Application of Genetic Algorithms and Visual Simulation in a Real-Case Production Optimization

This paper presents a real case customized flexible furniture production optimization. Such a make-to-order production must be flexible to meet the customer’s needs, which are changing frequently. Hence, a frequent review of the production process is needed to ensure near-optimal production schedule to meet the minimal makespan constraint. In such a case we are confronted with a tradeoff between makespan and optimization runtime to ensure efficient realtime scheduling. The genetic algorithm production scheduling optimization is presented to solve a job shop scheduling problem with recirculation. Several initial population generators, selection methods, and crossover and mutation methods are discussed and tested. The visual model of the furniture production process was developed during the research. Such a model is closer to end-user perception and is used to clarify the results of numerical optimization. It also enables the implementation of end-user’s expert knowledge into the optimizer to reduce the GA search space with a goal of reducing the runtime to a minimum. Key-Words: Production scheduling, Job-shop, Recirculation, Genetic algorithms, Simulation, Optimization