基于SPC的数据采集和质量监控系统的技术研究

发布时间：2018-04-12 08:56

本文选题：质量监控 + SPC控制图模式　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：在2015年由国务院印发的《中国制造2025》行动纲领中,“质量为先”是其五大基本方针之一,表明产品质量已经成为国家的重点关注对象,也是目前制造企业生产过程的主要目标。因此,如何有效地对产品的生产过程进行质量监控以保证产品质量,是目前制造企业的重要研究问题。统计过程控制(Statistical Process Control,SPC)技术是目前生产制造领域对产品质量进行监控的主要技术。根据SPC控制图的图像特征可以判断生产过程是否出现异常,通过对SPC控制图模式的识别可以推断生产过程发生异常的原因。近年来研究人员采用多种智能算法用于对SPC控制图模式的识别研究,由于目前生产过程控制图模式的样本较少,而支持向量机(Support Vector Machines,SVM)可以有效解决小样本条件下的模式识别问题,因此本文将支持向量机作为控制图模式识别的主要研究工具。本文主要做了以下研究:1)对SPC控制图模式数据进行仿真,并通过统计特征和形状特征对其进行特征提取,通过研究、分析支持向量机原理及其多分类方式的特点,选择“有向无环图”(Directed Acyclic Graph,DAG)型支持向量机作为多分类器,并采用粒子群算法(Particle Swarm Optimization,PSO)对其参数进行优化,然后通过相应的仿真实验进行对比分析;2)为了进一步提高控制图模式的识别效率及准确率,提出两种改进方式:首先通过特征融合将原始数据和特征数据的识别优势相结合,然后通过主成分分析(Principal Component Analysis,PCA)算法对融合特征进行进一步的维数约简,提取出对分类影响较大的数据特征,从而提高了分类准确率和识别效率;对粒子群算法进行改进,通过增强粒子的主动搜索能力,解决其易陷入“局部最优”的缺陷,从而提高了分类器识别的准确率。仿真实验结果表明,通过对识别及优化算法的改进,分类器对控制图模式的识别效率和准确率方面都有了相应的提高,识别率能够达到95%以上,能够满足基本的生产需求,有效预防生产异常现象的出现。
[Abstract]:Therefore, how to effectively monitor the quality of the production process to ensure the quality of products is an important research issue in manufacturing enterprises at present.Statistical Process Control (SPC) technology is the main technology to monitor the product quality in the field of production and manufacturing.According to the image features of SPC control chart, the abnormal production process can be judged, and the reason of abnormal production process can be inferred by recognizing the pattern of SPC control chart.In recent years, researchers have used a variety of intelligent algorithms to study the pattern recognition of SPC control charts.Support vector machine (SVM) can effectively solve the problem of pattern recognition under the condition of small sample, so support vector machine is regarded as the main research tool of control chart pattern recognition in this paper.In this paper, we do the following research: 1) simulate the pattern data of SPC control chart, and extract the feature by statistical feature and shape feature. Through the research, we analyze the principle of support vector machine and the characteristics of multi-classification.The support vector machine "directed Acyclic directed Acyclic Graph DAG" is selected as multi-classifier, and its parameters are optimized by particle swarm optimization algorithm (PSO).In order to further improve the recognition efficiency and accuracy of the control chart, two improved methods are proposed: firstly, the recognition advantages of the original data and the feature data are combined through feature fusion.Then, the principal component analysis (PCA) algorithm is used to reduce the dimension of the fusion features, which has a great influence on the classification, which improves the classification accuracy and recognition efficiency, and improves the particle swarm optimization algorithm.By enhancing the active searching ability of particles, the defect of local optimum is solved, and the accuracy of classifier recognition is improved.The simulation results show that by improving the recognition and optimization algorithm, the classifier can improve the recognition efficiency and accuracy of the control chart pattern, and the recognition rate can reach more than 95%, which can meet the basic production requirements.Effective prevention of abnormal production phenomenon.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB114.2

【参考文献】