基于BP-NSGA的注塑参数多目标智能优化设计  

Multi-objective intelligent optimization of injection molding parameters based on BP-NSGA

作　　者：程锦[1] 谭建荣[1] 金莉莉[1] 

机构地区：[1]浙江大学流体传动及控制国家重点实验室,浙江杭州310027

出　　处：《计算机集成制造系统》2009年第10期1900-1906,共7页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(50705084);浙江省自然科学基金资助项目(Y1090512);浙江省教育厅科研资助项目(Y200805452)~~

摘　　要：为获得成型性能最优的注塑参数设计方案,提出了基于BP神经网络和非支配排序遗传算法的注塑参数多目标优化方法。将注塑模结构尺寸参数和注塑工艺参数作为待优化的设计变量,建立了以高质量、低成本、高效率为优化目标的注塑参数优化设计模型。基于非支配排序遗传算法获取给定参数范围内的所有Pareto最优解,并通过建立多输入和多输出的BP神经网络来快速获得非支配排序遗传算法优化进程中所有个体的适应度值。开发了基于BP神经网络与非支配排序遗传算法集成的注塑参数智能优化设计系统,并通过鼠标注塑参数设计实例,验证了其适用性和有效性。To get the optimal parameter schemes with their molding ability evaluated by comprehensive criteria, a multi-objective optimization approach for injection molding parameters based on Back Propagation （BP） neural network and Non-dominated Sorted Genetic Algorithm （NSGA） was proposed. The mathematical model with high product quality, low cost, and high efficiency as objectives was established for optimizing injection molding parameters related to both mold structures and molding processes. All the Pareto-optimal solutions within the specified parameter region were located based on NSGA and a multi-input multi-output BP network was developed for fast computing the fitness of every individual during the evolution of NSGA. The system for intelligent optimization of injection molding parameters based on the integration of BP network and NSGA was developed, and the feasibility and validity were demonstrated through the optimization of injection molding parameters for manufacturing mice.

关 键 词：注塑参数 多目标 智能优化 PARETO最优集 BP神经网络 遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]