A comparative analysis of sequencing procedures for mixed-model assembly lines in a just-in-time production system

Abstract

In this paper several procedures for sequencing products on a mixed-model assemblyline in a just-in-time production system are statistically compared using simulation analysis. The procedures include two 'goal chasing' heuristics developed at Toyota that focus on constant component usage, an algorithm developed by Miltenberg designed to achieve uniform production rates for each model, a time spread method developed by the authors that smooths the work load at each assemblyline station, and the familiar batch sequencingprocedure frequently used in practice. These five sequencing procedures are evaluated according to four measures of assembly line inefficiency, work not completed, worker idleness, worker station time and a measure of variability in uniform component usage. The results provide assembly line managers with guidelines for selectingthe most effective sequencing procedure for achieving their operational objectives. IntroductionJust-in-time (JIT) production systems frequently include mixed-model assembly lines to produce a variety of different product models while maintaining quality without incurring excessive inventories. Monden (1983) in his book on JIT at Toyota specifically addresses the use of mixed-model assembly lines in a JIT environment. In his discussion of this topic, Monden points out that if products with relatively longer processing times are successively fed into the assembly line (i.e. usage is not constant), a delay in model completion will eventually occur, which may stop the (conveyor) line. In order to avoid this problem, the processing time at each station must be smoothed by sequencing models so that, in general, a model with relatively short processing time at a station follows soon after a model with relatively long processing time at that station. Also, the quantity of each part used per unit time must be as near constant as possible. It is crucial for the processes preceding the assembly line supplying components, that the demand for these components be uniform. Uniform demand allows the JTT 'pull' system to minimize work-in-process inventories. Thus there are two major considerations when sequencing models: avoiding unfinished work which may stop the conveyor and assuring uniform demand for the components feeding the assembly line.As the popularity of JIT has evolved during the 1980s a number of different sequencing methods have been developed and reported in the research literature. The purpose of this paper is to compare several of the more prominent procedures for sequencing products on a mixed-model assembly line in a JIT environment. Two of the sequencing procedures included in this evaluation, referred to as 'goal chasing method' and 'goal chasing method II', are heuristics developed at Toyota and

AEGIS—attribute experimentation guiding improvement searches

“…This connection was also noted by Sumichrast et al (1992) in his studies. Likewise, the second observation led to a further investigation of the spacing issue.…”

Genetic algorithms for sequencing problems in mixed model assembly lines

“…(1992) is the minimization the overall line length. Sumicharast, Russell, and Taylor (1992) proposed a 'Time spread method' based on the idea similar to the goal-chasing method, and made a comparative analysis with other approaches. Stiener and Yeomans (1993) developed a graph-theoretic approach to determine an optimal solution for the second goal by formulating a new, non-convex objective function.…”

“…Three researches done by Duplaga and Bragg (1998), Leu et al (1996), and Sumicharast et al (1992) addressed the comparison of mixed-model line sequencing heuristics for smoothing component parts usage. They compared the heuristics by considering the variation only for single level.…”

Sequencing mixed-model assembly lines: Survey, classification and model critique

“…Preliminary results are provided by Sumichrast et al (1992), who schedule resulting model sequences obtained by workload-oriented level scheduling in a simulated real-world production environment. Unfortunately, they do not compare these results to mixed-model and car-sequencing, so that further research seems recommendable.…”