Solving fms loading problem with genetic algorithm: a strategy for solution generation and recombination

摘要

This paper attempts to develop a strategy for using genetic algorithm to solve the part selection and machine loding problem in a flexible manufacturing system (FMS). The strategy is guided by integer programming (IP) formulation of the problem. The solution chromosome consists of part type selection and operation assignment genes. For generation of solutions, constraints are segregated in to direct and indirect ones. Direct constraints (large in number) are satisified by assigning context-dependent values to decision variables. Indirect constraints require the penalty functions. A two-step crossover (part swap followed by route swap) is suggested which retains the solution feasibility across generations. The mutation is allowed only for operation assignment genes of selected part types. Implementation of methodology for small FMSs indicates that with proper penalty scheme and GA parameters, the proposed problem-specific GA can provide optimum solutions.