基于改进的离散PSO算法的FJSP的研究

Study on Flexible Job-shop Scheduling Problem Based on Improved Discrete Particle Swarm Optimization Algorithm

柔性作业车间调度问题(Flexible Job-shop Scheduling Problem,FJSP)是经典作业车间调度问题的一个扩展,前者更接近于实际生产。以最小化最大完工时间为目标,提出了一种改进的离散粒子群优化算法。传统粒子群优化算法一般适用于优化连续模型问题,FJSP作为复杂度比较高的组合优化问题,是一种典型的离散模型。提出的算法采用机器负荷平衡机制初始化粒子种群,在粒子的更新过程中引入了3个操作算子来更新粒子的工序排序部分和机器分配部分,这3个算子分别为基于工序排序或机器分配的变异、与个体最优位置之间进行工序先后顺序保留的交叉(POX)操作、与全局最优位置进行随机点保存的交叉(RPX)操作。先后执行以上3个算子以完成粒子的一次更新。这种操作能够使种群较快地收敛于最优解。对标准测试案例进行实验的结果表明,所提算法对解决FJSP具有有效性,并且能够快速地搜索到近似最优解;与其他同类算法相比,所提算法在求解效果和收敛速度上均具有优越性。

Flexible job-shop scheduling problem is an extension of the classical job-shop scheduling problem.The former is much closer to the practical production.Aiming at minimizing the maximum completion time,this paper proposed an improved discrete particle swarm optimization algorithm.The traditional particle swarm optimization algorithm is applicable to optimize the continuous models.As a combinatorial optimization problem with high complexity,FJSP is a typically discrete model.The proposed algorithm utilizes the load balancing mechanism for the machines to initiate the population,and introduces three operators into the procedure of updating the individuals’status in the population.The three operators are respectively described as follows:the mutation based on the operation sequencing or the machine assignment,the precedence preserving order based crossover between current particle and the individual optimal,and rand-point preservation crossover between current particle and the global optimal.A particle is completely updated by using three operators sucessively.This method makes the population converge to the optimal solution very fast.The experimental results of benchmark instances show that the proposed algorithm can practically solve the flexible job-shop scheduling problem and search the near optimal solutions very fast.The proposed algorithm outperforms the other similar algorithms with respect to searching efficiency and convergence speed.