Stability radii of optimal assembly line balances with a fixed workstation set

https://doi.org/10.1016/j.ijpe.2016.07.016Get rights and content

Abstract

For an assembly line, it is required to minimize the line's cycle time for processing a partially ordered set of the assembly operations on a linearly ordered set of the workstations. The operation set is partitioned into two subsets, manual and automated. The durations of the manual operations are variable and those of the automated operations are fixed. We conduct a stability analysis for this problem. First, we derive a sufficient and necessary condition for the optimal line balance to have an infinitely large stability radius. Second, we derive formulas and an algorithm for calculating the stability radii for the optimal line balances. Third, we report computational results for the stability analysis of the benchmark instances. Finally, we outline managerial implications of the stability results for choosing most stable line balances, which save their optimality in spite of the variations of the operation durations, and for identifying the right time for the re-balancing of the assembly line.

Introduction

The assembly line consists of m workstations, which are linked by a conveyor belt (or another equipment) moving an in-process product from one workstation to the next at a constant pace. The set V of n assembly operations is fixed. Each workstation needs to perform a specific subset of the operations from the set V within the line's cycle-time. All the m workstations start simultaneously to process their own operations. A partial order on the operation set V arises due to technological and economical considerations, which are represented by the precedence digraph G=(V,A) with the set A of arcs. A Simple Assembly Line Balancing Problem is to find an optimal assembly line balance, i.e. an assignment of the operations V to the m workstations such that the cycle-time is minimal. The abbreviation SALBP-2 for denoting this problem has been introduced by Baybars (1986). The problem SALBP-2 is NP-hard (Gutjahr and Nemhauser, 1964, Wee and Manoj, 1982) since the bin-packing problem is NP-hard and is a special case of the problem SALBP-2, where in the bin-packing problem, the digraph G=(V,A) has no arcs, A=.

Throughout this paper, it is assumed that the set V consists of two specific subsets of the assembly operations. The non-empty subset V˜V includes all the manual operations and the subset VV˜ includes all the automated operations. The initial vector t=(t1,t2,,tn) of the operation durations is known before solving the problem SALBP-2. However, for the subset V˜V of the manual operations jV˜, each duration tj may vary due to different factors such as the operator skill, motivation, learning effect, etc. In contrast to the manual operations, the duration ti of each automated operation iVV˜ is fixed. We assume that V˜={1,2,,n˜} and VV˜={n˜+1,n˜+2,,n}, 1n˜n. The vectors of the operation durations are denoted as follows: t˜=(t1,t2,,tn˜), t¯=(tn˜+1,tn˜+2,,tn), t=(t˜,t¯)=(t1,t2,,tn). Let a subset Vkbr of the set V be assigned to the workstation Sk, where k{1,2,,m}. The assignment br:V=V1brV2brVmbr of the operations V to the ordered workstations (S1,S2,,Sm), VkbrVlbu=, 1k<lm, is called a line balance, if the following two conditions hold.

Condition 1

The assignment br does not violate the partial order given on the set V by the precedence digraph G=(V,A), i.e. each arc (i,j)A implies that operation iV is assigned to workstation Sk and operation jV is assigned to workstation Sl in a way such that 1klm.

Condition 2

The assignment br uses all the m workstations, i.e. the subset Vkbr is not empty for each workstation Sk, k{1,2,,m}.

Let B(G) denote the set of all assignments br satisfying Condition 1. The subset B(G,m)={b0,b1,,bh} of the set B(G) consists of all line balances. The cycle-time c(br,t) for the line balance br with the vector t=(t˜,t¯) of the operation durations is determined as c(br,t)=maxk=1miVkbrti, where the sum iVkbrti:=t(Vkbr) is a workstation time. The line balance b0 is optimal with the operation durations t=(t˜,t¯) if it achieves a minimal cycle-time c as follows:

Condition 3

c=c(b0,t)=min{c(br,t):brB(G,m)}.

Note that Condition 2 allows us to restrict a set of the line balances since the set B(G,m) contains the optimal line balance without fail. Let B(G,m,t) denote a set of all the optimal line balances, B(G,m,t)B(G,m), with the vector t=(t˜,t¯) of the operation durations. If operation i belongs to the set VV˜, its duration ti is fixed. Without loss of generality, we assume that ti>0 for each automated operation iVV˜ since the automated operation with the fixed zero duration has no influence on a solution to the problem SALBP-2. The initial duration ti is a strictly positive real number ti>0 for each operation iV. A value of the duration tj>0 of the manual operation jV˜V can vary during the assembly line lifespan. The varied duration tj may be even equal to zero, which means that the manual operation j from the setV˜kbr:=VkbrV˜is processed by an additional operator in parallel with the processing of other operations assigned to workstation Sk. Due to the additional operator, the processing of the manual operation j does not increase the workstation time, i.e.t(Vkbr)=iVkbrti=iVkbr{j}ti,where t indicates the modified vector t=(t˜,t¯)=(t1,t2,,tn˜,tn˜+1,tn˜+2,...,tn):=(t1,t2,,tn˜,tn˜+1,tn˜+2,...,tn), for which the workstation time iVkbrti is calculated. The second equality in (2) is valid because of holding equality tj=0. We summarize the above in the following remark.

Remark 1

The initial duration ti is a strictly positive real number for each operation iV. A value of the duration tj>0 of the manual operation jV˜ can vary during the assembly line lifespan. The varied duration tj may be equal to zero: tj0.

The aim of this paper is to investigate the stability of the optimal line balance with respect to variations t˜t˜ of the operation durations. The stability radius ρb0(t) of the optimal line balance b0 is interpreted as a maximum of simultaneous and independent variations t˜ of the durations t˜ of operations V˜ without violating the optimality of the line balance b0, i.e. b0B(G,m,t)B(G,m,t). A formal definition of the stability radius is given in Section 2.1 along with a sufficient and necessary condition for a zero stability radius. In Section 3, it is shown that the stability radius may be infinitely large, ρb0(t)=. Formulas for calculating the stability radius ρb0(t) for the line balance b0B(G,m,t) are given in Section 4.1. The calculation of the stability radius is illustrated in 2.2 Example 1, 3.3 Example 2, 4.3 Example 3. In Section 4.2, it is shown on how to restrict a subset of the set B(G,m){b0}, which must be compared with the line balance b0B(G,m,t) for calculating the stability radius ρb0(t). An algorithm for calculating the stability radius is presented in Section 5. Section 6 reports the computational results for the stability analysis of the benchmark instances from the old dataset and the recent one (Otto et al., 2013) tested in Morrison et al., 2014, Otto and Otto, 2014). In Section 7, the managerial implications are spelled out on how to use the stability results in the assembly industry. Concluding remarks and perspectives are discussed in Section 8.

Section snippets

Contributions of this work, previous results, and related literature

The assembly lines are widely used in a mass production for assembling components into final products. An effectively balanced assembly line allows a factory to increase its efficiency via reducing a production cost. Since the production conditions may change over time, the need of a re-balancing of the assembly line may arise from time to time in order to serve customer demands in the competitive market environment. The assembly re-balancing is tedious procedures requiring significant costs

An infinite stability radius of the optimal line balance

Theorem 1 gives sufficient and necessary condition for a zero stability radius. To present a criterion for the infinitely large stability radius, we need to introduce the simple assembly line balancing problem SALBP*-2, which has the same input data as the original problem SALBP-2, except that the initial vector t* of the operation durations: t*=(t˜*,t¯):=(0,0,,0,tn˜+1,tn˜+2,,tn). The line balance buB(G,m) is optimal for the problem SALBP*-2, if buB(G,m,t*).

The calculation of stability radii for the optimal line balances

Theorem 1, Theorem 2 give criteria for the extreme values of the stability radii: ρb0(t)=0 or ρb0(t)=. In 4.1 Formulas for calculating the stability radius, 4.2 The redundant line balances for calculating the stability radius, we show how to calculate the stability radius if 0<ρb0(t)<.

An algorithm for the calculation of stability radii for the optimal line balances

If the set B(G,m) is constructed, it is not difficult to find the set B(G,m,t), having chosen the line balances from the set B(G,m) with the minimal cycle-time. Using Theorem 1, one can choose all unstable optimal line balances B0(G,m,t)B(G,m,t). Using Theorem 2, one can choose all optimal line balances B(G,m,t) with infinite stability radii if they exist. Next, we present an algorithm for constructing sets B(G,m), B(G,m,t), B0(G,m,t), B(G,m,t), and calculating the stability radii for all

Computational results

Algorithm RAD was implemented in C++ and tested on the benchmark instances available on http://www.assembly-line-balancing.de The computational experiments were run on a laptop with the following characteristics: Intel(R), Pentium(R), CPU 2020 M @2.40 GHz 2.40 GHz, 4.00 GB Internal Memory. The main characteristics of the benchmark instances, which determine their complexity, are presented in Table 1 for the benchmark dataset given in Scholl (1999), and in Table 2 for the benchmark dataset given in 

The managerial implications of the stability results

We answer the following three questions: 1. Why and how are our stability results useful? 2. How does one obtain a stable optimal line balance if it exists? 3. How does one overcome a large size of the instance SALBP-2 for making the stability analysis either for all optimal line balances or for one of them? First of all, one can observe that an exact algorithm used for solving the NP-hard problem SALBP-2, traditionally, terminates once an optimal line balance is constructed, which may make

Conclusion

Since the durations of manual operations are variable, it is impossible to construct the optimal line balance, which will be the best for all variable operation durations. Nevertheless traditionally, the solution procedure for the problem SALBP-2 ends when the first optimal line balance b0B(G,m,t) is constructed. Due to the computational results presented in Table 3, Table 4, Table 5, Table 6, we can argue that for the problem SALBP-2 there is a high probability that the first constructed

Acknowledgment

The authors are grateful to the anonymous referees for their useful suggestions on the early draft of this paper. The first and second authors acknowledge the research grant from Ministry of Science and Technology, Taiwan.

References (33)

Cited by (25)

  • Assembly line balancing: What happened in the last fifteen years?

    2022, European Journal of Operational Research
  • Hybridizations in line balancing problems: A comprehensive review on new trends and formulations

    2022, International Journal of Production Economics
    Citation Excerpt :

    Zhao et al. developed an improved preference-based variable neighbourhood search algorithm with AR-dominance for assembly line balancing considering preventive maintenance scenarios, see (Zhao et al., 2022). Stability measures for obtained solutions were considered by Gurevsky et al. (2013a), Freiheit and Li (2017) and Lai et al. (2016). Intervals were also used for definition of task attributes, for example task processing time was defined as an interval in the following studies: Fang et al. (2020a); Gurevsky et al. (2013b); Hamta et al. (2013); Hazır and Dolgui (2013), (2015); Moreira et al. (2015); Pereira (2018b); Pereira and Álvarez-Miranda (2018); Xiao et al. (2021); Zhang et al. (2019a).

  • The impact of unequal processing time variability on reliable and unreliable merging line performance

    2021, International Journal of Production Economics
    Citation Excerpt :

    Furthermore, since the effect of machine unreliability on the performance of unbalanced patterns in merging lines has also been shown to be significant (Shaaban and Romero-Silva, 2020), this paper studies the performance of merging lines with reliable and unreliable machines, and station operation time variability imbalance, to develop intuition about the behavior of merging lines. This study provides valuable, complementary insights to the traditional assembly line studies which focused on balancing the station's cycle times (see, e.g., Lai et al., 2016; Özcan, 2019) and buffer assignment optimization (see, e.g., Demir et al., 2014; Weiss et al., 2019), and examines whether a simple balanced assignment of CV along the line is better than an unbalanced assignment. We attain this objective by simulating a merging assembly line with two parallel lines and assessing its performance in terms of throughput and average buffer level.

View all citing articles on Scopus
View full text