were classi® ed as `best’ and `second
best’, so that two values could be used in the simulation program for sensitivity
analysis. Tables 2 and 3 show the results grouped at best and second best combina-
tions for the on-line and o -line models respectively. The seven subassembly-assem-
bly stations with their corresponding values for evaluation factors are shown. The
reader is also referred to Figs 3 and 4 to visualize the relationship between the
assembly-subassembly stations.
Tables 2 and 3 are interpreted as follows. For SA1-ASS Y 1 on Table 3 (o -line
model) the UL S recommended as the `best’ option or ® rst choice is 27. A total of six
NPK of size 27 are to be kept between the subassembly and the assembly stations to
avoid production stoppages. The total WIP necessary to keep between these two
stations may be calculated by multiplying the UL S times the NPK, which gives 162
pieces. With these UL S and NPK values this workstation has a PS of 0. 01. The
second best option for this station, has UL S = 27, NPK= 3, and a PS of 0. 06. For
sensitivity analysis purposes a UL S = 108 is also included.
The data obtained in this section is used to develop the simulation models. The
simulation models and the results obtained are described in the next section.
4. Simulation analysis
This section describes the simulation models that were used for the strategy
analysis. In § 3 the recommended design parameters for these arrangements were
calculated by using an analytical method. These parameters included the assembly
process capacity, the subassembly process capacity, the number of production
kanbans, and the unit load sizes for subassemblies. They were chosen based on
the capability to comply with the daily production requirement, the probability of
stockout, and the WIP inventory required.
Evaluating JIT in the electric wire-harness industry 2003
1 SA1-ASSY1 1 6 0. 01 27 6 0. 01 1 4 0. 06 27 3 0. 06
2 SA2-ASSY3 1 6 0. 01 27 5 0. 01 1 3 0. 06 27 3 0. 07
3 SA3-ASSY5 1 6 0. 01 27 6 0. 01 1 3 0. 06 27 3 0. 09
4 SA4-ASSY6 1 4 0. 01 27 4 0. 01 1 2 0. 04 27 2 0. 04
5 SA5-ASSY7 1 22 0. 01 27 11 0. 09 1 11 0. 09 27 5 0. 29
6 SA6-ASSY8 1 30 0. 01 27 6 0. 34 1 6 0. 34 27 4 0. 47
7 SA7-ASSY9 1 3 0. 01 27 3 0. 06 1 2 0. 02 27 2 0. 03
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
The objective of this section is twofold. First, it seeks corroboration that the
results obtained by the analytical procedure allowed the assembly process to meet
customer requirements. Second, it seeks evidence to recommend one strategy over
the other. A secondary objective is to obtain sensitivity information on the design
parameters. The experimental design used is presented in § 4.1 and the results
obtained in § 4.2.
The simulation package chosen for this research was ProModelPC. All simula-
tions were performed on a 486DX/66 IBM compatible PC.
4.1. Experimental design
The experimental design used to simulate the models corresponds to a fractional
factorial design identi® ed as 2 - . A ® xed-e ects model was used with a de® ning
5 1
relation for the design of I= ABCDE. This gives a design resolution type V. Table
4 shows the factors considered for the experimental design as well as the levels at
which the experiment was run. For the `number of kanbans’ factor, the levels are
identi® ed as best and second best. By that it is meant that for each subassembly
station the number of kanbans that yields the lowest PS and achieves the daily
requirements is used for the + level. The second best `number of kanbans’ is that
number that also achieves the daily requirements, but with a higher probability of
stockout. The UL S values are obtained from § 3. The `number of ASS Y stations’
factor is included to analyse the e ect caused to the assembly line when it is balanced
2004 F. Estrada et al.
No Factor - + - +
Production requirements
(7) Total subassembly time may be distributed among the subassembly stations.
To go from seven to ® ve stations the corresponding time from the eliminated
stations is equally distributed among the remaining ® ve.
(8) Only 5% of the pieces require rework at the quality control inspection
station.
Both models were run for a period of 320 simulated hours, which represents a full
month of production with two shifts per day. To reduce the e ect of initial condi-
tions, a warm-up period of 40 hours was used before the collection of statistical data.
Also to reduce this same e ect, the system started with all the kanbans full. Only the
assembly stations started without partial assemblies. That is, with the maximum
recommended number of kanbans located in front of each assembly station, but
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
without partial assemblies in the assembly stations. Tables 6 and 7 summarize the
results of the simulation experiment. To evaluate compliance to daily production
requirements, the models were allowed to run without material restrictions during
the complete simulation period. That is, raw material was provided constantly
throughout the simulation.
4.2. Results
From Tables 6 and 7, one can see that only one combination (number 13 in Table
7) of the experimental factors failed to meet the minimum production requirement.
This was expected since the calculated PS value associated with this combination was
known in advance. Additionally, by observing Table 3 which gives the analytical
results calculated in § 3, it can be seen that the combination failing to meet the
minimum production uses the second best NPK and UL S. This combination also
uses the highest number of assembly and subassembly stations and the highest
number of jobs. The simulation results in Tables 6 and 7 gives evidence of the
adequacy of the analytical method proposed in this paper for the determination of
the design parameters.
10 1 27 20 7 5
11 2 108 20 7 5 12 989. 57 540. 09 22. 86 1363. 58
12 1 108 20 7 1 13 630. 77 789. 43 26. 28 2097. 48
13 2 27 24 7 5 11 025. 17 28. 91 19. 3 471. 54
14 1 27 24 7 1 14 177. 14 356. 7 27. 38 918. 86
15 2 108 24 7 1 15 001. 86 523. 63 28. 9 1331. 08
16 1 108 24 7 5 14 218. 91 689. 92 24. 65 2084. 38
Additionally, analysis of variance for data in Tables 6 and 7 was performed for the
following performance measures: mean total throughput, mean lead time, mean
work in process and utilization rate for subassembly equipment.
Four performance measures are used in this study to evaluate the models. They
are: throughput, lead time, work in process (WIP), and subassembly process utiliza-
tion.
The mean total throughput is determined by,
å
m
k= 1 Jk
mean total throughput =
mr
Where Jk is the total number of complete pieces for the kth simulation, m is the
number of replicated simulations, and r is number of days simulated. To avoid
misinterpreting the results obtained, the throughput value is calculated in hours of
work content, and not in pieces.
The mean lead time is determined by,
å
m
Tik
k= 1
mean lead time =
mrqi
Where Tik is the total accumulated time spent in the system by part i for the kth
simulation, m is the number of replicated simulations, r is the number of days
simulated, and qi is total quantity of parts i that enter the system.
The mean work in process is determined by,
å
m
k= 1 W ik
mean wip =
mr
Where W ik is the total pieces contained in a queue station i for the kth simulation, m
is the number of replicated simulations, and r is number of days simulated.
Statistical tests of total mean throughput criteria showed the on-line strategy not
to be statistically di erent from that of the o -line strategy. This was expected since
Evaluating JIT in the electric wire-harness industry 2007
the critical parameters were designed as discussed in § 3, so that both models will
comply with the assembly requirements. The factors that seemed to have the greatest
in¯ uence were: number of assembly stations and number of jobs.
Results of hypothesis tests for lead-time suggested the on-line strategy to have a
superior performance to the o -line strategy. The factors that seemed to have the
greatest impact over this performance criterion were: number of kanbans and unit
load size. This was expected since the on-line strategy used smaller UL S’s than the
o -line stategy that had to wait longer to complete each UL S and be able to move
parts.
Statistical tests for the SA equipment utilization criterion showed the on-line
strategy to be statistically equal to that of the o -line strategy. The factors that
seemed to have the greatest impact on this criterion were the number of subassembly
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
conditions and the assumptions used in this paper. One assumption is that 5% of the
pieces require rework at the quality control inspection station. Two di erent oper-
ating times are used at this station to ® t this assumed situation. Though it is not very
far from reality, the exact scenario may be reproduced. The quality control area
o ers a wide opportunity for research in this environment.
Another assumption made in this paper is that repair and maintenance activities
were performed during non-productive hours. This could be modi® ed and include a
more precise scenario. One goal of this study was to develop a computer model that
can be used in future research. Improvements can be made to the model to make it
more versatile, to suit di erent operating conditions.
Also a more comprehensive study of the economic tradeo between an o -line
subassembly arrangement shared by two or more lines and an on-line arrangement
Downloaded by [University of California Santa Cruz] at 10:38 05 January 2015
may be investigated.
The present study answered some questions of interest to its authors. At the same
time, it also set the basis for calculating and comparing other scenarios that may be
of interest in the wire-harness manufacturing industry.
References
Harrell , C. R., Bateman, R. E., Gogg Thomas, J. and Mott , J. R. A., 1992, System
Improvement Using Simulation (Orem Utah: JMI Consulting Group and
PROMODEL Corporation).
Karmark ar , U. S., Kekre, S., Kekre, S., Freeman , S. 1985, Lot sizing and lead-time
performance in a manufacturing cell. Interfaces, 15 (2), 1± 9.
Law , A. M. and Kelton , W. D., 1982, Simulation Modeling Analysis, (New York: McGraw-
Hill).
Lisy , G. and LeBeau , J. M., 1991, Flow rate process: how to improve supply chain manage-
ment. American Production and Inventory Control Society, 34th International
Conference Proceedings, pp. 580± 582, October 20± 25, Seattle, WA.
Ma rtin , N., 1993, Fast track. Automotive Industries, September.
Mehra , S. and Inman , R. A., 1992, Determining the critical elements of Just-In-Time imple-
mentation, Decision Sciences, 23, 160± 174.
Meja bi, O. and Wasserman , G. S., 1992, Simulation constructs for JIT modelling.
International Journal of Production Research, 30 (5), 1119± 1135.
ProModel PC, User Manual, Version 5.01, 1991, Promodel Co., Provo, Utah.
Su z a ki, K., 1987, The New Manufacturing Challenge. Techniques for Continuous Improvement
(New York: Free Press).
Tompkins, J. and God da rd , W., 1992, Strike at GM points up JIT’s risks rewards. Modern
Materials Handling, November, pp. 14± 15.
Trev in˜o-Uribe, J. , 1986, Design Procedures for Pull Production System. PhD thesis, Georgia
Institute of Technology.