Multiobjective process planning and scheduling using improved vector evaluated genetic algorithm with archive
Multiobjective process planning and scheduling using improved vector evaluated genetic algorithm with archive
复制标题
DOI：
10.1002/tee.21726
复制
发表时间：
2012-05
影响因子：
1
通讯作者：
Wenqiang Zhang;S. Fujimura
中科院分区：
文献类型：
--
作者：
Wenqiang Zhang;S. Fujimura
多目标工艺规划与调度（PPS）是制造系统中一个重要而又棘手的组合优化问题。许多研究人员已经使用多目标进化算法（moEAs）来解决这些问题;然而，这些方法在功效（质量，即，收敛和分布）和效率（速度）。非支配排序遗传算法II（NSGA-II）和SPEA 2作为经典的moEAs算法，可以获得很好的效果，但需要大量的CPU时间。矢量评价遗传算法（VEGA）也不能应用，因为它的效率低。本文提出了一种改进的带存档的VEGA（iVEGA‐A）来处理多目标PPS问题，考虑了最小化完工时间和机器工作负载变化。该方法巧妙地结合了VEGA机制对帕累托前沿边缘区域的偏好以及广义帕累托尺度无关适应度函数（gp-siff）具有向帕累托前沿中心区域收敛趋势的特征。这两种机制不仅保持了收敛速度，而且保证了更好的分布性能。此外，一些问题相关的交叉，变异和局部搜索方法被用来提高算法的性能。完整的数值比较表明，iVEGA-A在效率上明显优于VEGA，收敛性能也优于NSGA-II和SPEA 2，而分布性能与之相当，效率明显优于它们。© 2012日本电气工程师协会。由John Wiley & Sons公司出版
Multiobjective process planning and scheduling (PPS) is a most important practical but very intractable combinatorial optimization problem in manufacturing systems. Many researchers have used multiobjective evolutionary algorithms (moEAs) to solve such problems; however, these approaches could not achieve satisfactory results in both efficacy (quality, i.e., convergence and distribution) and efficiency (speed). As classical moEAs, nondominated sorting genetic algorithm II (NSGA‐II) and SPEA2 can get good efficacy but need much CPU time. Vector evaluated genetic algorithm (VEGA) also cannot be applied owing to its poor efficacy. This paper proposes an improved VEGA with archive (iVEGA‐A) to deal with multiobjective PPS problems, with consideration being given to the minimization of both makespan and machine workload variation. The proposed method tactfully combines the mechanism of VEGA with a preference for the edge region of the Pareto front and the characteristics of generalized Pareto‐based scale‐independent fitness function (gp‐siff) with the tendency to converge toward the central area of the Pareto front. These two mechanisms not only preserve the convergence rate but also guarantee better distribution performance. Moreover, some problem‐dependent crossover, mutation, and local search methods are used to improve the performance of the algorithm. Complete numerical comparisons show that the iVEGA‐A is obviously better than VEGA in efficacy, and the convergence performance is also better than NSGA‐II and SPEA2, while the distribution performance is comparable to and the efficiency is obviously better than theirs. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.