多品种小批量制造模式下工序质量控制研究

发布时间：2018-01-22 21:16

本文关键词： 多品种小批量工序质量控制相似性模式识别工序质量诊断　出处：《西安电子科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：质量是现代工业社会和各国经济建设中一个受到普遍关注的突出问题,质量在国家的战略高度、企业的竞争力和客户的需求等层面上都有极其重要的意义。研究证明,控制产品质量的根本,在于对产品加工过程即生产工序的控制,而非仅对加工成品的检验。另外,随着新技术的不断涌现和经济全球化进程的推进,制造企业的生存环境发生了巨大的变化,市场多元化、顾客需求多样化、零件个性化使产品更新换代的速度不断增加,多品种小批量生产方式日趋成为主要生产方式。但小批量生产的产品少,样本少,控制系统具有复杂性和不稳定性,因此无法获取稳定的质量特征值。而且,由于质量特征数据量少,无法满足经典统计过程控制(SPC)技术的统计量要求。本文立足于多品种、小批量制造模式的特点,结合统计过程控制在产品质量控制中的研究与应用现状,以多品种小批量生产制造模式下的工序质量控制为目标,对这种模式下的工序质量控制方法进行研究,主要研究工作如下： 1.通过对多品种小批量生产的实际制造过程进行分析,提出运用相似元理论,构建成组工序相似性识别模型,对小批量制造工序进行相似性评判,找出符合相似原则的工序,并划分同组工序,以充分利用工序之间的相似性信息,来拓展样本空间、增加样本容量。 2.基于成组相似工序,提出了质量特征数据转换算法和控制图的设计方法,为控制图的模式识别奠定了基础。提出了控制图基本异常模式的识别方法,并针对目前的控制图特殊异常模式识别方法适应性差、训练慢等问题,提出基于Elman神经网络的控制图特殊异常模式识别方法,并对Elman神经网络进行了网络设计。 3.为了给工序质量诊断提供更详尽的信息,构造了3个相似的BP神经网络对控制图特殊异常模式的特征参数进行估计。在对当前常见的工序质量诊断方案分析比较的基础上,将神经网络技术引入工序质量诊断中,提出使用神经网络的方法对工序质量进行诊断,并构造了工序质量诊断的神经网络模型。另外,根据工序质量异常原因,提出了工序质量控制策略。
[Abstract]:Quality is a prominent problem in modern industrial society and the economic construction of various countries, and quality is in the strategic height of the country. The competitiveness of enterprises and the needs of customers are of great significance. The research has proved that the fundamental to control the quality of products lies in the control of the process of product processing, that is, the production process. In addition, with the continuous emergence of new technologies and the advancement of economic globalization, the living environment of manufacturing enterprises has undergone tremendous changes, market diversification, customer demand diversification. The personalization of parts makes the speed of product renewal increase, and the production mode of multi-variety and small-batch is becoming the main mode of production day by day, but the products produced in small batch are less and the sample is less. The control system is complex and unstable, so it can not obtain stable quality eigenvalue. It can not meet the statistical requirements of classical statistical process control (SPC) technology. This paper is based on the characteristics of multi-variety, small-batch manufacturing model. Combined with the research and application status of statistical process control in product quality control, the process quality control method in this mode is studied, aiming at the process quality control in multi-variety and small-batch manufacturing mode. The main work of the study is as follows: 1. Through the analysis of the actual manufacturing process of multi-variety and small-batch production, the similarity recognition model of group process is constructed by using similarity element theory, and the similarity evaluation of small batch manufacturing process is carried out. In order to make full use of the similarity information between the processes, the sample space can be expanded and the sample size can be increased by finding out the processes that conform to the principle of similarity and dividing them into the same group. 2. Based on the similar working procedure in groups, the algorithm of quality feature data conversion and the design method of control chart are proposed, which lays the foundation for pattern recognition of control chart, and puts forward the recognition method of basic abnormal pattern of control chart. Aiming at the problems of poor adaptability and slow training of current control chart special abnormal pattern recognition methods, a special abnormal pattern recognition method based on Elman neural network is proposed. The Elman neural network is designed. 3. To provide more detailed information for process quality diagnosis. Three similar BP neural networks are constructed to estimate the characteristic parameters of the special abnormal pattern of control chart. This paper introduces neural network technology into process quality diagnosis, proposes a neural network method to diagnose process quality, and constructs a neural network model for process quality diagnosis. In addition, according to the cause of abnormal working procedure quality, the neural network technology is applied to process quality diagnosis. The strategy of process quality control is put forward.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TB114.2;TH16

【参考文献】