Using simulation to evaluate the assembly line balancing solutions proposed by heuristic algorithms under stochastic task times

In this work, a benchmark assembly line balancing problem is solved by using heuristic procedures that are commonly used in industry and broadly represent the two main families of commercial algorithms, namely COMSOAL procedures and genetic algorithms. Then, considering stochastic task times, the proposed solutions are implemented via a simulation testbed in order to examine the effect of the variability of the task times on the number of finished products for a given simulation time (sensitivity analysis). The results show that existing line balancing indexes do not characterize the quality of the proposed solutions very effectively when stochastic variations in task times are taken into consideration. At this level, simulation appears to be a reliable tool for decision making in assembly line design and task assignment to workstations, as it represents the actual characteristics of an assembly line more effectively.