Abstract
We address a proactive scheduling problem with stochastic machine breakdown, controllable processing time and deterioration effect considerations arising from steel production. The problem is to determine the pre-compression amount of each job's processing time and the job sequence for the rolling process so as to achieve a robust predictive schedule in response to machine breakdown. Both robustness and stability of the predictive schedule are considered, in correspondence with the mean and variance of rescheduling cost that consists of match-up time cost and additional resource cost. Since the scenario-based approach to robustness evaluation of a predictive schedule is cursed with high computational burden, a surrogate-assisted multi-objective evolutionary algorithm based on Elitist non-dominated sorting genetic algorithm is proposed to solve the proactive scheduling problem under consideration. Support vector regression model is introduced to approximate the robustness of the each alternative schedule which surrogates the time-consuming simulation-based fitness evaluation process and saves more time for solution space search. In addition, a probabilistic sequencing strategy which takes advantage of each job's ability to absorb disruption at low cost is introduced to guide the evolutionary search. Computational experiments of numerical and practical data indicate that the proposed proactive scheduling approach performs well in response to stochastic machine breakdown. The support vector regression model and the probabilistic sequencing strategy improve the performance of the proposed algorithm with respect to the convergence and diversity of the obtained Pareto front.