Sunteți pe pagina 1din 20

European Journal of Operational Research 183 (2007) 674–693

www.elsevier.com/locate/ejor

Production, Manufacturing and Logistics

A classification of assembly line balancing problems


a,*
Nils Boysen , Malte Fliedner a, Armin Scholl b

a
Universität Hamburg, Von-Melle-Park 5, D-20146 Hamburg, Germany
b
Friedrich-Schiller-Universität Jena, Carl-Zeiß-Straße 3, D-07743 Jena, Germany

Received 27 February 2006; accepted 20 October 2006


Available online 14 December 2006

Abstract

Assembly lines are special flow-line production systems which are of great importance in the industrial production of
high quantity standardized commodities. Recently, assembly lines even gained importance in low volume production of
customized products (mass-customization). Due to high capital requirements when installing or redesigning a line, its con-
figuration planning is of great relevance for practitioners. Accordingly, this attracted attention of many researchers, who
tried to support real-world configuration planning by suited optimization models (assembly line balancing problems). In
spite of the enormous academic effort in assembly line balancing, there remains a considerable gap between requirements
of real configuration problems and the status of research. To ease communication between researchers and practitioners,
we provide a classification scheme of assembly line balancing. This is a valuable step in identifying remaining research chal-
lenges which might contribute to closing the gap.
 2006 Elsevier B.V. All rights reserved.

Keywords: Configuration of assembly lines; Assembly line balancing; Classification

1. Introduction Originally, assembly lines were developed for a


cost efficient mass-production of standardized prod-
An assembly line is a flow-oriented production ucts, designed to exploit a high specialization of
system where the productive units performing the labour and the associated learning effects (Shtub
operations, referred to as stations, are aligned in a and Dar-El, 1989; Scholl, 1999, p. 2). Since the
serial manner. The workpieces visit stations succes- times of Henry Ford and the famous model-T, how-
sively as they are moved along the line usually by ever, product requirements and thereby the require-
some kind of transportation system, e.g., a conveyor ments of production systems have changed
belt. dramatically. In order to respond to diversified cos-
tumer needs, companies have to allow for an indi-
vidualisation of their products. For example,
*
German car manufacturer BMW offers a catalogue
Corresponding author. Tel.: +49 40 42838 4640.
E-mail addresses: boysen@econ.uni-hamburg.de (N. Boysen),
of optional features which, theoretically, results in
fliedner@econ.uni-hamburg.de (M. Fliedner), a.scholl@wiwi. 1032 different models (Meyr, 2004). Multi-purpose
uni-jena.de (A. Scholl). machines with automated tool swaps allow for

0377-2217/$ - see front matter  2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejor.2006.10.010
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 675

facultative production sequences of varying models Three theoretical reasons might explain the
at negligible setup costs. This makes efficient flow- aforementioned deficit: (i) Researchers have not
line systems available for low volume assembly- considered the ‘‘true’’ real-world problems so far.
to-order production (Mather, 1989) and enables (ii) The problems were covered, but could not be
modern production strategies like mass-customiza- solved to satisfaction. (iii) Scientific results could
tion (Pine, 1993), which in turn ensures that the not be transferred back to practical applications,
thorough planning and implementation of assembly e.g., because solutions for special case studies could
systems will remain of high practical relevance in not be extended to general problems.
the foreseeable future. Any of these reasons might result from a funda-
Due to the high level of automation, assembly mental problem in communication, which is
systems are associated with considerable investment expressed by an inconsistent use of terms and defini-
costs. Therefore, the (re)-configuration of an assem- tions for the various types of balancing problems
bly line is of critical importance for implementing a (Becker and Scholl, 2006). This is not only impeding
cost efficient production system. Configuration the communication within the research community,
planning generally comprises all tasks and decisions but also the knowledge transfer to practice.
which are related to equipping and aligning the pro- A first, yet decisive step to resolve this problem
ductive units for a given production process, before lies in a consistent, authoritative classification of
the actual assembly can start. This includes setting ALB problems including all relevant constraints
the system capacity (cycle time, number of stations, and objectives. A uniform classification enables
station equipment) as well as assigning the work practitioners to compare their individual problem
content to productive units (task assignment, settings with those covered by research and to single
sequence of operations). out suitable solution techniques. Furthermore,
In light of the high practical relevance, it is not future research challenges can be identified by struc-
astounding that a massive body of academic litera- turing the existing body of literature according to
ture covers configuration planning of assembly sys- the classification scheme. The primary aim of this
tems. In the scientific discussion, the term assembly article is to develop such a classification.
line balancing (ALB) is used to subsume optimiza- Apparently, the existing distinction of SALB and
tion models which seek to support this decision pro- GALB introduced by Baybars (1986) has become
cess. Since the first mathematical formalization of insufficient to reflect the heterogeneity of GALB
ALB by Salveson (1955), academic work mainly problems. Especially now, when the problem struc-
focused on the core problem of the configuration, ture of SALB is well examined and powerful solu-
which is the assignment of tasks to stations. Because tion techniques exist, it is to be expected that
of the numerous simplifying assumptions underly- future publications will mainly focus on GALB
ing this basic problem, this field of research was problems, some of which might show a similar
labeled simple assembly line balancing (SALB) in problem structure as SALB, whereas others will
the widely accepted review of Baybars (1986). Sub- deviate considerably. Ghosh and Gagnon (1989)
sequent works however, more and more attempted extended Baybars efforts by further distinguishing
to extend the problem by integrating practice rele- special characteristics of GALB problems which
vant aspects, like u-shaped lines, parallel stations had been covered by academic work at that time.
or processing alternatives (Becker and Scholl, However, their list lacks a systematic approach
2006). In spite of these efforts, which are referred and is long outdated because a great variety of addi-
to as general assembly line balancing (GALB), there tional constraints has been introduced in the mean-
seems to be a wide gap between the academic discus- time. Several researchers have tried to overcome this
sion and practical applications. Empirical surveys weakness by developing individual names for their
stemming from the 1970s (Chase, 1974) and 1980s considered problem extensions, which were mostly
(Schöniger and Spingler, 1989) revealed that only oriented towards the existing nomenclature.
a very small percentage of companies were using a Although these efforts certainly help experts in the
mathematical algorithm for configuration planning field to recognize particular problem characteristics,
at that time. The apparent lack of more recent scien- it might also lead to confusion as long as no guide-
tific studies on the application of ALB algorithms lines exist as to when an extension is different
indicates that this gap still exists or even has enough in order to receive a new label or when
widened. two extensions with individual names are to be
676 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

combined. In any case, this policy hardly reveals ancing) the assembly work among all stations with
relations between problems and therefore cannot respect to some objective. For this purpose, the total
replace a classification in structuring the literature. amount of work necessary to assemble a workpiece
Before a new classification is proposed, the sub- is split up into a set V = {1, . . ., n} of elementary
ject to be classified has to be characterized unambig- operations named tasks. Tasks are indivisible units
uously. Therefore, Section 2 will start out with a of work and thus each task j is associated with a
description of the core problem of ALB, which will processing time tj also referred to as task time.
be used as a basis for the classification scheme pre- Due to technological and/or organizational require-
sented in Section 3. Then, the classification scheme ments, tasks cannot be carried out in an arbitrary
is applied to structure the existing literature in Sec- sequence, but are subject to precedence constraints.
tion 4. Section 5 provides a first interpretation of The general input parameters of any SALB
achieved results and aims at providing hints on instance can be conveniently summarized and visu-
how to close the gap between research and practice alized by a precedence graph. This graph contains
by identifying promising areas of future research a node for each task, node weights which equal
and characterizing practice relevant problem exten- the task times and arcs reflecting direct as well as
sions which have not been covered so far. paths reflecting indirect precedence constraints.
Fig. 1 shows an example precedence graph with
2. A definition of assembly line balancing (ALB) n = 9 tasks having task times between 2 and 9 (time
units).
In the following, we define the scope of the clas- A feasible line balance, i.e., an assignment of
sification. First we characterize the SALB problem tasks to stations, has to ensure that no precedence
as the core decision problem in configuration plan- relationship is violated. The set Sk of tasks assigned
ning in its very basic version. Afterwards, the basic to a station k constitutes its station load orP
work con-
assumptions of SALB are examined of how they tent, the cumulated task time tðS k Þ ¼ j2S k tj is
have to be adopted for setting the more general called station time.
assumptions of ALB. That way, a definition of SALB further assumes that the cycle time of all
ALB, the field to be classified, can be derived. stations is equal to the same value c. Assembly lines
Among the family of ALB problems, the best- with this attribute are called paced, as all stations
known and best-studied is certainly the SALB prob- can begin with their operations at the same point
lem. Although it might be far too constrained to in time and also pass on workpieces at the same
reflect the complexity of real-world line balancing, rate. As a consequence, all station times of a feasible
it nevertheless captures its main aspects and is right- balance may never exceed c, as otherwise the
fully regarded as the core problem of ALB. In fact, required operations could not be completed before
vast varieties of more general problems are direct the workpiece leaves the station. Station times can
SALB extensions or at least require the solution of however be smaller than the cycle time, in which
SALB instances in some form. In any case, it is well case a station k has an unproductive idle time of
suited to explain the basic principles of ALB and c  t(Sk) time units in each cycle.
introduce its relevant terms. A comprehensive For the example of Fig. 1, a feasible line balance
review of SALB and its solution procedures is pro- with cycle time c = 11 and m = 5 stations is given by
vided by Scholl and Becker (2006). station loads S1 = {1, 3}, S2 = {2, 4}, S3 = {5, 6},
According to the underlying concept of any S4 = {7, 8} and S5 = {9}.
SALB formulation, an assembly line consists of In order to ensure high productivity, any good
k = 1, . . ., m (work) stations arranged along a con- balance should cause as few idle times as possible.
veyor belt or a similar mechanical material handling
device. The workpieces (jobs) are consecutively
launched down the line and are hence moved on
6 6 4
from station to station until they reach the end of 1 2 7
the line. A certain set of operations is performed 2 9
repeatedly on any workpiece which enters a station, 8 9
whereby the time span between two entries is 4 5 4 5
3 4 5 6
referred to as cycle time. In general, the line balanc-
ing problem consists of optimally partitioning (bal- Fig. 1. Precedence graph.
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 677

With regard to the objective function considered, real-life, the same task might be performed in dif-
SALB problems can further be distinguished (cf. ferent modes and/or with different resources.
Scholl, 1999, ch. 2.2) into four types: For a given More powerful machinery might for instance be
cycle time c, minimizing the sum of station idle purchased at higher cost, so that a time-cost
times is equal to minimizing the number of opened trade-off influences the balancing decision. Under
stations. SALB problems under this objective are the limiting assumptions that processing modes
called SALBP-1. Conversely, if the number of sta- are fixed for any task and each station is equally
tions is given, then minimizing the cycle time guar- equipped an explicit consideration of resources
antees minimum idle times, which is known as (machine, men, material and/or tools) is not nec-
SALBP-2. If both, number of stations and the cycle essary. As a consequence the decision problem
time, can be altered, the line efficiency E is used to of SALB comes down to assigning tasks to sta-
determine the quality of a balance. The line effi- tions. As soon as assumptions (S-2) and (S-9)
ciency corresponds to the productive fraction of have to be given up, line balancing needs also to
the line’s total operating time tsum and is typically decide on an assignment of required resources to
defined as E = tsum/(m Æ c). As the total idle time is stations along with the respective tasks. At the
equal to tsum  (m Æ c), a maximization of E also same time capacity-oriented objectives of SALB
minimizes idle times. The corresponding SALB become insufficient as resource costs need to be
problem was hence labeled SALBP-E. Finally, the considered.
problem of finding a feasible balance for a given In order to account for the large variety of ALB
number of stations and a given cycle time is known problems discussed in the literature, some SALB
as SALBP-F. assumptions need to be modified whereas others
As already mentioned above, all four types of are to be given up completely. The following set
SALB are based on a set of limiting assumptions of assumptions (A-1) to (A-5) is considered to be
(c.f. Baybars, 1986; Scholl, 1999, ch. 2.2; Scholl constitutive for ALB in general (SALB and other
and Becker, 2006): problem versions as well are contained as special
cases with additional assumptions):
(S-1) Mass-production of one homogeneous
product. (S-1) ! (A-1) The products to be manufactured
(S-2) All tasks are processed in a predetermined (one or more) are known with certainty.
mode (no processing alternatives exist). (S-2) ! (A-2) A set of processing alternatives is
(S-3) Paced line with a fixed common cycle time given.
according to a desired output quantity. (S-3) ! (A-3) The line is to be configured such
(S-4) The line is considered to be serial with no that target production quantities are satisfied
feeder lines or parallel elements. for a certain planning horizon. This might be
(S-5) The processing sequence of tasks is subject realized by setting the (average) cycle time(s)
to precedence restrictions. and, thus, production rate(s) or by seeking to
(S-6) Deterministic (and integral) task times tj. produce as much as possible if maximum sales
(S-7) No assignment restrictions of tasks besides are not a limiting factor.
precedence constraints. (S-4) ! (A-4) The line flow is unidirectional.
(S-8) A task cannot be split among two or more (S-5) = (A-5) The processing sequence of tasks is
stations. subject to precedence restrictions.
(S-9) All stations are equally equipped with (S-6) – (S-9) are given up.
respect to machines and workers.
Finally, this leads to the following definition of
Nearly all of these assumptions have been ALB. All optimization models which aim at sup-
relaxed or somehow modified by various model porting that part of assembly line configuration
extensions considered in the literature. This poses which deals with grouping tasks and their required
the question which, if any, of the assumptions of resources to stations and are in line with the pre-
SALB is also constitutive for ALB in general. sented five assumptions (A-1)–(A-5) are considered
With regard to real-world line balancing, one to be assembly line balancing problems. All of those
of the strongest simplifications of SALB is cer- approaches are hence the subject of this work and
tainly caused by assumptions (S-2) and (S-9). In will be included in the presented classification.
678 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

3. The classification scheme for ALB bined arbitrarily. As all attribute values are chosen
such that they are unique, it is not necessary to spec-
In the following, we present a scheme for classi- ify the attribute designators within the tuples.
fying the research field of ALB based on the distinc-
tion made in Section 2. Because there are very recent 3.1. Precedence graph characteristics
surveys on SALB and GALB research (cf. Scholl
and Becker, 2006; Becker and Scholl, 2006) we A precedence graph consists of nodes which rep-
restrict the presentation to very short descriptions resent the tasks of the production process and a set
of the relevant characteristics of assembly systems of arcs which represent the precedence relations
and ALB problems. For further descriptions we between the tasks. Moreover, node and arc weights
refer to, e.g., Buxey et al. (1973), Baybars (1986), can be considered which reflect important attri-
Shtub and Dar-El (1989), Ghosh and Gagnon butes, like task times and processing alternatives,
(1989), Erel and Sarin (1998), Scholl (1999, ch. 1), resource requirements and zoning restrictions.
and Rekiek et al. (2002b). These precedence graph characteristics a are repre-
The classification scheme is constructed as sented by the six attributes a1 to a6:
follows:
3.1.1. Product specific precedence graphs:
• The assumptions of SALB as outlined in Section a1 2 {mix, mult, }
2 are chosen as the common reference for classi- This attribute determines whether only a single
fying problem characteristics. That means, if not product and, thus, a single precedence graph is con-
further stated it is supposed that the SALB sidered or several ones have to be taken into
assumptions apply, so that only deviations from account simultaneously. However, for the line bal-
SALB are explicitly provided. ancing problem it is not the actual number of differ-
• The basic classification scheme has been adopted ent products which is decisive, but the degree of
from the widely accepted and successful classifi- homogeneity among precedence graphs.
cation scheme for machine scheduling of Graham
et al. (1979). Another scheme which successfully a1 = mix. Mixed-model line: Varying models are
helped structuring a complex research field is manufactured on the same production system,
the one provided by Brucker et al. (1999) for pro- the production processes of which are similar
ject scheduling, which also employs the tuple- enough so that setup times are not present or
notation. negligible. Thus, the units of the different models
• As has been established in Section 2, any ALB are produced in an arbitrarily intermixed
problem will at least consist of three basic ele- sequence (cf. Scholl, 1999, p. 7). The balancing
ments: A precedence graph which comprises all problem is often based on an average model-
tasks and resources to be assigned, the stations mix. In order to anticipate the later sequencing
which make up the line and to which those tasks problem (cf. Yano and Bolat, 1989) adequately,
are assigned and some kind of objective to be a horizontal balancing objective is usually utilized
optimized. Accordingly, the presented classifica- which attempts to equalize the work content of
tion will be based on those three elements which stations over all models (cf. Scholl, 1999, ch.
are noted as tuple [ajbjc], where: 3.3 and 3.4; Boysen, 2005, ch. B.2).
a precedence graph characteristics; a1 = mult. Multi-model line: Different products
b station and line characteristics; are manufactured in batches. Whenever another
c objectives. batch is to be processed, a setup occurs which
requires time and resources. Therefore, the
One major advantage of the tuple-notation is that ALB problem is not only connected to a (batch)
any default value, represented by the symbol , can sequencing but also to a lot sizing problem (cf.,
be skipped when a tuple is actually specified. As e.g., Burns and Daganzo, 1987; Dobson and
explained above, the default values of all attributes Yano, 1994). However, both additional problems
constitute the SALB problem. In the following nota- are typically not part of the long/medium-term
tion, the symbol * always indicates that for the configuration decisions.
respective attribute the alternative values (except a1 = . A single product is manufactured or the
for ) do not exclude each other and can be com- production processes of multiple products are
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 679

(almost) identical so that they need not be 3.1.5. Assignment restrictions:


distinguished. a5 2 {link, inc, cum, fix, excl, type, min, max, }*
Besides precedence relations, assignment restric-
tions might affect the grouping of tasks to station
3.1.2. Structure of the precedence graph: loads by forcing or forbidding certain combi-
a2 2 {spec, } nations.
Some research papers restrict themselves to pre-
cedence graphs with special structures, mainly for a5 = link. Subsets of tasks are linked such that
developing more efficient specialized algorithms. these tasks must be assigned to the same station,
e.g., because they employ the same resource
a2 = spec. The research is restricted to prece- which cannot be duplicated.
dence graphs with some kind of special structure, a5 = inc. Subsets of tasks are incompatible and
e.g., linear, diverging or converging graphs. must not be assigned to the same station, e.g.,
a2 = . The precedence graph can have any acy- because the tasks disturb each other (drilling
clic structure. and measuring) or require the workpiece in
incompatible positions.
a5 = cum. The assignment of tasks to a station is
3.1.3. Processing times: a3 2 {tsto, tdy, }* subject to constraints on the cumulated value of
In reality, task times can vary in time, for particular task attributes (e.g., restricted space
instance due to complicated manual operations or for storing material).
defaults of machinery. ALB models can consider a5 = fix. Some tasks can only be assigned to par-
this phenomenon in different ways. ticular stations, e.g., a required resource cannot
be moved when reconfiguring a line.
a3 = tsto. Stochastic processing times: It is a5 = excl. Some tasks must not be assigned to
assumed of processing times follow a known (or particular stations, e.g., because needed resources
even unknown/partially known) distribution cannot be installed there.
function. a5 = type. Some tasks have to be assigned to a
a3 = tdy. Dynamic variations of processing times: station from a certain type set, e.g., for working
These variations are, e.g., due to learning effects underneath a workpiece.
of operators. a5 = min. When assigning a task, minimum dis-
a3 = . Processing times are considered to be sta- tances to other tasks measured in time, space or
tic and deterministic. sequence positions have to be observed, e.g., for
drying the paint.
a5 = max. Maximum distances between tasks
3.1.4. Sequence-dependent task time increments: have to be observed, e.g., because a required
a4 2 {Dtdir, Dtind, }* physical condition can only be maintained for a
Such task time increments occur whenever the short time.
sequence of operations influences their processing a5 = . No assignment restrictions are con-
times. sidered.

a4 = Dtdir. If two tasks are executed at a station


one directly after the other, additional time might 3.1.6. Processing alternatives: a6 2 {pak, }
be required for setup operations or tool changes If processing alternatives exist, the production
(Wilhelm, 1999) and repositioning of workpieces process and, hence the precedence graph is subject
(Arcus, 1966; Bautista and Pereira, 2002). to change, so that an additional decision problem
a4 = Dtind. Indirect sequence-dependent time arises, which is to select processing alternatives out
increments occur if the status achieved by complet- of the set available (cf. Pinto et al., 1983).
ing particular tasks has an effect on the processing
time of other tasks which are executed later in the a6 = pak. Processing alternatives can be distin-
same or another station (Scholl et al., 2006). guished according to the effects they have on
a4 = . Sequence-dependent time increments are the precedence graph by defining k 2 {, prec,
not considered. subgraph}:
680 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

k = : Processing alternatives only deviate in # = div: (Local) cycle times diverge between
processing times and costs (time-cost trade-off ). stations or models.
k = prec: The chosen alternative not only b1 = unpack. An unpaced line is not strictly
affects times and costs, but also precedence rela- restricted by a cycle time. Instead, it advances
tionships between tasks. For example, an alterna- when stations have completed their tasks; with
tive placement of one component at the k 2 {, sync}.
workpiece might change the precedence relations k = : Asynchronous workpiece movement: As
for installing another. soon as a station completes its work, the workpiece
k = subgraph: Processing alternatives alter is moved to the next station or a buffer in front of
complete parts of the production process, so that this station unless blocking occurs. The ALB is
whole subgraphs are substitutable. This might, accompanied by the additional decision problem
e.g., occur whenever a set of options can either of positioning and dimensioning buffers (e.g., Baker
be installed separately or completely be replaced et al., 1990; Hillier et al., 1993; Malakooti, 1994).
by a purchased module. k = sync: In synchronous lines, the movement
a6 = . Processing alternatives are not con- of workpieces is coordinated between stations.
sidered. The workpieces are transferred to the respective
next station when all stations have completed
their current workpiece.
Remark. Whenever processing alternatives are
regarded, the aforementioned precedence graph 3.2.2. Line layout: b2 2 {, uk}
characteristics (except for a1 and a6) might be
related to alternatives and tasks rather than tasks b2 = . The stations are arranged in a serial man-
only. This can be visualized by attaching pa to the ner along the flow of the line.
respective attribute. b2 = uk. A U-shaped line layout with crossover
stations is used (cf. Miltenburg and Wijngaard,
1994; Monden, 1998; Scholl and Klein, 1999);
3.2. Station and line characteristics with k 2 {, n}:
k = : The line forms a single U.
Stations and their arrangement in the assembly k = n: An n-U line composed of multiple
line can be classified using the six attributes b1 to U-shaped segments is considered (cf. Miltenburg,
b6 of the second tuple entry b: 1998; Sparling, 1998).

3.2.1. Movement of workpieces: b1 2 {k#, unpack}


Remark. If other special layouts will be introduced,
the classification can easily be extended.
b1 = k#. In paced lines, a cycle time restricts
the station time at each workstation with k 2
{, each, prob} and # 2 {, div}: 3.2.3. Parallelization:
k = : The (average) work content per cycle of b3 2 {plinek, pstatk, ptaskk, pworkk, }*
each station over all workpieces is restricted by
the cycle time. While this restriction is strict in b3 = plinek. More than one parallel line is to be
single-model production, it has just to be fulfilled balanced or the number of lines installed is part
on average in the case of mixed-model produc- of the decision problem (cf. Geoffrion and
tion, some type of parallelization and/or stochas- Graves, 1976).
tic task times. b3 = pstatk. When stations are parallelized, their
k = each: Each single model must strictly fulfill resources and work contents are duplicated so
the cycle time restriction in mixed- or multi- that they process all assigned tasks alternately
model production. (cf. Buxey, 1974; Pinto et al., 1981; Sarker and
k = prob: The cycle time restriction is obeyed Shanthikumar, 1983; Bard, 1989).
with a given probability (i.e., stochastic task b3 = ptaskk. A parallelized task is assigned to
times) or proportion (i.e., uncertain model-mix). more than one station. In addition to their regu-
# = : All stations and models have to regard lar work content, stations process the parallelized
the same global cycle time. task interchangeably (cf. Pinto et al., 1975;
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 681

Bukchin et al., 2002; Bukchin and Rabinowitch, k = 01: If more than one task can be per-
2005). formed on the same resource (a tool or
b3 = pworkk. Several working places simulta- machine), investment cost is reduced if those
neously work on the same workpiece at different tasks are assigned to the same station, because
mounting places such that they do not obstruct the resource needs to be installed only once.
each other (Falkenauer, 2005). The balancing That is, a 0-1 investment decision has to be made
problem is connected to a detailed scheduling for each station-resource combination (install if
problem which has to avoid that different opera- any task requires the resource, do not install
tors interfere with each other. A special case of else).
a line with parallel workplaces is the two-sided k = max: If tasks differ with respect to
line where each station consists of a left-hand- resource quality (concerning, e.g., speed,
side and a right-hand-side workplace (Bartholdi, capabilities, qualification), they require the
1993; Kim et al., 2000a; Lee et al., 2001). resources to be selected such that they fulfill
b3 = . Neither type of parallelization is the maximum demand level of assigned tasks.
considered. For example, if operators require a specific
qualification to perform (difficult) tasks, the
most challenging task assigned to a station
Remark. If a certain fixed or maximum level of
defines the necessary qualification level of the
parallelization is constitutive for the problem, this is
operator and, hence, the wage costs to be paid
visualized by the superscript k 2 {2, 3,. . .}. Such a
at that station.
constraint might be due to technological or space
k = : Resources are modeled explicitly to
restrictions. For example, the stations of a two-
account for another type of synergy and/or
sided line consist of two parallel working places, i.e.,
dependency.
b3 = pwork2. If the level of parallelization is uncon-
b4 = . Resources are not considered explicitly
strained or can be set to an arbitrary value by the
but might influence the balancing decisions via
planner, k =  is used.
assignment restrictions, processing alternatives,
etc. implicitly.
3.2.4. Resource assignment:b4 2 {equip, resk, }*
Typically, multiple resources, like operators,
machines and tools which provide the skills and/or Remark. If resources are modeled explicitly, a part
technological capabilities, forming the station of the task-station assignment restrictions (a5) might
equipment, are necessary to perform tasks at be replaced by task-resource and resource-station
stations. assignment restrictions.

b4 = equip. For each station exactly one equip-


ment is to be chosen out of a set of prespecified 3.2.5. Station-dependent time increments:
equipment alternatives. So, the balancing problem b5 2 {Dtunp, }
is connected to an equipment selection problem b5 = Dtunp. Some part of the station time is con-
which simultaneously has to decide on the equip- sumed by unproductive activities, e.g., workpiece
ment to be installed in each station. In the litera- transportation (Bard, 1989), walking distances to
ture this combined problem is referred to as the change sides in a U-line (Sparling, 1998; Sparling
assembly line design problem (Baybars, 1986). and Miltenburg, 1998) or workers’ return to the
b4 = resk. Instead of selecting a single equipment beginning of a station at the end of the cycle.
out of a set of predetermined equipment alterna- b5 = . Station-dependent time increments are
tives, a stations’ equipment is configured along not regarded.
with the task assignment. If several tasks require
the same resources, synergies can be realized by
combining these tasks into the same station 3.2.6. Additional aspects of line configuration:
loads, because the resources are needed only once b6 2 {buffer, feeder, mat, change, }*
and investment costs are at a minimum (cf. Boy- Depending on the production system, additional
sen, 2005, pp. 105–108). Two basic types of syn- technical requirements might need to be considered
ergies are denoted with k 2 {, 01, max}*: for balancing the line:
682 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

b6 = buffer. Buffer storages are required and have natives exist. If each station is equipped with a
to be allocated and dimensioned. worker who can perform any task, wage costs
b6 = feeder. One or more feeder lines which might be directly assigned to stations. An in-
flow into a main line necessitate a simultaneous depth discussion of cost factors in ALB is pro-
coordination of task assignments and cycle vided by Steffen (1977) and Amen (1997), Amen
time (Hautsch et al., 1972; Lapierre and Ruiz, (2000a) Amen (2000b), Amen (2001), Amen
2004). (2006).
b6 = mat. Boxes which contain the required c = Pr. The profit, which is defined as the
material need to be positioned and dimensioned difference between the revenue (which depends
(cf. Bukchin and Meller, 2005). on the production rate and, hence, the cycle
b6 = change. If certain tasks require the work- time) and the costs, is maximized. A further
piece to be in a particular position (lifted, tilted, distinction shall not be provided because of
etc.), a decision has to be made, whether posi- the aforementioned issue of cost type assign-
tions are fixed within stations or special machines ment.
are assigned to stations which allow for a change c = SSLk. Instead or in addition to the objectives
of positions. mentioned above, the station times should be
b6 = . No additional aspects of line configura- smoothed. Two ways of smoothing are given by
tion are regarded. k 2 {stat, line}:
k = stat: In mixed-model production the vary-
ing station times caused by different models are
3.3. Objectives to be smoothed (horizontal balancing; Merengo
et al., 1999).
Finally, the optimization of ALB will be guided k = line: Station times are balanced over all
by some objective which evaluates solutions. In stations of the line (vertical balancing; Rachama-
the case of multi-objective optimization more than dugu and Talbot, 1991).
a single objective can be selected out of the set c = score. The objective is to minimize or maxi-
c 2 {m, c, E, Co, Pr, SSLk, score, }*. mize some composite score which is related to
one or more attributes describing bottleneck
c = m. Minimize the number of stations m sub- aspects or further measures of efficiency, e.g.,
ject to a given output target for a certain plan- required grip strengths, quality or number of
ning horizon (specified by the cycle time c or workpiece position changes.
the production rate). In case that parallel stations c = . No objective function is required, only fea-
or workplaces are included in m, the number of sible solutions are searched for.
sequential stages might be restricted (maximal
line length).
c = c. Minimize the cycle time c or, equivalently,
maximize the production rate subject to a given 4. Classifying the literature
number of stations m.
c = E. Maximize the line efficiency E (productive With this scheme on hand, we classify that part
part of the line capacity) such that restrictions on of the ALB literature which deals with GALB prob-
the production rate and/or the number of sta- lems. For the versions of SALB, we only specify the
tions are fulfilled. tuple-notation, because there is no need to structure
c = Co. Cost minimization for a given output this field of research (instead, we refer to Scholl and
target. A classification of cost factors shall not Becker, 2006):
be provided as cost types can be assigned to
various elements of the assembly system, like SALBP-F: [j j] SALBP-1: [j j m]
tasks, stations, processing alternatives and/or SALBP-2: [j j c] SALBP-E: [j j E]
resources, depending on the considered model
formulation. Wage costs might, e.g., depend on The following table assigns the tuple-notation to
the selected processing alternative, but can also each relevant publication and its contribution to the
be directly assigned to a task as long as no alter- research field of GALB. If more than one problem
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 683

M mathematical B bound computation version is treated in a paper the most complex tuple-
model notation is reported. Further we distinguish
SA sensitivity/ HS heuristic start between the following contributions made by the
stability procedure for initial respective publication:
analysis solution, mostly
priority rule based
HI heuristic
Remark. The following list does not include the
improvement
numerous papers which treat related decision
procedure
problems of configuration planning which assume
Exact solution Meta-heuristics:
a given line balance. For example, a massive
procedures:
body of literature covers the buffer allocation
DP dynamic TS tabu search
problem in unpaced asynchronous lines (see the
programming
review papers of Dallery and Gershwin, 1992;
B&B branch & GA genetic algorithm
Papadopoulos and Heavey, 1996; Gershwin,
bound
2000). Since a change in station loads typically
GR graph SA simulated annealing
affects optimal buffer allocation considerably, a
theoretic
simultaneous optimization would be desirable to
approach
identify the systems’ global optimum. However,
B&C branch & cut ANT ant system approach
such simultaneous approaches are currently not
CG column GRASP greedy randomized
on-hand.
generation adaptive search
procedure
ENU enumeration
approach

Publication Notation Contribution


Aase et al. (2003), Aase et al. (2004) [j u j m] M, B, B&B
Agnetis et al. (1995) [spec, inc, fix j j SSLline] DP
Agrawal (1985) [j j score] HS
Ajenblit and Wainwright (1998) [j u j m, SSLline] GA
Akagi et al. (1983) [j pwork j] HS
Amen (1997), Amen (2000a), [j resmax j Co] M, B, B&B, HS
Amen (2000b), Amen (2001),
Amen (2006)
Arcus (1966) [mix, Dtdir, cum, fix j resmax, Dtunp, pwork j E] HS
Askin and Zhou (1997) [mix j pstat j Co] M, B, HS
Bard (1989) [j pstat, ptask, Dtunp j Co] DP
Bartholdi (1993) [fix, type j pwork2 j m] HS
Bautista and Pereira (2002) [inc, Dtdir j j m, score] ANT
Bautista and Pereira (2006) [cum j j m, c, score] M, HI, ANT
Bautista et al. (2000) [inc j j m, score] HS, GRASP
Baykasoglu and Özbakir (2006) [tsto j prob, u j m] GA
Berger et al. (1992) [spec j u j m] B, B&B
Bhattacharjee and Sahu (1987) [link, inc, fix j station j m, SSLline] HS
Boysen and Fliedner (2006) [tsto, link, inc, cum, pa j u, pstat, ptask, res01, HS, GR
resmax j Pr]
Bukchin and Rabinowitch (2005) [mix j div, ptask, res01 j Co] M, B, B&B
Bukchin and Rubinovitz (2003) [pa j pstat, equip j Co] M, B, B&B
Bukchin and Tzur (2000) [pa j equip j Co] M, B, B&B
(continued on next page)
684 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

Table (continued)
Publication Notation Contribution
Bukchin et al. (1997), [j pwork j m, score] B, B&B
Bukchin and Masin (2004)
Bukchin et al. (2002) [mix j ptask j score] M, HS, HI
Buxey (1974) [Dtdir, link, inc, max j pstat j score] HS
Capacho and Pastor (2004) [pasubgraph j j m] M
Carnahan et al. (2001) [cum j j c, score] HS, GA
Carraway (1989) [tsto j prob j m] DP
Carter and Silverman (1984) [tsto j j Co] HS
Chakravarty (1988) [tdy j j E] HS, DP
Chakravarty and Shtub (1985) [mult j div j Co] M, HS
Chakravarty and Shtub (1986) [mult, tsto j div j Co] HS
Chiang and Urban (2002) [tsto j prob j m] HS, HI
Dar-El and Rabinovitch (1988) [mult, tdy j j Co] M
Deckro (1989) [link, inc, max j j m, c] M
Dolgui and Ihnatsenka (2004), [link, inc j pwork j Co] M, HS, B,
Dolgui et al. (1999), Dolgui et al. (2001a), GR, B&B
Dolgui et al. (2001b) Dolgui et al. (2001c),
Dolgui et al. (2003), Dolgui et al. (2006)
Domschke et al. (1996) [j j E, SSLstat] B&B
Erel and Gökcen (1999), [mix j div j m] M, GR
Gökcen and Erel (1998)
Erel et al. (2001) [j u j m] SA
Erel et al. (2005) [tsto j u j Co] HS
Gadidov and Wilhelm (2000) [type, pa j res01 j Co] M, HS, B&C
Gamberini et al. (2004) [tsto j u j Co, score] HS
Gökcen and Agpak (2006) [j u j m, c, score] M
Gökcen and Erel (1997) [mix, link, inc j j score] M
Gökcen et al. (2005) [j u j m] M, GR
Haq et al. (2005) [mix j j m] HS, GA
Hautsch et al. (1972) [link j pwork2, resmax, feeder j E] HS
Henig (1986) [tsto j prob j Co] DP
Johnson (1983) [type j div, ptask j m] B, B&B
Johnson (1991) [type j j m] B, B&B
Kao (1976), Kao (1979) [tsto j prob j m] DP
Karabati and Sayin (2003) [mix j unpacsync j score] M, HS
Karini and Herer (1995) [mult, tdy j unpacsync j score] M, HS
Kim and Park (1995) [cum j equip j m] M, HS, B&C
Kim et al. (2000a) [fix, type j pwork2 j m] GA
Kim et al. (2000b), Kim et al. (2000c), [mix j u j SSLline] GA
Kim et al. (2006)
Kimms (2000) [mult, spec j unpac, equip j Co] M, B, CG
Klenke (1977) [tsto, link, inc j prob j Pr] M, ENU
Kottas and Lau (1973), Kottas and [tsto j j Co] HS
Lau (1976), Kottas and Lau (1981)
Lapierre and Ruiz (2004) [link, inc, type j div, pwork2, feeder j m] HS
Lapierre et al. (2006) [type j pwork2 j m] TS
Lee et al. (2001) [type j pwork2 j score] HS
Leu et al. (1994) [fix j j score] HS, GA
Levitin et al. (2006) [pa j equip j c] GA
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 685

Table (continued)
Publication Notation Contribution
Liu and Chen (2002) [cum j j c] M, HS
Lyu (1997) [tsto j j Co] HS
Macaskill (1972) [mix j j E] HS
Malakooti (1991) [j j m, c, Co] M, ENU
Malakooti (1994) [j unpac, buffer j m, c, Co, score] HS
Malakooti and Kumar (1996) [fix j unpac, buffer j m, c, Co, score] HS
Matanachai and Yano (2001) [mix j j SSLline, SSLstat] HS, HI
McMullen and Frazier (1997), [mix, tsto j pstat j Co, SSLline, score] SA, ANT
McMullen and Frazier (1998),
McMullen and Tarasewich (2003),
McMullen and Tarasewich (2006)
Merengo et al. (1999) [mix j each j m, SSLline, SSLstat] HS, HI
Miltenburg (1998) [fix j un, Dtunp j m, score] DP
Miltenburg (2002) [mix j u j SSLline, SLstat] GA
Miltenburg and Wijngaard (1994) [j u j m] HS, DP
Miralles (2005) [pa, link, cum j equip j c] M, B, B&B
Moodie and Young (1965) [tsto j j SSLline] HS
Nicosia et al. (2002) [pa j equip j Co] B, DP
Nkasu and Leung (1995) [tsto j prob j m, c, E] M
Park et al. (1997) [spec, inc, paprec j j c] HS, HI
Pastor and Corominas (2000) [link, inc, type, max j j SSLline] M, HS, TS
Pastor et al. (2002) [mult, cum, fix j j c, SSLline, SSLstat] HS, TS
Pinnoi and Wilhelm (1997a) [mult, link, inc, cum, type, pa j div, M
equip, pstat, pwork j Co]
Pinnoi and Wilhelm (1997b) [j j SSLline] M, B, BC
Pinnoi and Wilhelm (1998) [pa j equip j Co] M, B, BC, HS
Pinto et al. (1975) [j ptask j Co] M, B, B&B
Pinto et al. (1981) [j pstat j Co] M, B, B&B
Pinto et al. (1983) [pa j j Co] M, B, B&B
Ponnambalam et al. (2000) [j j m, E, SSLline] GA
Rachamadugu and Talbot (1991) [j j SSLline] M, HI
Raouf and Tsui (1982) [tsto, fix, excl j prob j m, SSLstat] HS
Reeve and Thomas (1973) [tsto j j SSLline] HS
Rekiek et al. (2001) [Dtdir, link, inc, fix, type, pa j j Co, SSLline] GA
Rekiek et al. (2002a) [link, fix, pa j j Co, SSLline] GA
Roberts and Villa (1970) [mix, link j each, ptask j m] M, GR
Rosenberg and Ziegler (1992) [j resmax j Co] M, B, HS
Rosenblatt and Carlson (1985) [j j Pr] M, HS, ENU
Rubinovitz and Bukchin (1993) [pa j equip j m] B, B&B
Sabuncuoglu et al. (2000) [j j m, SSLline] GA
Sarin and Erel (1990) [tsto j j Co] DP
Sarin et al. (1999) [tsto j j Co] HS, HI
Sarker and Shanthikumar (1983) [j pstat j Co, SSLline] HS, HI
Sawik (2002) [mix, cum j div, ptask j c] M
Schofield (1979) [mix, link, inc, fix, excl j Dtunp j c] HS
Scholl and Becker (2005) [j resmax j Co] M, B, B&B
Scholl and Klein (1999) [j u j E] B, B&B
Scholl et al. (2006) [Dtind j j m] M, ENU
(continued on next page)
686 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

Table (continued)
Publication Notation Contribution
sto
Shin and Min (1991) [t j prob j Co] HS
Shtub (1984) [tsto, pa j j Co] M, HS
Shtub and Dar-El (1990) [j j m, c, score] M, HS
Silverman and Carter (1986) [tsto j j Co] HS
Sniedovich (1981) [tsto j prob j m] DP
Sotskov et al. (2006) [tsto j j m] SE
Sparling (1998) [j un, pline, Dtunp j m] M, HS
Sparling and Miltenburg (1998) [mix j u, Dtunp j m, SSLstat] M, HS
Sphicas and Silverman (1976) [tsto j prob j m] M
Süer (1998) [j pline, pstat j m, c, score] HS
Suresh and Sahu (1994), [tsto j j SSLline, score] SA, GA
Suresh et al. (1996)
Thomopoulos (1970) [mix j j m, SSLstat] HS
Tsujimura et al. (1995) [tsto j j score] M, GA
Ugurdag et al. (1997) [j j c, SSLline] M, HS, HI
Urban (1998) [j u j m] M
Urban and Chiang (2006a) [tsto j prob, u, Dtunp j m] M, B
Urban and Chiang (2006b) [tsto j unpacasync j m] HS, TS
Van Hop (2006) [mix, tsto j div j m] M, HS, HI
Vilarinho and Simaria (2002) [mix, link, inc j prob, pstat j m, SSLline, SSLstat] M, SA
Vilarinho and Simaria (2006) [mix, link, inc, fix j pstat j m, SSLline, SSLstat] M, ANT
Visich et al. (2002) [mix j u j SSLstat] TS, SA
Wilhelm (1999) [Dtdir, pa j equip j Co] M, B, CG
Wilhelm and Gadidov (2004) [mix, cum, pa j res01 j Co] M, HS, BC
Wilson (1986) [tsto, pa j j Co] M
Zäpfel (1975) [link, inc j j Pr] M, ENU

5. Research challenges and conclusions reconfiguration is often considered by introducing


assignment restrictions. If machinery is too heavy
The classification of the literature reveals a num- for being moved to another station, the assignment
ber of important open fields of research which of tasks which require such a machine can be forced
require an in-depth discussion to narrow the gap to its current station by an assignment restriction
between research and practice. In the following, (a5 = fix). Nevertheless, often the reallocation of
we discuss those which seem to be most important. machines is not technically impossible, but rather
associated with moving costs (Gamberini et al.,
5.1. Rebalancing 2004). The same applies to operators at a station,
who might need to be trained with regard to the
Within the lifetime of a modern production sys- new work content. Costs can be decreased when
tem, ALB problems do not only occur once prior workers keep as much of their previous tasks as pos-
to its construction, but rather continuously as recon- sible. Accordingly, assembly line balancing should
figuration or rebalancing (Schofield, 1979; Falk- regard these moving costs.
enauer, 2005). These might be necessary to react
on shifts in the demand structure or whenever new 5.2. Cost synergies
production technologies become available. As a
consequence, stations will often have a unique iden- A similar phenomenon occurs when several tasks
tity expressed by a position in the workshop, require the same resources. In this case, cost of cap-
assigned resources or space restrictions (a5 = cum), ital can be decreased when tasks are assigned to the
instead of being abstract entities. In the literature, same station, which is typically enforced by zoning
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 687

restrictions (a5 = link). Whether the benefits associ- whole assembly system is questionable, especially
ated with avoiding an extra resource actually sur- if operators can perform tasks on both the feeder
mount the disadvantages of a more restricted and the main line at their contact point (Lapierre
balancing problem, can just be answered by explic- and Ruiz, 2004).
itly modeling cost synergies (b4 = res01).
In the literature, these cost aspects are often 5.5. Material supply
neglected and costs are directly assigned to tasks
or stations instead. This, however, does often not Another important but almost completely
reflect the real situation. In real life, costs only arise ignored aspect is the material supply of an assembly
by the purchase or usage of a resource which is posi- line (b5 = mat). Material is usually provided in some
tioned at a station. By exactly assigning costs to the kind of container along the line. A container could
resources instead of stations or tasks, a more realis- possibly be some traditional shelf or box, but also
tic modeling of the decision problem could be an automated guided vehicle (AGV) which delivers
attained. the material directly to the stations. In such a set-
ting, new decision problems arise concerning the
5.3. Sequence-dependent time increments dimensions of these containers for different compo-
nents (Bukchin and Meller, 2005) and their alloca-
Sequence-dependent relations between tasks also tion along the line. If more than one station is
require further investigation. Occasionally, prece- served by the same material container or if distances
dence-dependent time increments (Buxey, 1974; vary depending on the actual position of the opera-
Wilhelm, 1999; Rekiek et al., 2001; Bautista and tor, station times are affected by variable walking
Pereira, 2002) are regarded, which result from a tool distances for fetching the material. Due to these
or position change between predecessors and their interdependencies, it might be necessary in practice
direct successors (a4 = Dtdir). However, task times to solve line balancing and material supply prob-
can also be influenced by an indirect succession of lems simultaneously.
tasks (a4 = Dtind). These time increments occur if
the resulting transformation of a completed task 5.6. Parallel working places
obstructs the implementation of another, e.g., seat
and safety belt of a car. These dependencies can Furthermore, little attention has been paid to
even cover more than two tasks. For example, the parallel working places (b3 = pwork). Although,
assembly of a car’s hand-break might be obstructed simultaneous work at a station requires a suffi-
only after both front seats have been installed. ciently large workpiece, there are important prac-
Otherwise, the access is free from at least one side. tical applications in automobile and related
In an extreme case, the assembly of a hand break industries (Falkenauer, 2005). Workers are able to
might even be impossible if both seats are installed. perform operations at different mounting places at
These phenomena cannot be included in traditional the same time if tasks are compatible and do not
arcs of the precedence graph, since they depict require the workpiece to be in an exclusive position.
dependencies exclusively between two tasks. By dis- In case of parallel work, precedence constraints lead
regarding these dependencies assembly line balanc- to additional idle times whenever one operator has
ing might result in infeasible line balances. to wait for another one finishing a preceding task.
Accordingly, the station time can no longer be com-
5.4. Feeder lines puted by simply summing up the assigned task
times, but requires the solution of a detailed sched-
Little attention has been paid to the so called uling problem instead. To close the gap to practice
‘‘additional aspects of line configuration’’. Real in this field of research, optimization models have
world assembly lines often have one or more feeder to be developed which go far beyond a simple
lines (b5 = feeder) which flow into a main line assignment of tasks to working places.
(Tempelmeier, 2003). A possible approach to con-
figure such an assembly system is to balance the 5.7. Processing alternatives
main line first and use the retrieved cycle time to
balance each feeder line separately. Whether this Over the last few years, a lot of research has
decomposition results to a global optimum of the regarded processing alternatives. Most tasks can be
688 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

performed in different ways: A faster machine and/ ing a task. In reality, processing alternatives might,
or multiple manning decrease the processing time however, also alter precedence constraints between
but commonly raise cost of capital (time-cost tasks (a6 = paprec). For example, the automated
trade-off). Choosing processing alternatives and bal- assembly of an option using a robot can be
ancing the line simultaneously promises better con- obstructed by an already installed part whereas a
figurations than deciding successively. A successive manual assembly is still possible. Even whole sub-
planning approach selects a processing alternative graphs of the precedence graph may be substituted
for each task first and then balances the line with by a processing alternative (a6 = pasubgraph). An
regard to the chosen alternatives. Hints on the example is the assembly of a procured module
advantage’s extent when planning simultaneously which unites a set of optional features instead of
(maybe with less efficient algorithms) in comparison individually installing each option. Both approaches
to planning successively would be of great value in of processing alternatives need to be extended in this
this field of research. direction.
Two different approaches have been proposed to
incorporate processing alternatives into ALB. The 5.8. Disassembly line balancing
first is known as the equipment selection problem
[pa j equip j Co] (e.g., Pinnoi and Wilhelm, 1998; A related problem to ALB labeled disassembly
Bukchin and Tzur, 2000). It is based on the assump- line balancing (DLB) raised in the recent years
tion that there is a fixed set of equipments (complete within the field of reverse logistics (Fleischmann
set of resources) exactly one of which has to be et al., 1997). Though a lot of classified ALB charac-
selected and assigned to a station. The alternative teristics are relevant for DLB, too, some important
approach consists in assigning processes to tasks. peculiarities exist, which are exclusively related to
In addition to line balancing, for each task exactly the DLB-case. In contrast to ALB, where a finished
one processing alternative has to be chosen out of product leaves the assembly system each production
a set of possible ones. Because these processes cycle, a returned product is disassembled into its
require resources, the problem can be interpreted consisting parts, which build the output of the disas-
as an (implicit) equipment design problem, repre- sembly process. Moreover, hazardous tasks may
sented by [pa j j Co] in our classification scheme damage demanded parts disassembled later on in
(e.g., Pinto et al., 1983). The chosen processes are the process, which leads to sequence-dependent out-
usually not independent of each other, which has put rates. Excellent overviews about the peculiari-
to be reflected by considering possible synergies ties of DLB are given by Güngör and Gupta
arising from jointly using resources for several tasks (2002) as well as Lambert and Gupta (2005). To
at a station (b4 = res01) or different types of assign- avoid continual changes of the viewpoint between
ment restrictions (a5 = linkpa, a5 = incpa, a5 = the ALB and DLB perspective we consciously
cumpa). exclude DLB from our classification within this
Both concepts can in principle be used inter- paper. Nevertheless, an enhancement of our classifi-
changeably. However, the resulting mathematical cation scheme to the specific features of DLB would
models might differ considerably. While the equip- be a valuable contribution of further research.
ment selection approach excludes incompatible pro-
cessing alternatives prior to modeling at the cost of 5.9. Test beds and flexible solution procedures
a potentially large number of equipment alterna-
tives, the equipment design approach needs to Our survey finally revealed that there is a dire
exclude incompatible combinations within the need of systematic evaluations to identify those
model which can hence lead to more restrictions decision concepts – combining a (theoretical) prob-
and a higher complexity. Due to limitations in data lem type considered, the mathematical model
generation and model solving, both approaches derived and the (exact or heuristic) solution method
have to be limited to a subset of possible alternatives applied – which are best suited to solving real-world
in real-world settings. In order to find out which assembly line balancing problems. In particular, this
approach is better suited to different situations, includes the systematic generation of realistic test
comparative analyses should be performed. beds and the development of solution algorithms
In the literature, both approaches are so far flexible enough to jointly cover as many problem
restricted to covering different modes of accomplish- characteristics as possible. The structure provided
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 689

in this paper might therefore also be employed to Journal of Manufacturing and Operations Management 3,
develop new instance generators and solution proce- 200–223.
Bard, J.F., 1989. Assembly line balancing with parallel worksta-
dures which cover all or a certain systematic subset tions and dead time. International Journal of Production
of important problem characteristics. Research 27, 1005–1018.
We would thus like to encourage all researchers Bartholdi, J.J., 1993. Balancing two-sided assembly lines: A case
to make use of the provided scheme (and to extend study. International Journal of Production Research 31,
it if necessary), in order to establish a common ref- 2447–2461.
Bautista, J., Pereira, J., 2002. Ant algorithms for assembly line
erence for past as well as future publications con- balancing. Lecture Notes in Computer Science 2463, 65–75.
cerning assembly line balancing. This should ease Bautista, J., Pereira, J., 2006. Ant algorithms for a time and space
the coordination of remaining steps to close the sta- constrained assembly line balancing problem. European
ted gap between research and practice substantially. Journal of Operational Research 177, 2016–2032.
Bautista, J., Suarez, R., Mateo, M., Companys, R., 2000. Local
search heuristics for the assembly line balancing problem with
References incompatibilities between tasks. In: Proceedings of the 2000
IEEE International Conference on Robotics and Automa-
Aase, G.R., Schniederjans, M.J., Olson, J.R., 2003. U-OPT: An tion, San Francisco, CA, pp. 2404–2409.
analysis of exact U-shaped line balancing procedures. Inter- Baybars, I., 1986. A survey of exact algorithms for the simple
national Journal of Production Research 41, 4185–4210. assembly line balancing problem. Management Science 32,
Aase, G.R., Olson, J.R., Schniederjans, M.J., 2004. U-shaped 909–932.
assembly line layouts and their impact on labor productivity: Baykasoglu, A., Özbakir, L., 2006. Stochastic U-line balancing
An experimental study. European Journal of Operational using genetic algorithms. International Journal of Advanced
Research 156, 698–711. Manufacturing Technology. doi:10.1007/s00170-005-0322-4.
Agnetis, A., Ciancimino, A., Lucertini, M., Pizzichella, M., 1995. Becker, C., Scholl, A., 2006. A survey on problems and methods
Balancing flexible lines for car components assembly. Inter- in generalized assembly line balancing. European Journal of
national Journal of Production Research 33, 333–350. Operational Research 168, 694–715.
Agrawal, P.K., 1985. The related activity concept in assembly line Berger, I., Bourjolly, J.-M., Laporte, G., 1992. Branch-and-
balancing. International Journal of Production Research 23, Bound algorithms for the multi-product assembly line bal-
403–421. ancing problem. European Journal of Operational Research
Ajenblit, D.A., Wainwright, R.L., 1998. Applying genetic algo- 58, 215–222.
rithms to the U-shaped assembly line balancing problem. In: Bhattacharjee, T.K., Sahu, S., 1987. A heuristic approach to
Proceedings of the 1998 IEEE International Conference on general assembly line balancing. International Journal of
Evolutionary Computation, Anchorage, AK, pp. 96–101. Operations and Production Management 8, 67–77.
Akagi, F., Osaki, H., Kikuchi, S., 1983. A method for assembly Boysen, N., 2005. Variantenfließfertigung. Gabler, Wiesbaden.
line balancing with more than one worker in each station. Boysen, N., Fliedner, M., 2006. A versatile algorithm for
International Journal of Production Research 21, 755–770. assembly line balancing. European Journal of Operational
Amen, M., 1997. Ein exaktes Verfahren zur kostenorientierten Research, to appear.
Fließbandabstimmung. In: Zimmermann, U., et al. (Eds.), Brucker, P., Drexl, A., Möhring, R.H., Neumann, K., Pesch, E.,
Operations Research Proceedings 1996. Springer, Berlin, pp. 1999. Resource-constrained project scheduling: Notation,
224–229. classification, models and methods. European Journal of
Amen, M., 2000a. An exact method for cost-oriented assembly Operational Research 112, 3–41.
line balancing. International Journal of Production Econom- Bukchin, J., Masin, M., 2004. Multi-objective design of team
ics 64, 187–195. oriented assembly systems. European Journal of Operational
Amen, M., 2000b. Heuristic methods for cost-oriented assembly Research 156, 326–352.
line balancing: A survey. International Journal of Production Bukchin, Y., Meller, R.D., 2005. A space allocation algorithm for
Economics 68, 1–14. assembly line components. IIE Transactions 37, 51–61.
Amen, M., 2001. Heuristic methods for cost-oriented assembly Bukchin, Y., Rabinowitch, I., 2005. A branch-and-bound based
line balancing: A comparison on solution quality and solution approach for the mixed-model assembly line-balanc-
computing time. International Journal of Production Eco- ing problem for minimizing stations and task duplication costs.
nomics 69, 255–264. European Journal of Operational Research 174, 492–508.
Amen, M., 2006. Cost-oriented assembly line balancing: Model Bukchin, J., Rubinovitz, J., 2003. A weighted approach for
formulations, solution difficulty, upper and lower bounds. assembly line design with station paralleling and equipment
European Journal of Operational Research 168, 747–770. selection. IIE Transactions 35, 73–85.
Arcus, A.L., 1966. COMSOAL: A computer method of sequenc- Bukchin, J., Tzur, M., 2000. Design of flexible assembly line to
ing operations for assembly lines. International Journal of minimize equipment cost. IIE Transactions 32, 585–598.
Production Research 4, 259–277. Bukchin, J., Dar-El, E.M., Rubinovitz, J., 1997. Team oriented
Askin, R.G., Zhou, M., 1997. A parallel station heuristic for the assembly system design: A new approach. International
mixed-model production line balancing problem. Interna- Journal of Production Economics 51, 47–57.
tional Journal of Production Research 35, 3095–3105. Bukchin, J., Dar-El, E.M., Rubinovitz, J., 2002. Mixed-model
Baker, K.R., Powell, S.G., Pyke, D.F., 1990. Buffered and assembly line design in a make-to-order environment. Com-
unbuffered assembly systems with variable processing times. puters and Industrial Engineering 41, 405–421.
690 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

Burns, L.D., Daganzo, C.F., 1987. Assembly line job sequencing tions. In: Proceedings of the 28th International Conference on
principles. International Journal of Production Research 25, Computer and Industrial Engineering, Florida, USA, 2001,
71–99. pp. 541–547.
Buxey, G.M., 1974. Assembly line balancing with multiple Dolgui, A., Guschinsky, N., Levin, G., Harrath, Y., 2001c.
stations. Management Science 20, 1010–1021. Optimal design of a class of transfer lines with parallel
Buxey, G.M., Slack, N.D., Wild, R., 1973. Production flow line operations. In: Groumpos, P.P., Tzes, A.A. (Eds.), Manufac-
system design – A review. AIIE Transactions 5, 37–48. turing, modeling, management and control, Proceedings of
Capacho, L., Pastor, R., 2004. ASALBP: The alternative the IFAC Symposium, Patras, Greece, 2000, Elsevier,
subgraphs assembly line balancing problem, Working Paper, Amsterdam, pp. 36–41.
Universitat Politecnica de Catalunya, Spain. Dolgui, A., Guschinsky, N., Levin, G., 2003. Optimal design of
Carnahan, B.J., Norman, B.A., Redfern, M.S., 2001. Incorpo- automated transfer lines with blocks of parallel operations.
rating physical demand criteria into assembly line balancing. In: Camacho, E.F., Basanez, L., De la Puente, J.A. (Eds.),
IIE Transactions 33, 875–887. Proceedings of the 15th IFAC World Congress, Barcelona,
Carraway, R.L., 1989. A dynamic programming approach to Spain, 2002, Elsevier, Amsterdam (on cd-rom).
stochastic assembly line balancing. Management Science 35, Dolgui, A., Guschinsky, N., Levin, G., 2006. A special case of
459–471. transfer lines balancing by graph approach. European Journal
Carter, J.C., Silverman, F.N., 1984. A cost-effective approach to of Operational Research 168, 732–746.
stochastic line balancing with off-line repairs. Journal of Domschke, W., Klein, R., Scholl, A., 1996. Antizipative Leis-
Operations Management 4, 145–157. tungsabstimmung bei moderner Variantenfließfertigung.
Chakravarty, A.K., 1988. Line balancing with task learning Zeitschrift für Betriebswirtschaft 66, 1465–1490.
effects. IIE Transactions 20, 186–193. Erel, E., Gökcen, H., 1999. Shortest route formulation of mixed-
Chakravarty, A.K., Shtub, A., 1985. Balancing mixed model lines model assembly line balancing problem. European Journal of
with in-process inventories. Management Science 31, 1161– Operational Research 116, 194–204.
1174. Erel, E., Sarin, S.C., 1998. A survey of the assembly line
Chakravarty, A.K., Shtub, A., 1986. A cost minimization balancing procedures. Production Planning and Control 9,
procedure for mixed model production lines with normally 414–434.
distributed task times. European Journal of Operational Erel, E., Sabuncuoglu, I., Aksu, B.A., 2001. Balancing of U-type
Research 23, 25–36. assembly systems using simulated annealing. International
Chase, R.B., 1974. Survey of paced assembly lines. Industrial Journal of Production Research 39, 3003–3015.
Engineering 6 (2), 14–18. Erel, E., Sabuncuoglu, I., Sederci, H., 2005. Stochastic assembly
Chiang, W.-C., Urban, T.L., 2002. A hybrid heuristic for the line balancing using beam search. International Journal of
stochastic U-line balancing problem, Working Paper, Uni- Production Research 43, 1411–1426.
versity of Tulsa, Oklahoma, USA. Falkenauer, E., 2005. Line balancing in the real world. In:
Dallery, Y., Gershwin, S.B., 1992. Manufacturing flow line Proceedings of the International Conference on Product
systems: A review of models and analytical results. Queueing Lifecycle Management PLM 05, Lumiere University of Lyon,
Systems Theory and Applications 12, 3–94. France, 2005 (on cd-rom).
Dar-El, E.M., Rabinovitch, M., 1988. Optimal planning and Fleischmann, M., Bloemhof-Ruwaard, J.M., Dekker, R., van der
scheduling of assembly lines. International Journal of Pro- Laan, E., van Nunen, J.A.E.E., Van Wassenhove, L.N., 1997.
duction Research 26, 1433–1450. Quantitative models for reverse logistics: A review. European
Deckro, R.F., 1989. Balancing cycle time and workstations. IIE Journal of Operational Research 103, 1–17.
Transactions 21, 106–111. Gadidov, R., Wilhelm, W., 2000. A cutting plane approach
Dobson, G., Yano, C.A., 1994. Cyclic scheduling to minimize for the single-product assembly system design problem.
inventory in a batch flow line. European Journal of Opera- International Journal of Production Research 38, 1731–
tional Research 75, 441–461. 1754.
Dolgui, A., Ihnatsenka, I., 2004. Branch and bound algorithm for Gamberini, R., Grassi, A., Gamberi, M., Manzini, R., Regattieri,
optimal design of transfer lines with multi-spindle stations, A., 2004. U-shaped assembly lines with stochastic tasks
Working paper, Ecole Nationale Suoerieure des Mines, Saint- execution times: heuristic procedures for balancing and re-
Etienne, France. balancing problems. In: Proceedings of the Business and
Dolgui, A., Guschinski, N., Levin, G., 1999. On problem of Industry Symposium, 2004 Advanced Simulation Technolo-
optimal design of transfer lines with parallel and sequential gies Conference, Arlington, Virginia, <http://www.scs.org/
operations. In: Fuertes, J.M. (Ed.), Proceedings of the 7th scsarchive/getDoc.cfm?id=1719>.
IEEE International Conference on Emerging Technologies Geoffrion, A.M., Graves, G.W., 1976. Scheduling parallel pro-
and Factory Automation, Barcelona, Spain, vol. 1, 1999, pp. duction lines with changeover costs: Practical application of a
329–334. quadratic assignment/LP approach. Operations Research 24,
Dolgui, A., Guschinsky, N., Levin, G., 2001a. Decomposition 595–610.
methods to optimize transfer line with parallel and sequential Gershwin, S., 2000. Design and operation of manufacturing
machining. In: Binder, Z. (Ed.), Management and control of systems: The control-point policy. IIE Transactions 32, 891–
production and logistics, Proceedings of the 2nd IFAC 906.
Conference, Grenoble, France, 2000, vol. 3. Elsevier, Amster- Ghosh, S., Gagnon, R.J., 1989. A comprehensive literature
dam, pp. 983–988. review and analysis of the design, balancing and scheduling of
Dolgui, A., Guschinsky, N., Levin, G., 2001b. A mixed integer assembly systems. International Journal of Production
program for balancing of transfer line with grouped opera- Research 27, 637–670.
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 691

Gökcen, H., Agpak, K., 2006. A goal programming approach to Kim, Y.K., Kim, S.J., Kim, J.Y., 2000c. Balancing and sequenc-
simple U-line balancing problem. European Journal of ing mixed-model U-lines with a co-evolutionary algorithm.
Operational Research 171, 577–585. Production Planning and Control 11, 754–764.
Gökcen, H., Erel, E., 1997. A goal programming approach to Kim, Y.K., Kim, J.Y., Kim, Y., 2006. An endosymbiotic
mixed-model assembly line balancing problem. International evolutionary algorithm for the integration of balancing and
Journal of Production Economics 48, 177–185. sequencing in mixed-model U-lines. European Journal of
Gökcen, H., Erel, E., 1998. Binary integer formulation for mixed- Operational Research 168, 838–852.
model assembly line balancing problem. Computers and Kimms, A., 2000. Minimal investment budgets for flow line
Industrial Engineering 34, 451–461. configuration. IIE Transactions 32, 287–298.
Gökcen, H., Agpak, K., Gencer, C., Kizilkaya, E., 2005. A Klenke, H., 1977. Ablaufplanung bei Fließfertigung. Gabler,
shortest route formulation of simple U-type assembly line Wiesbaden.
balancing problem. Applied Mathematical Modelling 29, Kottas, J.F., Lau, H.-S., 1973. A cost-oriented approach to
373–380. stochastic line balancing. AIIE Transactions 5, 164–171.
Graham, R.L., Lawler, E.L., Lenstra, J.K., Rinnooy Kan, Kottas, J.F., Lau, H.-S., 1976. A total operating cost model for
A.H.G., 1979. Optimization and approximation in determin- paced lines with stochastic task times. AIIE Transactions 8,
istic sequencing and scheduling: A survey. Annals of Discrete 234–240.
Mathematics 5, 287–326. Kottas, J.F., Lau, H.-S., 1981. A stochastic line balancing
Güngör, A., Gupta, S.M., 2002. Disassembly line in product procedure. International Journal of Production Research 19,
recovery. International Journal of Production Research 40, 177–193.
2569–2589. Lambert, A.J.D., Gupta, S.M., 2005. Disassembly modeling for
Haq, A.N., Jayaprakash, J., Rengarajan, K., 2005. A hybrid assembly, maintenance, reuse, and recycling. CRC Press,
genetic algorithm approach to mixed-model assembly line Boca Raton.
balancing. International Journal of Advanced Manufacturing Lapierre, S.D., Ruiz, A.B., 2004. Balancing assembly lines: An
Technology. doi:10.1007/s00170-004-2373-3. industrial case study. Journal of the Operational Research
Hautsch, K., John, H., Schürgers, H., 1972. Taktabstimmung bei Society 55, 589–597.
Fließarbeit mit dem Positionswert-Verfahren. REFA-Nach- Lapierre, S.D., Ruiz, A., Soriano, P., 2006. Balancing assembly
richten 25, 451–464. lines with tabu search. European Journal of Operational
Henig, M.I., 1986. Extensions of the dynamic programming Research 168, 826–837.
method in the deterministic and stochastic assembly-line Lee, T.O., Kim, Y., Kim, Y.K., 2001. Two-sided assembly line
balancing problems. Computers and Operations Research 13, balancing to maximize work relatedness and slackness.
443–449. Computers and Industrial Engineering 40, 273–292.
Hillier, F.S., So, K.C., Boling, R.W., 1993. Toward character- Leu, Y.Y., Matheson, L.A., Rees, L.P., 1994. Assembly line
izing the optimal allocation of storage space in production balancing using genetic algorithms with heuristic-generated
line systems with variable processing times. Management initial populations and multiple evaluation criteria. Decision
Science 39, 126–133. Sciences 25, 581–606.
Johnson, R.V., 1983. A branch and bound algorithm for Levitin, G., Rubinovitz, J., Shnits, B., 2006. A genetic algorithm
assembly line balancing problems with formulation irregular- for robotic assembly line balancing. European Journal of
ities. Management Science 29, 1309–1324. Operational Research 168, 811–825.
Johnson, R.V., 1991. Balancing assembly lines for teams and Liu, C.-M., Chen, C.-H., 2002. Multi-section electronic assembly
work groups. International Journal of Production Research line balancing problems: A case study. Production Planning
29, 1205–1214. and Control 13, 451–461.
Kao, E.P.C., 1976. A preference order dynamic program for Lyu, J., 1997. A single-run optimization algorithm for stochastic
stochastic assembly line balancing. Management Science 22, assembly line balancing problems. Journal of Manufacturing
1097–1104. Systems 16, 204–210.
Kao, E.P.C., 1979. Computational experience with a stochastic Macaskill, J.L.C., 1972. Production-line balances for mixed-
assembly line balancing algorithm. Computers and Opera- model lines. Management Science 19, 423–434.
tions Research 6, 79–86. Malakooti, B., 1991. A multiple criteria decision making
Karabati, S., Sayin, S., 2003. Assembly line balancing in a mixed- approach for the assembly line balancing problem. Interna-
model sequencing environment with synchronous transfers. tional Journal of Production Research 29, 1979–2001.
European Journal of Operational Research 149, 417–429. Malakooti, B., 1994. Assembly line balancing with buffers by
Karini, R., Herer, Y.T., 1995. Allocation of tasks to stations in multiple criteria optimization. International Journal of Pro-
small-batch assembly with learning: Basic concepts. Interna- duction Research 32, 2159–2178.
tional Journal of Production Research 33, 2973–2998. Malakooti, B., Kumar, A., 1996. A knowledge-based system for
Kim, H., Park, S., 1995. A strong cutting plane algorithm for the solving multi-objective assembly line balancing problems.
robotic assembly line balancing problem. International Jour- International Journal of Production Research 34, 2533–2552.
nal of Production Research 33, 2311–2323. Matanachai, S., Yano, C.A., 2001. Balancing mixed-model
Kim, Y.K., Kim, Y., Kim, Y.J., 2000a. Two-sided assembly line assembly lines to reduce work overload. IIE Transactions
balancing: a genetic algorithm approach. Production Plan- 33, 29–42.
ning and Control 11, 44–53. Mather, H., 1989. Competitive manufacturing. Prentice Hall,
Kim, Y.K., Kim, J.Y., Kim, Y., 2000b. A coevolutionary Englewood Cliffs, NJ.
algorithm for balancing and sequencing in mixed model McMullen, P.R., Frazier, G.V., 1997. A heuristic for solving
assembly lines. Applied Intelligence 13, 247–258. mixed-model line balancing problems with stochastic task
692 N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693

durations and parallel stations. International Journal of Pinnoi, A., Wilhelm, W.E., 1997a. A family of hierarchical
Production Economics 51, 177–190. models for assembly system design. International Journal of
McMullen, P.R., Frazier, G.V., 1998. Using simulated annealing Production Research 35, 253–280.
to solve a multiobjective assembly line balancing problem Pinnoi, A., Wilhelm, W.E., 1997b. A branch and cut approach
with parallel workstations. International Journal of Produc- for workload smoothing on assembly lines. INFORMS
tion Research 36, 2717–2741. Journal on Computing 9, 335–350.
McMullen, P.R., Tarasewich, P., 2003. Using ant techniques to Pinnoi, A., Wilhelm, W.E., 1998. Assembly system design: A
solve the assembly line balancing problem. IIE Transactions branch and cut approach. Management Science 44, 103–118.
35, 605–617. Pinto, P.A., Dannenbring, D.G., Khumawala, B.M., 1975. A
McMullen, P.R., Tarasewich, P., 2006. Multi-objective assembly branch and bound algorithm for assembly line balancing with
line balancing via a modified ant colony optimization paralleling. International Journal of Production Research 13,
technique. International Journal of Production Research 44, 183–196.
27–42. Pinto, P.A., Dannenbring, D.G., Khumawala, B.M., 1981.
Merengo, C., Nava, F., Pozetti, A., 1999. Balancing and Branch and bound and heuristic procedures for assembly
sequencing manual mixed-model assembly lines. International line balancing with paralleling of stations. International
Journal of Production Research 37, 2835–2860. Journal of Production Research 19, 565–576.
Meyr, H., 2004. Supply chain planning in the German automo- Pinto, P.A., Dannenbring, D.G., Khumawala, B.M., 1983.
tive industry. OR Spectrum 26, 447–470. Assembly line balancing with processing alternatives: An
Miltenburg, J., 1998. Balancing U-lines in a multiple U-line application. Management Science 29, 817–830.
facility. European Journal of Operational Research 109, 1–23. Ponnambalam, S.G., Aravindan, P., Naidu, G.M., 2000. A
Miltenburg, J., 2002. Balancing and scheduling mixed-model multi-objective genetic algorithm for solving assembly line
U-shaped production lines. International Journal of Flexible balancing problem. International Journal of Advanced
Manufacturing Systems 14, 119–151. Manufacturing Technology 16, 341–352.
Miltenburg, J., Wijngaard, J., 1994. The U-line line balancing Rachamadugu, R., Talbot, B., 1991. Improving the equality of
problem. Management Science 40, 1378–1388. workload assignments in assembly lines. International Jour-
Miralles, C., 2005. Solving procedures for the assembly line nal of Production Research 29, 619–633.
worker assignment and balancing problem: application to Raouf, A., Tsui, C., 1982. A new method for assembly line
sheltered work centres for disabled. XI Escuela Latinoamer- balancing having stochastic work elements. Computers and
icana de Verano en Investigación de Operaciones, Villa de Industrial Engineering 6, 131–148.
Leyva, Colombia, <http://elavio2005.uniandes.edu.co/ Reeve, N.R., Thomas, W.H., 1973. Balancing stochastic assem-
ResumenesParticipantes/Lunes/MirallesCristobal_R.pdf>. bly lines. AIIE Transactions 5, 223–229.
Monden, Y., 1998. Toyota production system – An integrated Rekiek, B., de Lit, P., Pellichero, F., L’Eglise, T., Fouda, P.,
approach to just-in-time, 3rd ed. Kluwer, Dordrecht. Falkenauer, E., Delchambre, A., 2001. A multiple objective
Moodie, C.L., Young, H.H., 1965. A heuristic method of grouping genetic algorithm for assembly line design. Journal
assembly line balancing for assumptions of constant or of Intelligent Manufacturing 12, 467–485.
variable work element times. Journal of Industrial Engineer- Rekiek, B., de Lit, P., Delchambre, A., 2002a. Hybrid assembly
ing 16, 23–29. line design and user’s preferences. International Journal of
Nicosia, G., Pacciarelli, D., Pacifici, A., 2002. Optimally balanc- Production Research 40, 1095–1111.
ing assembly lines with different workstations. Discrete Rekiek, B., Dolgui, A., Delchambre, A., Bratcu, A., 2002b. State
Applied Mathematics 118, 99–113. of art of optimization methods for assembly line design.
Nkasu, M.M., Leung, K.H., 1995. A stochastic approach to Annual Reviews in Control 26, 163–174.
assembly line balancing. International Journal of Production Roberts, S.D., Villa, C.D., 1970. On a multiproduct assembly
Research 33, 975–991. line-balancing problem. AIIE Transactions 2, 361–364.
Papadopoulos, H.T., Heavey, C., 1996. Queueing theory in Rosenberg, O., Ziegler, H., 1992. A comparison of heuristic
manufacturing systems analysis and design: A classification of algorithms for cost-oriented assembly line balancing. Zeits-
models for production and transfer lines. European Journal chrift für Operations Research 36, 477–495.
of Operational Research 92, 1–27. Rosenblatt, M.J., Carlson, R.C., 1985. Designing a production
Park, K., Park, S., Kim, W., 1997. A heuristic for an assembly line to maximize profit. IIE Transactions 17, 117–121.
line balancing problem with incompatibility, range, and Rubinovitz, J., Bukchin, J., 1993. RALB – A heuristic algorithm
partial precedence constraints. Computers and Industrial for design and balancing of robotic assembly lines. Annals of
Engineering 32, 321–332. the CIRP 42, 497–500.
Pastor, R., Corominas, A., 2000. Assembly line balancing with Sabuncuoglu, I., Erel, E., Tanyer, M., 2000. Assembly line
incompatibilities and bounded workstation loads. Ricerca balancing using genetic algorithms. Journal of Intelligent
Operativa 30, 23–45. Manufacturing 11, 295–310.
Pastor, R., Andres, C., Duran, A., Perez, M., 2002. Tabu search Salveson, M.E., 1955. The assembly line balancing problem. The
algorithms for an industrial multi-product and multi-objective Journal of Industrial Engineering 6 (3), 18–25.
assembly line balancing problem, with reduction of the task Sarin, S.C., Erel, E., 1990. Development of cost model for the
dispersion. Journal of the Operational Research Society 53, single-model stochastic assembly line balancing problem.
1317–1323. International Journal of Production Research 28, 1305–1316.
Pine, B.J., 1993. Mass customization: The new frontier in Sarin, S.C., Erel, E., Dar-El, E.M., 1999. A methodology for
business competition. Harvard Business School Press, Boston, solving single-model, stochastic assembly line balancing
Mass. problem. Omega 27, 525–535.
N. Boysen et al. / European Journal of Operational Research 183 (2007) 674–693 693

Sarker, B.R., Shanthikumar, J.G., 1983. A generalized approach Suresh, G., Sahu, S., 1994. Stochastic assembly line balancing
for serial or parallel line balancing. International Journal of using simulated annealing. International Journal of Produc-
Production Research 21, 109–133. tion Research 32, 1801–1810.
Sawik, T., 2002. Monolithic vs. hierarchical balancing and Suresh, G., Vinod, V.V., Sahu, S., 1996. Genetic algorithm for
scheduling of a flexible assembly line. European Journal of assembly line balancing. Production Planning and Control 7,
Operational Research 143, 115–124. 38–46.
Schofield, N.A., 1979. Assembly line balancing and the applica- Tempelmeier, H., 2003. Practical considerations in the optimiza-
tion of computer techniques. Computers and Industrial tion of flow production systems. International Journal of
Engineering 3, 53–59. Production Research 41, 149–170.
Scholl, A., 1999. Balancing and sequencing assembly lines, 2nd Thomopoulos, N.T., 1970. Mixed model line balancing with
ed. Heidelberg, Physica. smoothed station assignments. Management Science 16, 593–
Scholl, A., Becker, C., 2005. A note on an exact method for cost- 603.
oriented assembly line balancing. International Journal of Tsujimura, Y., Gen, M., Kubota, E., 1995. Solving fuzzy
Production Economics 97, 343–352. assembly-line balancing problem with genetic algorithms.
Scholl, A., Becker, C., 2006. State-of-the-art exact and heuristic Computers and Industrial Engineering 29, 543–547.
solution procedures for simple assembly line balancing. Ugurdag, H.F., Rachamadugu, R., Papachristou, C.A., 1997.
European Journal of Operations Research 168, 666–693. Designing paced assembly lines with fixed number of stations.
Scholl, A., Klein, R., 1999. ULINO: Optimally balancing U- European Journal of Operational Research 102, 488–
shaped JIT assembly lines. International Journal of Produc- 501.
tion Research 37, 721–736. Urban, T.L., 1998. Note. Optimal balancing of U-shaped
Scholl, A., Boysen, N., Fliedner, M., 2006. The sequence- assembly lines. Management Science 44, 738–741.
dependent assembly line balancing problem. Operation Urban, T.L., Chiang, W.-C., 2006a. An optimal piecewise-linear
Research Spectrum, doi:10.1007/s00291-006-0070-3. optimization of the U-line balancing problem with stochastic
Schöniger, J., Spingler, J., 1989. Planung der Montageanlage. task times. European Journal of Operational Research 168,
Technica 14, 27–32. 771–782.
Shin, D., Min, H., 1991. Uniform assembly line balancing with Urban, T.L., Chiang, W.-C., 2006b. Balancing unpaced synchro-
stochastic task times in just-in-time manufacturing. Interna- nous production lines, Working Paper, University of Tulsa,
tional Journal of Operations and Production Management Oklahoma, USA.
(11/8), 23–34. Van Hop, N., 2006. A heuristic solution for fuzzy mixed-model
Shtub, A., 1984. The effect of incompletion cost on line balancing line balancing problem. European Journal of Operational
with multiple manning of work stations. International Journal Research 168, 789–810.
of Production Research 22, 235–245. Vilarinho, P.M., Simaria, A.S., 2002. A two-stage heuristic
Shtub, A., Dar-El, E.M., 1989. A methodology for the selection method for balancing mixed-model assembly lines with
of assembly systems. International Journal of Production parallel workstations. International Journal of Production
Research 27, 175–186. Research 40, 1405–1420.
Shtub, A., Dar-El, E.M., 1990. An assembly chart oriented Vilarinho, P.M., Simaria, A.S., 2006. ANTBAL: An ant colony
assembly line balancing approach. International Journal of optimization algorithm for balancing mixed-model assembly
Production Research 6, 1137–1151. lines with parallel workstations. International Journal of
Silverman, F.N., Carter, J.C., 1986. A cost-based methodology Production Research 44, S.291–S.303.
for stochastic line balancing with intermittent line stoppages. Visich, J.K., Diaz-Saiz, J., Khumawala, B.M., 2002. Develop-
Management Science 32, 455–463. ment of heuristics to reduce model imbalance for the mixed-
Sniedovich, M., 1981. Analysis of a preference order assembly model, U-shaped assembly line, In: Proceedings of Annual
line problem. Management Science 27, 1067–1080. Meeting Decision Science Institute 2002, 1786–1791.
Sotskov, Y., Dolgui, A., Portmann, M.-C., 2006. Stability analysis Wilhelm, W.E., 1999. A column-generation approach for the
of optimal balance for assembly line with fixed cycle time. assembly system design problem with tool changes. Interna-
European Journal of Operational Research. 168, 783–797. tional Journal of Flexible Manufacturing Systems 11, 177–
Sparling, D., 1998. Balancing JIT production units: The N U-line 205.
balancing problem. Information Systems and Operational Wilhelm, W.E., Gadidov, R., 2004. A branch-and-cut approach
Research 36, 215–237. for a generic multiple-product, assembly-system design prob-
Sparling, D., Miltenburg, J., 1998. The mixed-model U-line lem. INFORMS Journal on Computing 16, 39–55.
balancing problem. International Journal of Production Wilson, J.M., 1986. Formulation of a problem involving assem-
Research 36, 485–501. bly lines with multiple manning of work stations. Interna-
Sphicas, G.P., Silverman, F.N., 1976. Deterministic equivalents tional Journal of Production Research 24, 59–63.
for stochastic assembly line balancing. AIIE Transactions 8, Yano, C.A., Bolat, A., 1989. Survey, development, and applica-
280–282. tion of algorithms for sequencing paced assembly lines.
Steffen, R., 1977. Produktionsplanung bei Fließbandfertigung. Journal of Manufacturing and Operations Management 2,
Gabler, Wiesbaden. 172–198.
Süer, G.A., 1998. Designing parallel assembly lines. Computers Zäpfel, G., 1975. Ausgewählte fertigungswirtschaftliche Optimi-
and Industrial Engineering 35, 467–470. erungsprobleme von Fließfertigungssystemen. Beuth, Berlin.

S-ar putea să vă placă și