Embedding optimization with deterministic discrete event simulation for assignment of cross-trained operators: an assembly line case study

In manufacturing systems where productivity is constrained by operatorsâ availability, cross-training strategies can be used to enable dynamic assignment of operators to workstations. However, finding an assignment approach that efficiently works under various system conditions is not trivial as, among other factors, the level of cross-training, the production duration and the initial conditions of the system can influence the assignment approach's performance. To overcome this issue, in the case study presented in this paper, operator assignments have been modeled using a simulation-based optimization approach, with an â outerâ optimizer that selects assignment-related parameters to simulate based on the system conditions, and an â innerâ optimizer integrated with a simulation model that generates optimal assignments. For the case described here, which is modeled using a deterministic simulation model, the â outerâ optimizer is an Ant Colony Optimizer (ACO) and the â innerâ optimizer is a Binary Integer Programming (BIP) model. The ACO will select weights for the assignment objectives of the BIP multi-objective function so that throughput is maximized. The BIP is called by the system simulation model at fixed intervals or when a system status changes to assign operators to workstations. Results show that the simulation-based optimization approach generates higher throughput performance than static WIP-base assignment, especially when longer production duration are considered. The effects of cross-training and production duration on production throughput are also investigated. The simulation-optimization approach used can be abstracted to a framework where the â outerâ and â innerâ optimizers may be applied to different domains than the case study addressed here and can also be applied to stochastic simulation models.