模糊人工神经网络在冷挤压工艺设计中的应用研究  

Research on applications of fuzzy ANN in cold extrusion process design

作　　者：高军[1] 季廷炜[1] 赵国群[1] 王广春[1] 阮立群[2] 

机构地区：[1]山东大学材料科学与工程学院,济南250061 [2]日本国立熊本大学知能生产系统工学科

出　　处：《塑性工程学报》2005年第3期22-26,共5页Journal of Plasticity Engineering

基　　金：山东省自然科学基金资助项目(Q2002F01)。

摘　　要：在分析冷挤压成形影响因素的基础上,将模糊理论和人工神经网络进行了有机集成,并用以解决判断冷挤压零件能否通过一次挤压成形的问题。首先将人工神经网络进行模糊化处理,然后在分析冷挤压零件能否一次成形的有关规则基础上,抽取若干个模糊变量,并利用隶属度函数将影响冷挤压成形的模糊变量模糊化,同时通过从领域专家或其它知识源获取的训练样本,对人工神经网络进行学习,获得能够进行一次成形判断的人工神经网络模型,最后选取典型冷挤压零件进行了运行测试,并且用商业化软件DEFORM对系统输出的判断结果进行数值模拟分析,模拟结果证明了系统的可行性。On the basis of analysis of cold extrusion processes, fuzzy logic was integrated with artificial neural networks (ANN), in order to solve the problem that whether or not the cold extrusion part can be formed in one blow. ANN was fuzzified at first, and fuzzy variables were extracted through the analysis of rules that whether or not the cold extrusion part can be formed in one blow. Then, fuzzy variables were fuzzified by different functions of membership degree. ANN was trained by data samples that came from experts or other knowledge resources. The system was tested through a typical cold extrusion part. In the end, FEM analysis validates the mapping results of fuzzy ANN.

关 键 词：冷挤压工艺设计 模糊逻辑 人工神经网络 隶属度函数 

分 类 号：TG376[金属学及工艺—金属压力加工]