JournalofManufacturingSystems5

3(2019)150–158

Contents lists available at

Journal of Manufacturing Systems

journal homepage:

Changeability and flexibility of assembly line balancing as a multi-objective

optimization problem
Johannes Fisel⁎, Yannick Exner, Nicole Stricker⁎, Gisela Lanza
wbk Institute of Production Science, KAISERSTRASSE 12, 76131, KARLSRUHE, GERMANY

ARTICLEINFO
ABSTRACT
Keywords:
Assembly line Current trends, such as customers' demand for individual products and shorter product life cycles, are addressed
balancing by companies through a greater variety of products and variants. With regard to the line balancing of flow
Changeability assembly systems, however, adjustments are associated with high investments, which requires a new planning
Flexibility approach for assembly line balancing. Existing approaches do not consider the reallocation of assembly tasks or
Multi-objective optimization the dimensioning of system-inherent flexibility and changeability according to requirements. Furthermore, they
neglect the uncertainty of the future market situation. The proposed approach aims at optimizing the line
balancing of flow assembly systems, taking into account the potential need for adaptation in order to meet this
uncertain planning environment. For this purpose, the exchange of occurring costs as well as flexibility and
changeability of the system is focused. Based on scenarios, potential future compositions of the variant mix are
investigated and the resulting implications for the assembly system are derived. By applying the approach, an
adequate adaptable assembly line balancing is generated by performing a mixed integer linear optimization.
Since the evaluation and identification of adequacy are subject to subjective factors, several potentially adequate
solutions are generated, which differ in terms of costs, flexibility and changeability. The result of the presented
approach is a front of pareto-optimal assembly line balancing configurations. In order to show its practical
applicability, a use case in automotive assembly line balancing is presented.

1. Introduction
over or under adapt a production system in order to meet the customer
The current environment of industrial production is characterized demand. To mitigate that risk, a production system can be designed in a
by a variety of megatrends, including an increasing customer demand way that options to easily change the production system are integrated.
for individual products [1]. A frequently used strategy to respond to Since changes require effort in equipment, manpower, knowledge and
this market demand is to increase the variety of products and variants time [9], changes need to be kept at a reasonable limit.
[2]. As a result, the pressure for new and further product Automotive assembly typically is organized in the form of a variant
developments is intensifying, resulting in shorter product life cycles flow assembly line [10], where changes represent a significant chal-
[3]. Moreover, this is reinforced by regulations, such as subsidies or lenge. In this organizational form, assembly tasks are carried out at
emission thresh- olds for the automotive industry. The enlarged interlinked stations within a given cycle time [11]. The planning pro-
product portfolio and increased dynamic in customer demand lead to cess for assigning assembly tasks to assembly stations is referred to as
market conditions which can be considered challenging for industrial assembly line balancing [12]. The consideration of various product
companies [4]. However, in order to adapt a company’s production variants is mostly realized by the use of an average variant mix [12].
program, often times corresponding changes in the production Changes in this variant mix can hardly be reacted to once the
system are required. These changes may include production assembly line balancing has been determined [13].
equipment, logistics or work organization and result in corresponding The line balancing of an operating variant flow assembly line is to
costs, which are also subject to the frequency of change. In order to be designed under the premise of providing high efficiency and there-
predict fluctuations in demand different models are available [5,6]. fore economic advantages [14]. However, capacity reserves, such as
Since the future is unknowable, a certain level of uncertainty remains additional assembly stations or extended cycle times for preventive
[7]. An inherent characteristic of uncertainty is that it leads to risk addressing of demand fluctuations, reduce the economic efficiency of
[8], in this case the financial risk to the assembly line [15]. This leads to the challenge of designing the

⁎
Corresponding author.
E-MAIL ADDRESSES: [email protected] (J. Fisel), ni c o l e. s t ri c ker @ ki t .e du (N. Stricker).

https://doi.org/10.1016/j.jmsy.2019.09.012
Received 2 April 2019; Received in revised form 22 September 2019; Accepted 23 September 2019
0278-
6125/©2019TheSocietyofManufacturingEngineers.PublishedbyElsevierLtd.Allrigh
tsreserved.
J. Fisel, et AL. JourNAlofMANUfActuringSys
tems53(2019)150–158

assembly line balancing to be highly efficient, but also maintaining correlating operations [40], the separation into highly flexible and not
adequate capacity reserves in order to adapt the assembly line balan- flexible stations [41], as well as the definition of an absolute and re-
cing to changed requirements such as fluctuations in the variant mix. lative upper limit of variant-specific fluctuating operations per as-
The ability to adapt to changed requirements is described by various sembly station [42]. However, the effects of stochastic dependence of
terms in the literature such as changeability [9] or reconfigurability assembly tasks on the aggregated processing time of a workplace
[16], whereas the term flexibility focusses on maintaining the ability were not explicitly modelled by any approach.
to production under fluctuating influences [17]. The novel approach The design of changeability of assembly line balancing configura-
ad- dresses the presented line balancing challenge of variant flow tions typically focuses on the reallocation of assembly tasks.
assembly lines by introducing a model for integrated optimization of Altemeier
costs, changeability and flexibility regarding scenarios of the future [40] describes an approach to reconfigure flow assembly systems by
variant mix. The characteristic of flexibility hereby describes the reallocating tasks on the basis of negative correlation of the affected
design of an assembly line balancing, which is able to uphold tasks. The evaluation of potential reallocation operations is conducted
production within a certain level of requirement fluctuation without subsequently by an expert estimate, on the basis of a stationary
the need of adjustment. Furthermore, the changeability of a line model of operating resources. Quantitative approaches for the
balancing configuration relates to the organizational and physical evaluation of reallocations could also be identified. Makssoud et al.
reallocation of assembly tasks in- cluding assembly equipment. The [43] evaluate the effectiveness of the change of an assembly line
remainder of the paper is organized as follows. In Section 2, a balancing based on the number of operation reallocations. Gamberini
literature review is presented. Section 3 is devoted to the mixed et al. [44,45] propose a comparable approach by evaluating the
integer linear program applied to obtain changeable and flexible line assembly line balancing before and after a possible change by
balancing configurations. Additionally the evaluation of flexibility and applying a similarity index. Yang et al.
changeability is presented. In Section 4, the computational results are [38] weight the reallocation costs by the aggregated processing time
presented. Finally these results are discussed in Section 5. of the assembly tasks to be shifted as well as the requalification cost
of the employees. The selection of suitable manufacturing equipment
2. Literature review for a changeable assembly line balancing can be determined by
assessing equipment in the dimensions of cost, flexibility and
Assembly line balancing is of high practical relevance in industrial production speed [46].
production. Since the mathematical formulation of the line balancing In conclusion, it can be stated that singular approaches to the
problem by Salveson [18] various research activities have been carried concepts of flexibility and changeability of assembly line balancing
out in this field. The key task of assembly line balancing is to allocate have been developed. However, all reviewed approaches feature lim-
specific assembly tasks and their required resources to a station of the itations. Regarding flexibility, the modeling of stochastic dependency
assembly line, typically with the aim of minimizing cycle time or the of assembly tasks is a novel contribution. Furthermore, with regard to
number of assembly stations. This basic problem is referred to as the changeability, it can be noted that the reconfiguration effort of an as-
Simple Assembly Line Balancing Problem (SALBP) [19]. sembly line balancing is evaluated by means of auxiliary valuations
However, the models of the SALBP are not suitable for modelling such as indices and not by the actual resulting costs. Thus also no
realistic problems in sufficient detail. As a result, in the research area of bottom-up evaluation of the reconfiguration costs on the basis of re-
the General Assembly Line Balancing Problem (GALBP), extensions of allocating assembly tasks and equipment is considered. Finally, it
the basic problem are applied [10]. This includes extensions such as the should be noted that no approach provides an integrated evaluation
assembly of multiple products [12,20], parallel stations [15] or mul- of the trade-off between costs, flexibility and changeability, especially
tiple workers per station [21–23], dynamic [24,25], stochastic [26] and with regard to several future variant mix scenarios.
fuzzy [27] process times, unpaced asynchronous lines [28], resource
rescrictions [29] and U-lines [30–32]. 3. Multi-objective optimization problem
Likewise, approaches can be identified in the literature that
address the flexibility and changeability of assembly line balancing. In In this section the description of the multi-objective optimization
this paper, variant flow assembly lines are focused. If the assembly of model for the design of a changeable and flexible assembly line bal-
multiple variants is conducted in this organizational form, the flex- ancing with regard to different variant mix scenarios s S is given. The
ibility to vary the shares in the variant mix is beneficial. This ability is objective functions are individually presented and subsequently com-
referred to as variant mix flexibility [33] and is subsequently focused. bined in an objective function vector supplemented by problem
The line balancing of multiple variants requires the merging of in- specific constraints. The problem to be solved by optimization is the
dividual assembly precedence graphs into a unified assembly pre- identifi- cation of a greenfield assembly line balancing for a given
variant mix. This scenario is referred to as scenario 0 (hereinafter: s0).
In addition, the line balancing should result in low costs with regard to
defined
cedence graph [12]. For the changes in the variant mix
determination of the processing (scenarios s
times in 0) through flexibility
automotive industry, typically a Boysen hereto presents an
weighted average of the overview of formulations and
individual processing times is applications [15]. Fisel et al.
used [34]. An approach to present an approach that, in
generate variant mix flexibility addition to the expected change,
proposes using the maximum also takes its uncertainty re-
processing time of a variant presented by the standard
instead of using an average deviation into account [36]. The
value [35]. Furthermore, the use balanced allocation of tasks with
of a time premium for assembly the goal of a high flexibility of a
tasks based on an expected line balancing with fluctuating
change in the variant mix variant shares is also the subject
represents a further approach. of research. Thomo- poulos

15
J. Fisel, et AL. JourNAlofMANUfActuringSys
tems53(2019)150–158
presents a basic approach, buffers or cost-effective
which aims at an even reallocation of assembly tasks.
distribution of the processing With this novel approach,
time of each variant [12]. the uncertainty regarding the
Rachamadugu & Talbot [37] future variant mix is already
and Yang et al. [38], on the addressed during greenfield
other hand, demand an even planning of the line balancing
distribution of the utilization of through the concepts of
each assembly station. Further flexibility and changeability. A
approaches are the avoidance substantial improvement is the
of variant-related fluctuations integration of expected future
[39], the separation of line modification costs as a
result of a changed variant mix.
Existing ap- proaches fail to
handle this trade-off from
costs, flexibility and chan-
geability.

3.1. Model of flow ASSEMBLY lines

The organizational structure

of a flow assembly line is based
on a hierarchical system (see
Fig. 1). As formulated in the
SALBP, the flow assembly line is
divided into assembly stations
[19] k K which re- present the
highest hierarchical level.
Within an assembly station the
cycle time C is available for
processing the assigned
assembly tasks

15
 built. Likewise, the costs of workplaces result from the total number of
opened workplaces yj,k,s0 and its cost rate cwp ( 0). The costs of equip-
ment is determined by the total number of required operating resources
rb,k,j,s0 and the specific cost rate of each type of equipment cinv,b ( 0).

fgc =
yk,s0 × cstat ( 0) + yj,k,s0 × cwp ( 0) rb,k,j,s0 × cinv,b (
k Ks0 Jk,s
0)
j + bB
0 (1)

3.3. Flexibility

Fig. 1. Structure of assembly line model.

The ability to assemble at different model mix ratios without in-
curring change costs corresponds to the concept of flexibility as stated
I . An assembly task is a distinct and not further divisible work by Zäh, Möller & Vogl [47], according to which flexibility describes the
element of the total assembly process [19]. The level of a workplace ability of a system to change its output without changing itself, and is
j I is below the level of an assembly station. The definition of more subsequently referred to as flexibility of assembly line balancing.
than one workplace per station results in parallel execution of In the presented approach, the ability to assemble at a specific
assembly tasks and therefore extends the amount of work that can be model mix ratio is considered valid, when the average processing time
handled at one assembly station. It should be noted that the at each assembly station does not exceed the cycle time. Otherwise an
assignment of parallel assembly tasks at different workplaces of an overload of the station occurs. If the demand of flexibility is transferred
assembly station, which are performed simultaneously, results in an to the allocation of assembly tasks to workplaces, this can be inter-
invalid line balancing in case of identical car body positions or preted as avoiding overload. The line balancing based on the average
precedence constraints of the respective tasks. The maximum processing time results in a positive or negative deviation from the
number of workplaces per station is constant for all stations. It is average processing time at the moment of the actual assembly process
important to note that there is no obligation to use all workplaces, depending on the variant, resulting in a workplace-related overload or
thus unused workplaces are taken into account accordingly in the cost idle time [48]. If multiple assembly tasks with model-specific proces-
evaluation. For the execution of assembly tasks, equipment b B may sing times are allocated to a workplace, the effect on the aggregated
be required. If equipment u B also meets the requirements of the processing time of the workplace has to be considered. If the assump-
assembly task, is considered a substitute of b. This equipment is tion applies that the processing time of an assembly task is not de-
assigned to a workplace and therefore also to a station through an as- pendent on other assembly tasks of the same workplace, this reflects
sembly task. Within a workplace, the same equipment can be used by stochastic independence of the assembly tasks. However, this assump-
different assembly tasks. If assembly tasks are reallocated as part of a tion is rejected regarding the assembly in the automotive industry be-
change in assembly line balancing, different options for satisfying the cause there is dependency between assembly tasks e.g. through the
changed equipment demand are to be evaluated (e.g. moving equip- offering of equipment packages. Therefore, it is desired to allocate as-
ment, using equipment of target workplace or buy new equipment). sembly tasks to workplaces in such a way that the influence of fluc-
The following assumptions are established for the optimization of tuations in the model mix on the maintaining of the cycle time is
the assembly line balancing. The processing time of an assembly task minimized.
is described by a discrete value. Learning effects in the early phase The objective function of flexibility therefore extends the approach
after production ramp-up are considered negligible. In addition, all of Fisel et al. [36] by including stochastic dependencies which are re-
resources show one hundred percent availability and the physical flected by the covariance of processing time for pairs of assembly tasks.
reallocation of equipment takes place simultaneously. Due to the one This is illustrated by the following example of two assembly tasks for
hundred percent availability of the equipment, no increased line optional equipment. The processing time of these tasks be either not 0 s
speed is required and the cycle time of all stations is therefore the (optional equipment is ordered) or 0 s (optional equipment is not or-
same. Furthermore, the assembly line balancing only includes dered). If the assembly tasks are part of the same equipment bundle,
assembly stations. Technical stations for the relocation of the vehicle stochastic dependency is assumed, because it is more likely that both
as well as buffer stations are excluded from the consideration. The tasks have to be conducted. In this case, allocating these tasks to the
output quantity of the assembly system per period of time is same workplace results in either 0 s or a high total processing time of
constant. A possible quantity adjustment is to be realized by change the workplace. From this perspective it is recommended to separately
of the working shifts. With regard to the workers, it is assumed that allocate these tasks. The variance of a workplace is hence defined as the
they move at infinite speed. Thus, the time that a worker needs to aggregated variance and the pairwise determined covariance of its al-
return to the start of the station after completing the assembly located assembly tasks. Based on the variance calculation of dependent
activity on a vehicle is considered negligible. random numbers by Laux & Schabel [49], Eq. (2) has been derived to
describe the variance of an assembly line balancing for a scenario . The
3.2. Cost of greenfield ASSEMBLY line formula summarizes the variance of processing time for each workplace
j at each station k by summarizing the variance of all workplace
In this objective function, the focus is exclusively on the costs of allocated assembly tasks (xi,j,k,s = 1) and the covariance of all
assembly line balancing prior to a change. This assembly line combinations of assembly tasks and . In order to obtain an aggregated
balancing is therefore referred to as assembly line balancing for variance evaluation, the consideration of the scenario-specific
scenario 0. weighting of the variance is given by the factor ws in Eq. (3).
The key task of assembly line balancing is the allocation of assembly |Is |
tasks to workplaces of an assembly line system. In order to ensure a
realistic modelling of the assembly process, parallel processing of as- i=1
sembly tasks at up to WP different workplaces of an assembly station is |Is | |Is |
supported. Moreover, the costs for assembly stations as well as for
fflex,s = xi,j,k,s × vari,s + 2 × xi,j,k,s × xv,j,k,s
workplaces and equipment is taken into account for s0 in the target k Ks j Jk,s i=1 v=i+1
function fgc in Eq. (1). The costs of an assembly station are determined
by the total number of opened assembly stations yk,s0 and the cost of an × cov (i;
assembly station cstat ( 0) at time 0 on which the assembly system is v) (2)

fflex = fflex,s × ws variable qtrans,b,s represents the required number of reallocations of

sS
(3) equipment b. If equipment b is to be integrated in the event of a
change by purchase or relocation at workplace j of station k, this is
indicated

3.4. CHANGEABILITY by mint,b,j,k,s. Likewise, mdis,b,j,k,s indicates disintegration. Furthermore, qstat,s

describes the difference amounts of stations between the scenarios s0
Line balancing that is different from one another is characterized and s|s s0 while stat,s describes whether this difference is positive or
by non-identical allocations of assembly tasks and equipment to additional stations are required. A negative indicator is not used as
work- obsolete stations are assumed to have no residual value. The time
places. The transition from an assembly line balancing lbA to a period in which indirect costs of change are incurred depends on the
different
assembly line balancing lbB is referred to as change. This includes the most complex reallocation process of an equipment b which is marked
reallocations of assembly tasks and equipment, as well as the resulting by b,s.
costs. The objective of a changeable design of the assembly line bal-
ancing is to minimize these costs. Thus, the design of a changeable fplan,s = (xi,j,k,s xi,j,k,s0) × xi,j,k,s × cplan,s ( )
assembly line balancing can be related to the problem of the realloca- i I s k Ks j Jk,s (4)
tion of assembly tasks. The ability to adjust assembly line balancing at
the lowest possible cost is therefore referred to as the changeability of finv,s = |qb,s| pos,b,s × cinv,b ( )
assembly line balancing. × (5)
bB
The classification of change costs is based on preliminary work by
Wiendahl and Heger [50]. The reallocation of assembly tasks corre- fdisinv,s = |qb,s| × neg,b,s × cdisinv,b ( )
sponds to the relocation of change objects, so that the change costs are bB
(6)
divided into transformation object costs (TOC) as well as direct (DIC)
and indirect implementation costs (IIC). The resulting modular classi- fint,s = mint,b,j,k,s × cint,b ( )
fication for cost evaluation is visualized in Fig. 2 by the reallocation of bBk Ks |s 0 Jk,s |s 0 (7)
j
the assembly task i from a workplace at assembly station A to a
fdisint,s = mdis,b,j,k,s × cdisint,b ( )
workplace at assembly station B. The values 0 and 1 in brackets indicate
the state before (0) and after the change (1). The transformation object b B k Ks0 j Jk,s0 (8)
costs consist of investment costs cinv and disinvestment costs cdisinv of
assembly equipment, whereby both positive and negative cash flows ftrans,s = qtrans,b,s × ctrans,b ( )
can occur with regard to disinvestments depending on the remaining bB
(9)
value. The direct implementation costs describe the costs directly as-
sociated with the change process. They are represented by the costs of flost,s = b,s
× clost,b ( )
integration cint , disintegration cdisint , transport ctrans and planning cplan. bB
(10)
Disintegration describes the removal of equipment from the flow as-
fadd,s = qstat,s × stat,s × cstat ( )+ yj,k × cwp ( )
sembly system, whereas integration describes its addition. If a
k Ks |s 0 j 0 (11)
removed equipment is to be moved, the costs of transport are Jk,s |s

applied. Organi-
zational planning of the reallocation of an assembly task also results in f = (f +f +f +f +f +f +f
costs that are independent of equipment. These costs are grouped under ch plan,s inv,s desinv,s int,s desint,s trans,s
lost,s s
planning costs. The indirect implementation costs result indirectly from S
+ fadd,s ) × ws (12)
the change process and are referred to as lost profits clost .
The expected time of the change is taken into account when eval-
uating changeability. The resulting change costs are therefore de-
scribed as functions of the point in time of change c ( ). Eqs. (4)–(12) 3.5. NOTATION of the OPTIMIZATION
describe functions f for determining the respective subordinate costs. problem
In this context, the variable qb,s denotes the difference in equipment, The objective functions described in sections 3.2 to 3.4 represent
whereby the binary variable pos,b,s indicates that additional equipment the multi-objective optimization problem and are grouped together in
b is required and neg,b,s indicates that less equipment is required. The Eq.
(13) to form the objective function vector z .
 Fig. 2. Modular cost model of change costs for assembly equipment.

fgc in scenario needs to be determined for use within the objective

min z function (37). In order to limit the workplaces per station, the max-
fflex
= imum represented by the parameter WP must not be exceeded (38).
fch (13) An overview of the used notation is given in Appendix A.1.

The presented objective functions are subject to the following con-

straints. The basic requirement to fulfil all tasks is met by setting the xi,j,k,s = 1 i I s; s S
binary variable xi,j,k,s to a value of exactly once per task and scenario k K j Jk,s (14)
s
(14). Further, all allocated tasks per workplace and assembly station xi,j,k,s × trobust,i,s C j Js; k K s; s S
need to be performed within the externally given cycle time C (15). For i (15)
Is
all task pairs that have a mutual precedence relation (h, i) Ivr,s, it is to
be ensured that the successor is allocated to the same or a succeeding (k × xh,k,s) (k × xi,k,s) 0 (h, i) Ivr,s ; k Ks (16)
assembly station k than its predecessors h (16). In case of a task allo-
cation to an assembly station k and a workplace j, both, the assembly xi,j,k,s M × yj,k,s × yk,s 0 j J s; k K s; s S
station and the workplace need to be opened yk,s = yj,k,s = 1. Therefore, i Is (17)
a sufficiently large number M > |Is is implemented (17).
In order to secure the precedence relations, the variable gh,i,j,k,s is xh,j,k,s + xi,j,k,s gh,i,j,k,s 1 j Js; k Ks ; s S; h, i Is |h
utilized. This binary variable is set to its required values according to i
the allocation of task h and (18–19). To minimize worker walking (18)
times, tasks such as h and that are allocated to the same car body xi,j,k,s + gh,i,j,k,s 0 j Js; k Ks ; s S; h, i Is |h i (19)
position (h, i) IKA,s are required to be performed at the same work-
place in the correct order if allocated to the same assembly station (20).
WP
Allocation to the same workplace in case of a shared assembly station xi,j,k,s + xh,j,k,s gh,i,j,k,s 1 (h, i) IKA,s; k K s; s S
holds for all tasks that have a mutual priority relation, too. The variable j=1 (20)
gh,i,j,k,s takes a value of in order to ensure this (21).
Eq. (22) sets the allocation variable bmb,i of a specific assembly WP x +x g 1 (h, i) I ;k K; s S

equipment b for a specific task to 0, in case that b is not intended or i,j,k,s h,j,k,s h,i,j,k,s vr,s s
needed by . Therefore, a two-dimensional matrix Bmb,u,i with b, u B (21)
j=1

has been provided externally stating whether equipment type b can be

substituted by equipment type for a certain task . Hereto, it is ensured bmb,i = 0 b B; i Is | Bmb,u,i = 0
that exactly one out of all possible substitutes is used for task (23). uB
(22)
Given these pre-requisites, the model allocates the assembly equipment 1
to workplaces according to the chosen substitute in Eq. (23) by using bmu,i = 1 b B; i Is | Bmb,u,i

the binary variable rb,j,k,s (24). Again, a sufficiently large number u B|Bmb,u,i=1 (23)
uB
M > |Is is used (24). bm ×x M×r 0 b B; j J; k K; s S

The method allows to adapt the line balancing to different scenarios. b,i i,j,k,s b,j,k,s s s
i sI
Therefore, allocations of assembly equipment might change according
to different scenarios which results in disintegrating and integrating (24)
cost within the objective function, activated by the binary variables 0
mdis,b,l,m,s and mint,b,j,k,s. For all scenarios s s0 that differ from the in- itial rb,j,k,s + mdes,b,l,m,s 1
scenario s0 (mdis,b,l,m,s = 1), it is ensured that assembly equipment is not
allocated at the former workplace j anymore after moving it to a different
workplace (25). Vice versa, Eq. (26) ensures that the moved equipment is
allocated at the new workplace, if equipment gets in- tegrated (mint,b,j,k,s = rb,j,k,s + mint,b,j,k,s
1).
b B; j Js0; l Js |l = j; k Ks0; m
Ks |m = k; s S| s s0
(25)

b B; j J s; k K s; s S|s s0 (26)
 The variable qb,s is set in Eq. (27) and represents the difference in rb,j,k,s rb,j,k,s0 b B; j Js; k Ks; s S|s s0
total demand of equipment type b after a change in scenarios. Binary k Ks j Jk,s k Ks0
j Jk,s
0

variables pos,b,s, neg,b,s and eq,b,s state whether the difference qb,s is positive,
qb,s = 0
negative or zero (28–30). Following that definition, exactly one of the
three binary variables must take a value of (31). (27)
The number of physical reallocations of assembly equipment that is
neither additionally acquired nor discarded during the adaption to a
rb,j,k,s0 pos,b,s rb,j,k,s b B; s S|s s0
different scenario is modelled by the variable qtrans,b,s (32). Further, binary
variable b,s indicates whether physical reallocation takes place for
k Ks 0 × k K j Jk,s
s
equipment type b during a shift from s0 to (33). Given the model’s 1j Jk,s0

assumption of simultaneous reallocation of assembly equipment during (28)

the change to a new scenario, the cost (lost profits due to time for re-
allocation) is determined by the most expensive (most time consuming)
equipment type b. This equipment type is determined in Eq. (34) rb,j,k,s neg,b,s rb,j,k,s0 b B; s S|s s0
through mutual comparison of all clost,b and stored within binary vari- k Ks × k Ks 0 j Jk,s0

able b,s by setting its value to 1 for only the most expensive equipment 1 j Jk,s

type (35). (29)

Due to different scenarios, the total number of assembly stations
opened can differ. The variable qstat,s is defined as the additional |qb,s| eq,b,s × M eq,b,s 0 b B; s S|s s0 (30)
×
amount of assembly stations needed in scenario s s0 (36). Further, a
binary indicator stat,s showing additional demand for assembly stations pos,b,s + neg,b,s + eq,b,s = 1 b B; s S|s s0 (31)
 (mint,b,j,k,s
|qb,s|) × pos,b,s + (mdisint,b,j,k,s |qb,s|) b B; s S|s s0

k Ks j Jk,s × neg,b,s + mint,b,j,k,s × eq,b,s

qtrans,b,s = 0

(32)
( b,s + qtrans,b,s ) (M b,s × qtrans,b,s) 0 b B; s S|s s0
× (33)
( b,s × lost,u × u,s
c ) ( b,s × lost,b × b,s
c ) 0
b B ; u B|u b;s S|s s0
Fig. 3. Scenarios for future line balancing changes.
(34)

b,s
=1
s S|s s0 constraints [52]. In this implementation of NSGA-II, the crossover op-
bB (35) erator for continuous problems is extended by a rounding rule for in-
yk,s tege problems. This type of reproduction is not suitable for assembly
yk,s0 qstat,s = 0 k Ks ; s S|s s0
k Ks0 line balancing problems, as this type of reproduction rounds off con-
k Ks (36)
tinuous variables, creating invalid solutions [53]. Therefore, the
( stat,s + qstat,s) (M stat,s × qstat,s) 0 b B; s S|s s0 (37) crossover operator is replaced by the one-point crossover operator
based on Bäck et al. [54]. The computational complexity of the NSGA-II
×
yj,k,s WP k K s; s S is O(MN²) where M in this context is the number of objectives and N
j Jk,s (38) the population size [55].
The results of the optimization problem contain a line balancing
configuration for each scenario and are therefore referred to as
4. Use case solution clusters. Fig. 4 shows the solution clusters with line balancing
solutions

The assembly line balancing is to be planned for a compact class 1 Focus ICE 100;0;0 0;0;0 10
2 High BEV 40;60;0 4;4;0 35
car with three different drive concepts based on real automotive
3 Multi Concept 50;25;25 5;5;5 10
industry data, collected in the research project “Energieeffiziente und
flexibel industriell herstellbare Elektrofahrzeugantriebe (16EMO0064
K)”. The drive concepts contain an internal combustion engine (ICE),
battery electric vehicle (BEV) and a bivalent natural gas engine
(bNGV). The different drive concepts are realized by a conversion
design and ac- cordingly integrated into existing flow assembly line
systems if re- quired. However, this leads to partially heterogeneous
processing times and equipment requirements as well as to
completely different as- sembly tasks, mainly due to different
technical properties. The case study contains one hundred assembly
tasks and seven types of assembly equipment.
In order to define the required flexibility and changeability, it is
necessary to define the possible future market developments in the form
of scenarios. The presented use case is based on a current scenario s0
(low BEV) and three future scenarios s1 to s3 (focus ICE, high BEV, multi
concept). These scenarios are formalized by normal distributions as
representations of the shares in the model mix. Table 1 shows the
quantitative characteristics of the expected value in the model mix μ,
the standard deviation σ and the probability of occurrence P of the
scenarios presented .
The scenario s0 describes the model mix of the initial greenfield
planning, whereas the scenarios s1 to s3 represent alternative changes of
the line balancing. Taking into account the probabilities of occurrence
Ps, the scenario tree of the planning problem shown in Fig. 3 results.
The complete data basis including processing times of assembly
tasks, precedence constraints, car body access areas, assembly equip-
ment and costs can be found in [51]. Additionally, a detailed pre-
sentation of results is included.
The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is used
according to Lin to calculate the solution clusters, since, in contrast to
MathWorks MATLAB’s implementation, it allows the usage of

Table 1
Planning scenarios of drive concepts (ICE; BEV; bNGV).

s name μ [%] σ [%] Ps [%]

0 Low BEV 80;20;0 3;3;0 45

for scenarios s0 to s3. In order to obtain a first evaluation of the solu- equipment 1, 3 and 6 which is allocated to station 11. In case of
tions, pareto fronts have been calculated. As shown, the solutions changing the line balancing from s0 to s1 or s2, no DIC or IIC occur from
are assigned to ten different pareto fronts. A rational decision maker reallocating this equipment.
should focus on solution clusters of the first pareto front. The presentation of the change costs focuses on TOC and DIC. For
An exemplary detailed view on the solution cluster 47, which is part a detailed breakdown of all change costs, readers are referred to [51].
of the first pareto front in Fig. 4, is given in Appendix A.2. A solution The physical relocation of equipment is evaluated by the TOC of the
cluster contains a line balancing for each scenario s0 to s3. Within a line investment (inv) and disinvestment (disinv) as well as the DIC of
balancing, the allocation of assembly tasks to a station and to one of its planning (plan), transport (trans), integration (int) and disintegration
two workplaces is shown by the number of the task (1 to 100). Also the (disint). Fig. 4 shows solution clusters that each contain a line
allocation of equipment to a workplace is given in the row b by the balancing for scenarios 0, 1, 2 and 3 (see Fig. A.1). Change costs occur
numbers 1 to 7. A positive aspect is the equipment allocation, which is when the assembly line balancing of scenario 0 is changed to the
displayed in Fig. A.1 in the row b for each workplace and every assembly line balancing of another scenario. Fig. 5 covers the change
sce- costs TOC and DIC of all solution clusters shown in Fig. 4 and
nario. The industrial manipulator (b = 1) with a load suspension device describes their dis-
for a door (b = 2) or cockpit (b = 3) and the stationary screw tech- tribution by quantiles. This is applied for the changes from scenario 0 to
nology (b = 6) are the most cost intensive equipment regarding both 1 (Fig. 5 s1), 0 to 2 (Fig. 5 s2), and 0 to 3(Fig. 5 s3). Here it is noticeable
the DIC and the IIC [51]. As can be seen in Fig. A.1 allocation of the that comparatively high investment costs in the change to scenario s1
industrial manipulator with the load suspension device for doors (sta- are not balanced by earnings from disinvestments, which implies an
tion 1 and 2) is the same for scenarios s0, s1 and s2. Hereby it should be additional need for equipment. This also results in an imbalance in
noted that assembly tasks 1 and 2 in s3 cannot be allocated to a integration and disintegration costs. With regard to the DIC, the plan-
common station due to the processing time constraints. The same ning costs represent the dominant cost driver in each scenario, whereby
applies to the planning effort in the change to scenario s3 causes the highest costs.
 Fig. 4. Pareto fronts as a result of the multi objective optimization problem.

5. Conclusion processing times is recommended, because the optimization algorithm

decides (based on these times) on the adherence to the cycle time and
The presented novel approach allows the design of an assembly therefore does not consider subsequent assembly process improve-
line balancing considering future scenarios as well as the preference ments.
of a decision maker in the objective conflict of costs, flexibility and
chan- geability. The approach can also be used to evaluate existing 5.2. COMPARISON with line BALANCING RELATED PUBLICATIONS in the
assembly line balancing configurations regarding these objectives.
JOURNAL of MANUFACTURING Systems (JMS)
This requires a manual formulation of the data basis and its
integration into the opti- mization model by defining the assembly
The presented approach is in line with the assembly line balancing
line balancing for scenario s0. Furthermore, this approach can also be topics of the Journal of Manufacturing Systems. The approaches pub-
applied to evaluate the change costs for existing assembly line lished in this journal can be categorized into extensions of the SALBP
balancing configurations for scenarios and algorithmic improvements.
that have not yet been considered. Here, the data basis of the scenario The extensions of the SALPB are carried out by adapting the pro-
must also be adjusted accordingly. blem constraints or the target function. For the adaptation of the pro-
blem constraints, contributions to the U-line [30–32], minimization of
5.1. Discussion the cycle time using robots [56], fuzzy assembly times [27], multiple
workers per station [21–23] and unpaced asynchronous lines [28] are
In order to determine valid solutions, the optimization problem identified. The presented approach offers links to fuzzy assembly
requires considerable computational power due to its large solution times and multiple workers per station. The fuzzy modeling of the
space. The modelling of the car body access areas and precedence processing times implies that those are not used as constant planning
constraints has a significant influence on the composition of the solu- parameters. In the approach presented, the processing time of an
tion space. If the start and end times have to be explicitly determined assembly task results from the proportionally weighted processing
for each assembly task within the station, this typically requires a times of different product and optional equipment variants. A change
dedicated decision variable, which in turn leads to a significant increase in the variant mix thus leads to a change in the planning parameter
in the solution space, depending on the granularity of the time scale. processing time. In- stead of applying fuzzy modelling, the presented
Another key influence on the solution space is the granularity of as- approach tackles the uncertainty of the processing time by the
sembly tasks. If, for practical reasons, assembly tasks are to be carried objective function of flex- ibility which takes the stochastic
out immediately one after the other, modelling as a joint assembly task dependence of assembly tasks into account. The extension of the
is an effective way of limiting the size of the solution space. SALPB for assigning multiple workers to one station is also applied in
Furthermore, the quality of the optimization result depends on the the presented approach. The upper limit of workers per station is an
quality of the underlying input data. The use of series production input variable and is defined by the parameter

Fig. 5. Overview of TOC and DIC.

WP. In contrast to the approach of Sungur & Yavuz [23], the same
qualification is assumed for all workers. Pearce et al. [22] introduce Acknowledgements
grouped tasks that are created prior to the optimization process from
the assembly precedence graph of a single product and are executed The authors would like to thank the Federal Ministry of Education
by the same worker. The presented approach does not create these and Research for the kind support and funding within the project
groups of operations explicitly, but implicitly through the same car “Energieeffiziente und flexibel industriell herstellbare
body po- sition (20) and precedence dependency between tasks (21). Elektrofahrzeugantriebe (16EMO0064K)”.
The objective functions of the approaches presented in the JMS
are predominantly mono-criterial [e.g. 22,23,30,41] or brought to a Appendix A. Supplementary data
mono- criterial form by combining the objectives [e.g. 21,29,57,58].
Typical objectives are to minimize the number of stations, workers Supplementary material related to this article can be found, in the
and cycle time. However, approaches with multi-criteria objective online version, at doi:https://doi.org/10.1016/j.jmsy.2019.09.012.
functions could also be identified [e.g. 27,56,59]. The present
approach uses a multi-criteria objective function in order not to References
preselect a weighting of the objectives and thus to enable sensitivity
studies relevant for in- dustrial applications. As is made clear by the [1] Westkämper E, Löffler C. Implementierung von Strategien in das System
articles published in the JMS, the flexibility [e.g. 41,60] and Produktion. In: Westkämper E, Löffler C, editors. Strategien der Produktion. Berlin
adaptability respectively re- configurability [e.g. 61,16] of production Heidelberg: Springer; 2016. p. 239–89.
[2] Vogel W. Complexity managed approach for resource planning in variant-rich
systems are essential com- petitive factors. In the presented product development. In: Bode C, Bogaschewsky R, Eßig M, Lasch R, Stölzleeditors.
approach, the flexibility and change- ability of assembly line balancing Supply management research. Wiesbaden: Springer; 2017. p. 83–128.
are formulated as separate objective functions. Thus, an optimization [3] Lee SM, Olson DL, Trimi S. Co-innovation: convergenomics, collaboration, and co-
creation for organizational values. Manage Decis 2012;50(5):817–31.
of the assembly line balancing is performed with a focus on the [4] Christopher M. The agile supply chain: competing in volatile markets. Ind Mark
uncertainty of the future variant mix. Manag 2000;29(1):37–44.
To solve the line balancing problem, the approaches identified in [5] Ghobbar AA, Friend CH. Evaluation of forecasting methods for intermittent parts
demand in the field of aviation: a predictive model. Comput Oper Res
the JMS preferably use heuristics, which are in some cases variants of
2003;30(14):2097–114.
genetic algorithms [e.g. 31,32]. However, further algorithms for solving [6] Zhao X, Xie J, Leung J. The impact of forecasting model selection on the value of
the line balancing problem are presented, such as a hybrid ant colony information sharing in a supply chain. Eur J Oper Res 2002;142(2):321–44.
[7] Stacey RD. Managing the unknowable: strategic boundaries between order and
algorithm [59], a hybrid bee colony algorithm [62], a branch and
chaos in organizations. San Francisco: John Wiley & Sons; 1992.
bound algorithm [30], a branch and bound based decomposition [63], [8] Morse E, Dantan JY, Anwer N, Söderberg R, Moroni G, Qureshi A, et al.
multi-objective evolution strategies [56] or a simulated annealing al- Tolerancing: managing uncertainty from conceptual design to final product.
gorithm [57]. In addition, multi-stage procedures are presented. Chen CIRP Ann 2018;67(2):695–717.
[9] Wiendahl HP, ElMaraghy HA, Nyhuis P, Zäh MF, Wiendahl HH, Duffie N, et al.
presents a procedure in which feasible solutions are identified in the Changeable manufacturing-classification, design and operation. CIRP Ann
first step and subsequently the optimal solution is identified by a si- 2007;56(2):783–809.
mulated annealing algorithm [21]. Chen et al. enhance this method and [10] Boysen N, Fliedner M, Scholl A. Production planning of mixed-model assembly
lines: overview and extensions. Prod Plan Control 2009;20(5):455–71.
replace the simulated annealing algorithm by a genetic algorithm [29]. [11] Papadopoulos CT, O’Kelly ME, Vidalis MJ, Spinellis D. Analysis and design of dis-
This method is again developed further by Chen et al. by replacing the crete part production lines. New York: Springer; 2009.
procedure for the identification of feasible solutions by a priority rule- [12] Thomopoulos NT. Mixed model line balancing with smoothed station assignments.
Manage Sci 1970;16(9):593–603.
based method [58]. Pearce et al. present a heuristic combining ranked
[13] Scholl A. Balancing and sequencing of assembly lines. Heidelberg: Physica-Verlag;
positional weighting, last-fit-improvement and an iterative blocking 1999.
scheme [22]. Since the scope of the presented approach is on opti- [14] Hu SJ, Ko J, Weyand L, ElMaraghy HA, Lien TK, Koren Y, et al. Assembly system
mizing the changeability and flexibility of a line balancing configura- design and operations for product variety. CIRP Ann 2011;60(2):715–33.
[15] Boysen N. Variantenfließfertigung. Wiesbaden: Deutscher Universitätsverlag; 2005.
tion according to the level of uncertainty regarding the future variant [16] Koren Y, Shpitalni M. Design of reconfigurable manufacturing systems. J
mix, a variant of the frequently used genetic algorithm is applied. Manuf Syst 2010;29(4):130–41.
[17] de Toni A, Tonchia S. Manufacturing flexibility. A literature review. Int J Prod
Res 1998;36(6):1587–617.
5.3. Outlook [18] Salveson ME. The assembly line balancing problem. J Ind Eng 1955;6(3):18–25.
[19] Baybars I. A survey of exact algorithms for the simple assembly line balancing
problem. Manage Sci 1986;32(8):909–32.
The following approaches can be applied to add further detail to [20] Dörmer J. Produktionsprogrammplanung bei variantenreicher Fließproduktion:
the evaluation of assembly line balancing configurations. The Untersucht am Beispiel der Automobilendmontage. Wiesbaden: Springer Gabler;
2013.
objective function of flexibility suppresses the formation of
[21] Chen Y-Y. A hybrid algorithm for allocating tasks, operators and workstations
workplaces with a high time spread and takes the stochastic in multi-manned assembly lines. J Manuf Syst 2017;42:169–209.
dependence of assembly tasks into account. An evaluation of the [22] Pearce B, Antani K, Mears L, Funk K, Mayorga M, Kurz M. An effective integer
resulting flexibility can be achieved by a subsequent analysis of program for a general assembly line balancing problem with parallel workers and
additional assignment restrictions. J Manuf Syst 2019;50:182–92.
flexibility corridors [27]. For this purpose, the variant mix of the [23] Sungur B, Yavuz Y. Assembly line balancing with hierarchical worker assignment.
respective scenarios can be systematically varied and evaluated in J Manuf Syst 2015;37:290–8.
terms of assembly feasibility. Furthermore, by con- sidering the costs [24] Boucher TO. Choice of assembly line design under task learning. Int J Prod
Res 1987;25(4):513–24.
of compensating for overload by, for example, sup- port staff, the [25] Chakravarty AK. Line balancing with task learning effects. IIE Trans
evaluation of assembly line balancing can be further improved. This 1988;20(2):186–93.
evaluation is to be carried out subsequently to the op- timization, [26] Tempelmeier H. Practical considerations in the optimization of flow
production systems. Int J Prod Res 2003;41(1):149–70.
since concrete instances of the variant mix are to be ana- lyzed for
[27] Kalayci C, Hancilar A, Gungor A, Gupta S. Multi-objective fuzzy disassembly line
each scenario. Multi attribute decision making methods can be used balancing using a hybrid discrete artificial bee colony algorithm. J Manuf Syst
to select one of the generated line balancing alternatives. 2015;37:672–82.
[28] Lopes T, Michels A, Sikora C, Magatão L. Balancing and cyclical scheduling of
asynchronous mixed-model assembly lines with parallel stations. J Manuf Syst
Declaration of Competing Interest 2019;50:193–200.
[29] Chen Y-Y, Cheng C-Y, Li J-Y. Resource-constrained assembly line balancing pro-
blems with multi-manned workstations. J Manuf Syst 2018;48:107–19.
The authors declare that they have no known competing financial [30] Ogan D, Azizoglu M. A branch and bound method for the line balancing problem
interests or personal relationships that could have appeared to influ- in U-shaped assembly lines with equipment requirements. J Manuf Syst

ence the work reported in this paper.

 2015;36:46–54.
[31] Rabbani M, Kazemi S, Manavizadeh N. Mixed model U-line balancing type-1 multi-objective optimization problem, Mendeley Data, v1; 2018. https://doi.org/
pro- blem: a new approach. J Manuf Syst 2012;31(2):131–8. 10.17632/pzv79h76jc.1.
[32] Tiacci L. Mixed-model U-shaped assembly lines: balancing and comparing with [52] Lin S. NGPM – a NSGA-II program in matlab v1.4. 2011 [Accessed 16 December
straight lines with buffers and parallel workstations. J Manuf Syst 2017;45:286– 2018]. https://de.mathworks.com/matlabcentral/fileexchange/31166-ngpm-a-
305. nsga-ii-program-in-matlab-v1-4.
[33] Olhager J. Manufacturing flexibility and profitability. Int J Prod Econ [53] Chong KE, Omar MK, Bakar NA. Solving assembly line balancing problem using
1993;30:67–78. genetic algorithm with heuristics-treated initial population. Proc World Congr Eng
[34] Keckl S, Kern W, Abou-Haydar A, Westkämper E. An analytical framework for 2008;2:1273–7.
handling production time variety at workstations of mixed-model assembly lines. [54] Bäck T, Fogel DB, Michalewicz Z. Evolutionary computation 1: basic algorithms
Procedia CIRP 2016;41:201–6. and operators. Boca Raton, New York, Oxon: Taylor & Francis; 2000.
[35] Aissi H, Bazgan C, Vanderpooten D. Min–max and min–max regret versi-ons [55] Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective
of combinatorial optimization problems: a survey. Eur J Oper Res genetic algorithm: NSGA-II. IEEE Trans Evol Comput 2002;6(2):182–97.
2009;197(2):427–38. [56] Yoosefelahi A, Aminnayeri M, Mosadegh H, Ardakani H. Type II robotic
[36] Fisel J, Exner Y, Stricker N, Lanza G. Variant flexibility in assembly line balancing assembly line balancing problem: an evoluation strategies algorithm for a multi-
under the premise of feasibility robustness. Procedia CIRP 2018;72:774–9. objective model. J Manuf Syst 2012;31(2):139–51.
[37] Rachamadugu R, Talbot B. Improving the equality of workload assignments in [57] Abdolr Roshani, Ar Roshani, Abdolh Roshani, Salehi M, Esfandyari A. A
as- sembly lines. Int J Prod Res 1991;29(3):619–33. simulated annealing algorithm for multi-manned assembly line balancing
[38] Yang C, Gao J, Sun L. A multi-objective genetic algorithm for mixed-model as- problems. J Manuf Syst 2013;32(1):238–47.
sembly line rebalancing. Comput Ind Eng 2013;65(1):109–16. [58] Chen J, Chen Y-Y, Chen T-L, Kuo Y-H. Applying two-phase adaptive genetic algo-
[39] Matanachai S, Yano CA. Balancing mixed-model assembly lines to reduce rithm to solve multi-model assembly line balancing problems in TFT-LCD
work overload. IIE Trans 2001;33(1):29–42. module process. J Manuf Syst 2019;52:86–99.
[40] Altemeier S. Kostenoptimale Kapazitätsabstimmung in einer getakteten [59] Zha J, Yu J-J. A hybrid ant colony algorithm for U-line balancing and
Variantenfließlinie Paderborn: Dissertation 2009. rebalancing in just-in-time production environment. J Manuf Syst
[41] Nazarian E, Ko J, Wang H. Design of multi-product manufacturing lines with the 2014;33(1):93–102.
consideration of product change dependent inter-task times, reduced changeover [60] Asadi N, Jackson M, Fundin A. Implications of realizing mix flexibility in assembly
and machine flexibility. J Manuf Syst 2010;29(1):35–46. systems for product modularity – a case study. J Manuf Syst 2019;52:13–22.
[42] Hazır Ö, Dolgui A. Assembly line balancing under uncertainty. Robust optimization [61] Maganha I, Silva C, Ferreira L. Understanding reconfigurability of
models and exact solution method. Comput Ind Eng 2013;65(2):261–7. manufacturing systems: an empirical analysis. J Manuf Syst 2018;48:120–30.
[43] Makssoud F, Battaïa O, Dolgui A. An exact method for the assembly line re-bal- [62] Akpinar S, Baykasoglu A. Modelling and solving mixed-model assembly line
ancing problem. Prabhu V, Taisch M, Kiritsis D, editors. Advances in production bal- ancing problem with setups. Part II: a multiple colony hybrid bees
management systems. Sustainable production and service supply chains, vol. 414. algorithm. J Manuf Syst 2014;33(4):445–61.
IFIP Advances in Information and Communication Technology; 2013. p. 159–66. [63] He C, Guan Z, Gong Y, Wang C, Yue L. Automated flexible transfer line design
[44] Gamberini R, Gebennini E, Grassi A, Regattieri A. A multiple single-pass heuristic problem: sequential and reconfigurable stages with parallel machining cells.
algorithm solving the stochastic assembly line rebalancing problem. Int J Prod J Manuf Syst 2019;52:157–71.
Res 2009;47(8):2141–64.
[45] Gamberini R, Grassi A, Rimini B. A new multi-objective heuristic algorithm for Johannes Fisel holds a M.Sc. degree in Industrial Engineering and Management from KIT.
solving the stochastic assembly line re-balancing problem. Int J Prod Econ Currently, he is a research associate at the wbk Institute of Production Science at KIT
2006;102(2):226–43. focusing on flexibility and changeability in production system planning.
[46] Bukchin J, Tzur M. Design of flexible assembly line to minimize equipment cost.
IIE Trans 2000;32(7):585–98.
Yannick Exner received his B.Sc. in Industrial Engineering and Management from KIT
[47] Zäh MF, Möller N, Vogl W. Symbiosis of changeable and virtual produc-tion‐the
and is currently engaged in his M.Sc. degree. He worked as a student research assistant
emperor’s new clothes or key factor for future success. Proceedings (CD) of the
at the wbk Institute of Production Science where he focused on adaptive production
international conference on changeable, agile, reconfigurable and virtual produc-
systems and agile factory planning.
tion 2005.
[48] Boysen N, Fliedner M, Scholl A. A classification of assembly line balanc-ing
pro- blems. Eur J Oper Res 2007;183(2):674–93. Dr.-Ing. Nicole Stricker is chief engineer of the group “Production Systems” at the wbk
[49] Laux H, Schabel MM. Subjektive Investitionsbewertung, Marktbewertung und Institute of Production Science at KIT. Focus topics of her research activities are robust
Risikoteilung. Grenzpreise aus Sicht börsennotierter Unternehmen und in- production systems and machine learning.
dividueller Investoren im Vergleich. Berlin Heidelberg: Springer; 2008.
[50] Wiendahl H-P, Heger CL. Justifying changeability. A methodical approach Prof. Dr.-Ing. Gisela Lanza holds the Professorship “Production System and Quality
to achieving cost effectiveness. J Manuf Sci Prod 2004;6(1–2):33–40. Management” at KIT and is head of the wbk Institute of Production Science. Her division
[51] [dataset] [51] J. Fisel, Changeability and flexibility of assembly line balancing as a mainly deals with the areas of global production strategies, production system planning,
and quality management in research, industrial application, and teaching.