An Empirical Comparison of Objective Functions for the Mixed-Model U-Shaped Assembly Line

Document Type



Request a copy of the paper from the author: John K. Visich


Several objective functions have been proposed in the literature to smooth (equalize) work load variations in mixed-model serial assembly lines. In this paper we modified and tested three mixed-model serial line objective functions for the mixed-model, U-shaped assembly line. These objective functions are the absolute deviation from cycle time (ADC), the maximum deviation from cycle time (MDC) and the sum of the cycle time violations (SCV). We compared these objective functions using the well-known smoothness index. We employed a four step smoothing process from the literature and swapped tasks between workstations to minimize our objective functions. Our task swapping procedure was guided by the great deluge algorithm heuristic. We tested three problem sizes of 19, 61 and 111 tasks, and a variety of cycle times and model sequences for a total of 68 problems. We analyzed our results using the non-parametric Friedman?s test in conjunction with the multiple-comparison for use with Friedman test. Our results indicated that across all problems the ADC and the MDC performed significantly better than the MDC. But, when we controlled for high and low cycle times, the MDC was significantly better than both ADC and SCV at low cycle times, while the ADC and SCV were significantly better than the MDC at high cycle times. From the results of this experiment we can conclude that the smoothness index is influenced by both the objective function and the cycle time in the mixed-model, U-shaped assembly line balancing problem. (Assembly Line Balancing, Mixed-Model Production, U-Lines, Layout)

This document is currently not available here.