Dynamic Rule-Based Genetic Algorithm For Large-Size Single-Stage Batch Scheduling

Yaohua He,Chi-Wai Hui

DOI: https://doi.org/10.1016/s1570-7946(06)80327-5

2006-01-01

Abstract:Single-stage multi-product scheduling problem (SMSP) with parallel units has been widely studied, very often solved by using mixed-integer linear programming (MILP). As the problem size increases linearly, the computational time of MILP increases exponentially, so that it is very difficult for MILP to obtain an acceptable solution to large-size problems within reasonable time. Hence, the preferred method in industry is the use of scheduling rules. However, due to the constraints in SMSP, the simple rule-based method may not guarantee the feasibility and quality of the solution. In this study, random search (RS) based on heuristic rules was first proposed. By exploring a set of random solutions, better feasible solutions were acquired. To improve the quality of random solutions. Genetic algorithm (GA) based on heuristic rules was then proposed. The heuristic rules are crucial to cut down the solution space; but there is no single universal rule, and the effectiveness of the rules depends on the scheduling objective and the prevailing shop or plant conditions. Hence, a dynamic rule selection method based on GA was finally proposed. To increase the search speed in solving the high-constrained problem, a penalty method was adopted. Through the comparison of the computational results of MILP and the new methods, the proposed methods have demonstrated their effectiveness and reliability to solve large-size scheduling problems.