改进型轴类零件计算机辅助工艺设计专家系统研究

发布时间：2018-02-10 21:25

本文关键词： CAPP 工艺推理轴类零件特征矩阵专家系统神经网络　出处：《燕山大学》2016年硕士论文　论文类型：学位论文

【摘要】：计算机辅助工艺设计(Computer Aided Process Planning,CAPP)是实现工艺设计智能化和自动化的途径,随着制造业的飞速发展,工艺设计对系统智能化、信息化的要求越来越高。而传统CAPP专家系统存在零件信息提取不完整、知识获取困难、匹配冲突及推理准确性差等问题,已难以满足现代制造业发展的需要。所以本文针对上述问题,利用特征建模、人工神经网络等技术,对传统专家系统进行了改进,主要对零件信息提取,知识库设计和工艺推理三个模块进行了探索和研究。首先,分析传统专家系统和神经网络的优缺点,对两者进行对比说明,构建基于神经网络的CAPP专家系统的总体框架,并对主要研究模块的实现方法和相关技术进行研究。其次,对轴类零件信息提取模块进行研究,采用特征技术和编码技术,参照标准编码体系编制适合本系统的特征编码规则,以构建特征矩阵的方式来解决零件信息提取不完整的问题。然后,对轴类零件工艺知识库模块进行研究,采用IF…THEN…规则表示和神经网络权值表示两种方法表达工艺知识,并存储于SQL Sever数据库中。再次,对工艺推理模块进行研究,采用神经网络和专家系统相结合,以构建混合推理机制的方式来解决匹配冲突和推理效率的问题。详细研究了专家系统推理策略,神经网络结构设计、学习算法、样本选择及训练等问题。最后,以Visual Studio2008作为系统开发工具,利用Matlab实现神经网络推理部分的功能,以VB作为开发语言,编制程序,设计工作界面,结合工艺推理实例,对上述模块进行验证,证明了上述研究方法的有效性。
[Abstract]:Computer Aided Process planning is a way to realize the intelligent and automatic process planning. With the rapid development of manufacturing industry, process planning is intelligent to the system. However, the traditional CAPP expert system has some problems, such as incomplete part information extraction, difficulty in obtaining knowledge, matching conflict and poor reasoning accuracy, etc. It is difficult to meet the needs of the development of modern manufacturing industry. Therefore, this paper improves the traditional expert system by using the techniques of feature modeling and artificial neural network, mainly to extract the information of parts. Three modules of knowledge base design and process reasoning are explored and studied. Firstly, the advantages and disadvantages of traditional expert system and neural network are analyzed, and the comparison between them is made to construct the overall framework of CAPP expert system based on neural network. And the main research modules and related technologies are studied. Secondly, the information extraction module of shaft parts is studied, using feature technology and coding technology, the characteristic coding rules suitable for the system are compiled according to the standard coding system. The problem of incomplete part information extraction is solved by constructing feature matrix. Then, the knowledge base module of shaft parts is studied. If... THEN... Rule representation and neural network weight representation are used to express process knowledge. It is stored in SQL Sever database. Thirdly, the process reasoning module is studied, and the neural network and expert system are combined. The problems of matching conflict and reasoning efficiency are solved by constructing hybrid reasoning mechanism. The reasoning strategy of expert system, neural network structure design, learning algorithm, sample selection and training are studied in detail. Using Visual Studio2008 as system development tool, using Matlab to realize the function of neural network reasoning, using VB as development language, programming, designing working interface, and combining process reasoning instance, the above modules are verified. The effectiveness of the above method is proved.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH162;TP391.72

【相似文献】