Balancing large assembly lines by a new heuristic based on differential evolution method

Evolutionary algorithms such as genetic algorithms have been applied on a variety of complex combinatorial optimization problems (COPs) with high success. However, in relation to other classes of COPs, there is little reported experimental work concerning the application of these heuristics on large size assembly line balancing problems (ALBPs). Moreover, very few works in the literature report comparative results on public benchmark instances of ALBPs for which upper bounds on the optimal objective function value exist. This paper considers the simple ALBP of type 2 (SALBP-2), which consists of optimally partitioning the tasks’ operations in an assembly line among the workstations with objective the minimization of the cycle time of the line. SALBP-2 is known to be intractable, and therefore the right way to proceed is through the use of heuristic techniques. To that purpose, a novel approach based on the differential evolution method has been developed and tested over public available benchmarks ALBPs. These benchmarks include test instances for several precedence graphs (representing the assembly restrictions) with up to 297 tasks. Extended comparisons with other previously published evolutionary computation methods showed a superior performance for the proposed approach.