An investigation into heuristics for alternative worker selection in discrete event simulation

In this paper, we analyse four different heuristics for qualified worker selection for machines in discrete event simulation. Conventional simulators simply select a capable worker randomly or from the top-of-the-stack (TOS) of candidates that are qualified to operate a machine, without considering the impact of removing that worker from the current available qualification pool (qPool). To investigate the efficacy of this approach, we compare these random and TOS approaches with two other worker selection rules: least number of qualifications (LENQ), and a heuristic that selects a worker with the lowest impact factor on the qualification pool (LIMP). LIMP ranks workers based on their contribution to the qPool and the constrainedness of each of their qualifications. We apply LENQ to a simulation model of a real company, and compared with the Random heuristic we observe a 44% reduction in the qualification resource constraint metric (RCMq) and a 2% reduction in the total lateness in sales-order satisfaction. For the LIMP heuristic, the RCMq reduction is 77%. However, LIMP yields no significant improvement in sales-order lateness over the simpler LENQ approach. The LENQ and LIMP heuristics also have the benefit of more closely modelling what happens in reality, as they are based on intuition that would be used in practice, rather than using a random or simple TOS approach followed in conventional simulation.