Professional Documents
Culture Documents
A R T I C LE I N FO A B S T R A C T
Keywords: Assembly line planning is the interface between agents of different professions, competencies, skills and ex-
Collaboration periences. Each agent is characterized by specific objectives and constraints, that must be considered when
Assembly line design different products and processes are merged across multiple levels of various decision making units. A systemic
Assembly line planning platform for the dependencies of goals, constraining relationships of elements to be assigned and the inter-
Collaborative planning
dependencies to properties of the elements is designed. Hereby, the objectives and the constraints are evaluated
Collaborative problem modelling
and mapped within the assembly line planning (ALP) process, resulting in a comprehensive graph with the
processes, structures, resources and their properties with its dependencies. The dependencies include time cri-
teria but also technical, processual, ergonomic, qualitative, part-dependent along with resource-dependent,
structural constraints. The dependency map is integrated within an existing ALP process to achieve a temporal
precedence on the incoming and outgoing information resulting in a dynamic process oriented presentation of
the objectives and constraints. The elements of the dependency map are categorized and analysed based on the
evaluations and a problem model is created for the presented ALP problem. The problem model is solved using a
genetic algorithm which is parameterized in accordance to the problem model complexity and characteristic.
⁎
Corresponding author
E-mail addresses: daniel.neumann@wirtschaft.tu-chemnitz.de (D. Neumann), jan.keidel@wirtschaft.tu-chemnitz.de (J. Keidel).
https://doi.org/10.1016/j.rcim.2018.05.002
Received 8 June 2017; Received in revised form 13 April 2018; Accepted 2 May 2018
0736-5845/ © 2018 Elsevier Ltd. All rights reserved.
Please cite this article as: Neumann, D., Robotics and Computer Integrated Manufacturing (2018), https://doi.org/10.1016/j.rcim.2018.05.002
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
2
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
Table 2 ▪ i before j
Input parameters for industrial assembly line design. ▪ i after j,
Input section Relevant planning data ▪ i parallel j,
▪ i assigned to j = j has assigned i
Assembly line section sub-assembly lines; cycle times ▪ i part of j = j consists of i,
materials/product handling technology
▪ i depends on j
Station machinery/manufacturing equipment offered
product height
Workplace assembly zone with different weights:
work force qualification
Task task description; task time; product/model ▪ must (not),
resources (machinery/manufacturing equipment)
▪ should (not),
container size per part/task
assembly location/assembly height/assembly zone ▪ independent,
Assembly process precedence relation
required parts and tasks and distance:
Logistics container size offered per part
equipment/container bundling
▪ i directly precede j: Position i = Position j − 1
▪ i precede by n elements j: Position i = Position j − n
planners in form of documents within the innovation process. These ▪ i assigned to j on nodeLevel k
planners give their restrictions and data to the subsequent innovation
process steps in form of documents. In case any changes must be made, The nodes represent planning elements, classes and properties. A
the planners will focus on the received documentation, regardless of planning element can be a task, a machine, a station or anything else
further possible changes to other areas of interest. All created docu- relevant to the assembly line planning process. A class represents a
ments are directly or indirectly linked to the individual objectives of the stereotype of the planning elements. These relations are used to model
participants within the innovation process, as shown in Table 2. the relation between elements and subsequently also the element
Due to the search methods for the underlying assembly line balan- classes as shown in Fig. 1. This process initiates the assembly line
cing problems (ALBP) for new products, it is common, that a pre- planning project and defines the content for the collaborative modelling
cedence relation is modelled. The precedence graph is an essential part and the solution evaluation system.
of the assembly line balancing, and many agents are directly or in- The assembly planning process in the traditional sense consists of
directly involved in the restriction model. Therefore, a collaborative following element classes, which are considered for the planning pro-
precedence modelling technique is presented in Fig. 1. The idea is that cess:
not only the precedence relations defined in the construction process
are used, but also preferences of the planner as well as restrictions are ▪ structure element,
being considered. The upfront modelling of this information accelerates ▪ process element,
the iterations allowing an earlier detection of conflicts. The search ▪ resource element,
quality benefits from this upfront information, enabling a conflict re- ▪ product element and
solution within the modelling process [49] (Table 3). ▪ properties.
To allow a search method to find different weighted interrelations
between the planning elements (elements, classes, properties), the re- Those classes build the basis for more detailed classes and for the
lations between a start node i and a target node j can be modelled as collaborative problem modelling system. As soon the interdependencies
sequence, assignment, dependency or mixed relation with following between the relations, the types of the connected nodes as well the level
directions: of the nodes are modelled, the interface for the solution search must be
defined. For the assembly line planning this is the highest level of
3
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
Table 3
Roles and documents required for assembly line design.
Planning Roles Document for assembly line design Individual objectives
process elements and the lowest level on structure elements, meanwhile further conglomeration of the planners can be required, considering
the elements of other classes are considered as additional components only the defined dependency flaws.
to the solution optimization. A backup approach would be the encapsulation of the inconsistent
The dependencies between the relations, the classes and the ele- dependencies within a placeholder that contains a summed-up proces-
ments are modelled by different. In contrast to the relations that con- sing time along with all related tasks and operators. The conflict can be
nect the elements and classes, the parameter relations of the properties solved manually after possible solutions have been found. The ad-
are considered as following: vantage is that fast processing is possible without continuous feedback
required for any imperfection of the data sets.
Table 4
Example for a multi-agent modelled dependency map.
Start node type Start node id Target node type Target node id Edge weight id Edge direction id Edge distance Agent id
4
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
yes
Problem modelling
no
yes
Problem Modeler
no
Analyse node content Analyse position in Evaluate probability
System
no
Solution
search
Final Solution
5
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
no
no yes
yes
Conflicts?
Temporarily eliminate most
no probable edges
no no
generated iteratively based on the outgoing list for so long no other number of individual within a population and the number of executed
connection to an outnode already used has an innode. Then the next iterations. There are a Binary Crossover operator as well as a Swap
node on the first list is used, until all nodes are assigned to a structural Mutation and an Inversion Mutation Operator. The best results are
element, i.e. a workplace or a workstation. As soon the available time is reached by using a combination of Swap and Inversion operator, which
smaller than the processing time required for the task to be assigned to is used within this example, too. The fitness function can be defined
the structural element, a new structure is generated. according to the underlying restrictions of the assembly line planning
The optimization is carried out using a Genetic Algorithm (GA) with problem.
swap mutation. Genetic Algorithms are stochastic optimization proce- In this example, the working height as well as the assembly line
dures from the field of metaheuristics. In this approach, a 1:1 assign- section of each operation, which has to be carried out at the assembly
ment between the set of genotypes (search space) and phenotype (so- line, is integrated in the fitness function. By adding an additional
lution space) is chosen. This approach uses a permutation chromosome iteration where tasks without indegree are checked to fill gaps in
coding. working stations before so that any task can be assigned to work-
The selected individuals are varied by the genetic operators stations, if available time of a workstation is bigger than the processing
Crossover and Mutation. While Crossover mixed the genetic material of time of the considered task.
parents randomly, Mutation varies genetic sequences of one individual. A reduced structure of the solution method is shown in Fig. 6. Be-
The new individuals called children are evaluated according to their sides the classes of a typical genetic algorithms, the initiation strategy,
fitness. After that, the iteration can be completed using a defined re- the problem structure based on the interdependencies of the element
placement scheme. This scheme is responsible to replace determined classes as well as the decode strategy are passed to the genetic algo-
individuals in the population by new generated one. rithm.
This process will be executed over a high number of iterations, For the evaluation the configuration parameters working height,
which causes in a perpetual development of the population. Therefore, assembly line section and sum up decode are tested. The evaluation in
the individuals get a better fitness because of the selection pressure. Fig. 7 shows that an increasing number of constraints results in an in-
Conceptual transferred to the optimization problem better solutions can crease of computation time (left y-axis), decrease in the fitness of the
be reached from iteration to iteration. Within this approach, several best found solution and an increase in the average acceptance rate of
setting and operators can be chosen. At first, it is possible to choose the the best found solution by the agents (right y-axis). The constraints
6
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
F1
K2 4
4 4 4 H C
6 F2
4 6 4 E
K1 1 4 1
S
4
3
K3 K5 V T
4
4 4
K4
F1
K2 4 4
4 4 H C
6 F2
K1 4 6 4 E
1 4 1
S 4
K3/K4/K5
3
K3 K5 V T
4
4 4
K4
considered are based on the assignment of process elements to the amount of time. In conclusion, it shows that an automated dependency
workstation, so the tasks are only assigned to a workstation with pro- conflict resolution, which is feasible, will result in a worse average
cess elements with exactly the same property items. Further, it is ob- station load. Instead, a rapid feedback can be given to the planning
vious that the sum up heuristic will find further improvements of as- personnel allowing a face-to-face resolution. This small example reveals
signing non-related process elements. a conflict resolution approach in precedence graphs used in collabora-
As the raw numbers will not justify the extra work for modelling tive assembly line planning.
additional edges the acceptance rate of a solution will be. As more
edges are modelled the higher the possibility of an acceptance of the
5. Summary
proposed solution by the algorithm.
However, a feasible assembly line can be found within a short
This paper, which is based on a collaborative problem solving
7
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
16 1.2
14
1
12
0.8
10
8 0.6
6
0.4
4
0.2
2
0 0
0,0,0 0,0,1 0,1,0 0,1,1 1,0,0 1,0,1 1,1,0 1,1,1
concept, shows that the distributed precedence modelling using parallel References
and precedence dependencies opens a new interdisciplinary scope of
identifying conflicts and resolving conflicts within the problem mod- [1] N. Boysen, A. Fliedner, A. Scholl, A classification of assembly line balancing pro-
elling for assembly line planning. This approach adds graph algorithms, blems, Eur. J. Oper. Res. 183 (2) (2007) 674–693.
[2] A. Scholl, C. Becker, State-of-the-art exact and heuristic solution procedures for
data preparation for algorithmic solution searches and distributed simple assembly line balancing, Eur. J. Oper. Res. 168 (3) (2006) 666–693.
modelling in scope of assembly line planning. As a starting point, this [3] C. Becker, A. Scholl, A survey on problems and methods in generalized assembly
paper presents three of many possible approaches for resolving conflicts line balancing, Eur. J. Oper. Res. (168) (2006) 694–715.
[4] A. Scholl, Balancing and Sequencing of Assembly Lines, Physica Verlag, 1999.
within a precedence graph, created by distributed agents using the [5] S. Bock, Modelle und verteilte Algorithmen zur Planung getakteter Fließlinien,
above-mentioned collaborative problem-solving concept. Further re- Gabler, Wiesbaden, 2000.
search of the graph resolution will be carried out over multiple graph [6] J. Rubinovitz, G. Levitin, Genetic algorithm for assembly line balancing, Int. J. Prod.
Econ. 41 (1-3) (1995) 343–354.
levels as well the automatic graph resolution of parallel and precedence [7] Y. Tsujimura, M. Gen, E. Kubota, Solving fuzzy assembly-line balancing problem
dependencies using artificial learning and data analysis. The solution with genetic algorithms, Comput. Ind. Eng. 29 (1-4) (1995) 543–547.
search is carried out using a combination of heuristic, genetic algo- [8] J. Plans, A. Corominas, Modelling and Solving the SALB-E Problem, IEEE,
Piscataway, NJ, 1999, pp. 356–360.
rithms and a problem-based decoding strategy to allow a solution
[9] R. Pastor, A. Corominas, Assembly Line Balancing with Work Zone and Station Load
search with multiple constraints. As the multitude of constraints and Time Windows Constraints, IEEE, Piscataway, NJ, 1999, pp. 448–453.
agents create conflicts on different levels a feasible solution can be [10] S.G. Ponnambalam, P. Aravindan, G. Mogileeswar Naidu, A multi-objective genetic
found. Regarding the problem modelling considerations have to be algorithm for solving assembly line balancing problem, Int. J. Adv. Manuf. Technol.
16 (5) (2000) 341–352.
made for further research to find an optimal measure for involving the [11] J. Bautista, R. Suarez, M. Mateo, R. Companys, Local Search Heuristics for the
information from multiple parties. Assembly Line Balancing Problem with Incompatibilities Between Tasks, Robotics
and Automation Society; IEEE Service Center, Piscataway, NJ, 2000, pp.
2404–2409.
[12] B. Rekiek, P. Lit, F. Pellichero, T. L'Eglise, P. Fouda, E. Falkenauer, A. Delchambre,
Acknowledgements A multiple objective grouping genetic algorithm for assembly line design, J. Intell.
Manuf. 12 (5/6) (2001) 467–485.
[13] J. Bautista, J. Pereira, Ant algorithms for a time and space constrained assembly
We gratefully acknowledge the support by the Federal Ministry of line balancing problem, Eur. J. Oper. Res. 177 (3) (2005) 2016–2032.
Education and Research, Germany for sponsoring the joint project [14] B. Rekiek, A. Delchambre, Assembly Line Design, Springer-Verlag London Limited,
MontAss (FID: 01IS15029B). The project aims for designing an assis- London, 2006, p. 1 online resource.
[15] P.M. Vilarinho, A.S. Simaria, ANTBAL, Int. J. Prod. Res. 44 (2) (2006) 291–303.
tance system for collaborative goal, precedence and restriction model-
[16] Z.-Q. Zhang, W. Cheng, L. Tang, Z. Bin, Ant Algorithm with Summation Rules for
ling with intelligent and autonomous agents. These input parameters Assembly Line Balancing Problem, first ed., Harbin Institute of Technology Press,
are the basis for automatic tasks assignment using a genetic algorithm Harbin, P.R. China, 2007, pp. 369–374.
[17] P. Su, Y. Lu, Combining Genetic Algorithm and Simulation for the Mixed-Model
for assembly line balancing.
Assembly Line Balancing Problem, IEEE, Piscataway, NJ, 2007, pp. 314–318.
In addition, we thank Prof. Dr. J. Käschel for his guidance, support [18] W. Zhang, M. Gen and L. Lin, A Multiobjective Genetic Algorithm for Assembly Line
and the possibility to work on this research project. Balancing Problem with Worker Allocation, Piscataway, NJ, IEEE, 2008, p.
3026–3033.
[19] Z.X. Guo, W.K. Wong, S. Leung, J.T. Fan, S.F. Chan, A genetic-algorithm-based
optimization model for solving the flexible assembly line balancing problem with
work sharing and workstation revisiting, IEEE Trans. Syst. Man Cybern. Part C 38
(2) (2008) 218–228.
[20] Z. Cao, S. Ma, Balancing and Sequencing Optimization of the Mixed Model
Assembly Lines, IEEE, Piscataway, NJ, 2008, pp. 732–736.
[21] J. Bautista, J. Pereira, A dynamic programming based heuristic for the assembly
line balancing problem, Eur. J. Oper. Res. 194 (3) (2009) 787–794.
[22] H.H. Kao and D.H. Yeh, "An Algorithm for Type-I Assembly Line Balancing Problem
8
D. Neumann, J. Keidel Robotics and Computer Integrated Manufacturing xxx (xxxx) xxx–xxx
with Optimal Workload Balance," Piscataway, N.J., IEEE, 2009, p. 822–826. [37] J.M. Nilakantan, S.G. Ponnambalam, An Efficient PSO For Type II Robotic Assembly
[23] C.L. Khaw, S.G. Ponnambalam, Multi-rule Multi-Objective Ant Colony Optimization Line Balancing Problem, IEEE, Piscataway, NJ, 2012, pp. 600–605.
for Straight and U-type Assembly Line Balancing Problem, IEEE, Piscataway, N.J., [38] L.M. Liao, C.J. Huang, A Multi-Agent Based Rescheduling Framework for Mixed-
2009, pp. 177–182. Model Assembly Line Balancing, IEEE, Piscataway, NJ, 2012, pp. 474–478.
[24] M. Chica, O. Cordon, S. Damas, J. Bautista, Adding Diversity to a Multiobjective Ant [39] A.A.J. Adham, R.B.M. Tahar, Enhancing Efficiency of Automobile Assembly Line
Colony algorithm For Time and Space Assembly Line Balancing (2009) 364–369. Using the Fuzzy Logical and Multi-Objective Genetic Algorithm, IEEE, Piscataway,
[25] V.L.V. Jonnalagedda, B.M. Dabade, Heuristic Procedure for Mixed Model Assembly NJ, 2012, pp. 1–7.
Line Balancing Problem, IEEE, Piscataway, N.J., 2010, pp. 552–556. [40] C. Jaturanonda, S. Nanthavanij, S.K. Das, Heuristic procedure for the assembly line
[26] O. Brudaru, C. Rotaru, Dynamic Segregative Genetic Algorithm for Assembly Lines balancing problem with postural load smoothness, Int. J. Occup. Saf. Ergon. 19 (4)
Balancing, IEEE, 2010, pp. 229–236. (2013) 531–541.
[27] Q. Tang, Z. Xiao, Y. Liang, M. Deng, Z. Xi, Novel Approach for Balancing Manual [41] H. Yu, W. Shi, A Genetic Algorithm for Multi-Model Assembly Line Balancing
Automobile Assembly Based on Genetic Algorithm, IEEE, Piscataway, N.J., 2010, Problem, IEEE, Piscataway, NJ, 2013, pp. 369–371.
pp. 2028–2032. [42] C. Liu, H. Wen, Hybrid Particle Swarm Algorithm For Assembly Line Balancing
[28] M. Chica, O. Cordon, S. Damas, J. Bautista, A Multiobjective Memetic Ant Colony Problem in Complicated Products, IEEE Press, Piscataway, N.J., 2013, pp. 902–905.
Optimization Algorithm for the 1/3 Variant of the Time and Space Assembly Line [43] U. Saif, Z. Guan, W. Liu, C. Zhang, B. Wang, Pareto based artificial bee colony
Balancing Problem, IEEE, Piscataway, NJ, 2011, pp. 1–7. algorithm for multi objective single model assembly line balancing with uncertain
[29] M. Chica, O. Cordon, S. Damas, Tackling the 1/3 Variant of the Time and Space task times, Comput. Ind. Eng. 76 (2014) 1–15.
Assembly Line Balancing Problem by Means of a Multiobjective Genetic Algorithm, [44] R. Davidrajuh, Solving Assembly Line Balancing Problems with Emphasis on Cost
IEEE, Piscataway, N.J., 2011, pp. 1367–1374. Calculations: A Petrinets Based Approach, IEEE, Piscataway, NJ, 2014, pp. 99–104.
[30] X. Qian, Q. Fan, Solving Multi-manned Assembly Line Balancing Problem by a [45] H. Triki, A. Mellouli, W. Hachicha, F. Masmoudi, Assembly Line Balancing in an
Heuristic-mixed Genetic Algorithm, IEEE, IEEE Computer Society, Piscataway, Los Automotive Cables Manufacturer Using a Genetic Algorithm Approach, IEEE,
Alamitos, California, 2011, pp. 320–323. Piscataway, NJ, 2014, pp. 297–302.
[31] M.N.I. Sulaiman, Y.-H. Choo, K.E. Chong, Ant Colony Optimization with Look [46] M. Chica, J. Bautista, S. Damas, O. Cordon, Adaptive IDEA for Robust
Forward Ant in Solving Assembly Line Balancing Problem, IEEE, Piscataway, NJ, Multiobjective Optimization, Application to the r-TSALBP-m/A, Conference
2011, pp. 115–121. Publishing Services, IEEE Computer Society, Los Alamitos, California; Washington;
[32] X. Hu, Heuristic Algorithm for Two-Sided Assembly Line Balancing Problem with Tokyo, 2015, pp. 1013–1020.
Multi-Objectives, IEEE, Piscataway, NJ, 2011, pp. 1407–1410. [47] C.G.S. Sikora, T.C. Lopes, H.S. Lopes, L. Magatao, Genetic Algorithm for type-2
[33] C. Zhuo, Y. Xu, Y. Tan, A Research of the Stochastic U-type Assembly Line Balancing Assembly Line Balancing, IEEE, Piscataway, NJ; Piscataway, NJ, 2015, pp. 1–6.
Problem, IEEE, Piscataway, NJ, 2011, pp. 874–878. [48] P. Sivasankaran, P. Shahabudeen, Literature review of assembly line balancing
[34] M.E. Noushabadi, U. Bahalke, K. Dolatkhahi, A.M. Yolmeh, A Simulation problems, Int. J. Adv. Manuf. Technol. 73 (9-12) (2014) 1665–1694.
Optimization Approach to Un-Paced Assembly Line Balancing Problem-II with [49] D. Neumann, J. Keidel, Collaborative formalization and modelling of multi-objec-
Additional Reworking Issue, IEEE, Piscataway, NJ, 2011, pp. 1–6. tive optimization problems, in: C.F. Oduoza (Ed.), Proceedings of the 25th
[35] N.M. Razali, J. Geraghty, Biologically Inspired Genetic Algorithm to Minimize Idle International Conference on Flexible Automation and Intelligent Manufacturing:
Time of the Assembly Line Balancing, IEEE, Piscataway, NJ, 2011, pp. 105–110. Designing for Advanced, High Value Manufacturing and Intelligent Systems for the
[36] M. Chica, Ó. Cordón, S. Damas, J. Bautista, Multiobjective memetic algorithms for 21st Century: 23-26 June, 2015, Wolverhampton, UK, vol. 2, The Choir Press,
time and space assembly line balancing, Eng. Appl. Artif. Intell. 25 (2) (2012) Wolverhampton, UK, 2015, pp. 124–131.
254–273.