Operation planning based on a genetic algorithm for multi-model assembly lines considering worker change in allocation

This paper focuses on multi-model assembly line production, where each model is produced in small lots, and operations differ obviously among models (e.g., number of operations, operation precedence constraints and so forth). At the same time, according to workers' different experiences and capabilities, their work skills differ largely among operations, and the processing time of a given operation is diverse among workers. Taking total processing time (makespan) as objective function, the line operation planning problem of such an assembly line is discussed, which includes three sub-problems: (i) assigning operations of each lot to workstations, (ii) allocating workers to workstations for each lot, and (iii) determining the lot production sequence for the line. In order to reduce the cycle time (CT) of each lot considering the obvious difference in worker skills, the worker allocation schemes of each lot may be different. Therefore, the switch time (SWT) caused by different worker allocation between adjacent lots has to be dealt with as well. Since this is a complicated and large-scale problem, an algorithm based on genetic algorithm (GA) is developed in this paper. A heuristic crossover procedure is proposed, which focuses on not only the workload balance loss of each lot but also the switch loss between adjacent lots. Our approach is tested and confirmed to be effective for shortening makespan compared to the following two production approaches: SPPT-oriented production approach and Cr-oriented production approach.