基于神经网络专家系统的工艺推理研究——以轴类零件为例

发布时间：2018-01-12 15:01

  本文关键词：基于神经网络专家系统的工艺推理研究——以轴类零件为例　出处：《现代制造工程》2017年09期 　论文类型：期刊论文

  更多相关文章： 工艺推理 专家系统 神经网络 特征矩阵

【摘要】：目前,传统专家系统工艺推理时存在零件信息提取不完整、知识获取困难和推理能力弱的问题,采用基于神经网络和规则的混合推理机制替代传统专家系统可以解决有效上述问题。首先,运用特征技术提取零件信息,将零件信息转换为特征矩阵,作为神经网络专家系统的输入;然后,根据特征矩阵搜索推理策略,基于轴类零件特征将神经网络分为精度、形状和热处理三类子网络,采用动量-自适应学习率BP算法训练网络;最后设计与实现了混合系统工艺推理过程。
[Abstract]:At present, there are some problems in the traditional expert system, such as incomplete information extraction, difficult knowledge acquisition and weak reasoning ability. Using hybrid reasoning mechanism based on neural network and rules to replace the traditional expert system can effectively solve the above problems. Firstly, the feature technology is used to extract part information and convert part information into feature matrix. As the input of neural network expert system; Then, according to the feature matrix search and reasoning strategy, the neural network is divided into three subnetworks: precision, shape and heat treatment based on the feature of shaft parts, and the momentum adaptive learning rate BP algorithm is used to train the network. Finally, the process of process reasoning in hybrid system is designed and implemented.
【作者单位】： 燕山大学经济管理学院;北京航空航天大学;迁安市九江线材有限责任公司;
【分类号】：TH16;TP18
【正文快照】： 3迁安市九江线材有限责任公司,唐山064400)0引言自20世纪60年代末CAPP系统诞生以来,一直受到国内外学者的重视,先后提出派生式、创成式、交互式等CAPP系统[1],而零件信息提取和工艺推理方法一直是研究的重点和难点。零件信息提取作为专家系统推理的前提和基础,至今仍然存在零，

本文编号：1414739