基于模式识别的动态过程质量监控与诊断

项目名称： 基于模式识别的动态过程质量监控与诊断

项目编号： No.71272207

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 管理科学

项目作者： 刘玉敏

作者单位： 郑州大学

项目金额： 55万元

中文摘要： 本项目针对复杂动态过程难以建立精确的过程模型和采用统计质量控制技术进行实时质量诊断的局限性，研究基于模式识别的动态过程质量监控与诊断方法。首先，将根据动态过程的离线测量数据确定动态数据流的质量异常模式；其次，开展适用于分析质量异常模式的小波变换算法研究，并基于该算法设计出自适应特征提取算法，其中要解决的关键问题是如何提高特征提取算法精度并降低特征维数；进而针对不同异常模式的特征，基于小波变换的异常检测方法，设计小波变换与最近邻分类器、神经网络、支持向量机相结合的多个子分类器，建立自适应特征分类算法。最后，在综合分析子分类器对各种质量异常模式的识别结果的基础上，确定动态过程的质量诊断准则，提供动态过程实时质量监控与诊断方法。本项目的研究结果不仅为石油、化工等过程工业和卷烟生产过程等自动化制造过程提供实时质量监控与故障诊断技术，而且为其他行业的动态过程提供在线质量监控的理论依据和分析途径。

中文关键词： 动态过程；质量异常模式；小波变换；神经网络；支持向量机

英文摘要： Because the complex dynamic process is difficult to establish a precise process model or the use of statistical quality control technique for on-line quality diagnosis, this project researches quality control and diagnosis method for dynamic process based on pattern recognition.Firstly,according to the off-line measurement data, the quality anomaly patterns for dynamic process are determined by analyzing the quality abnormal information. Secondly, the wavelet transformation algorithm for the quality abnormal pattern is researched, and the adaptive feature extraction algorithm is designed based on the algorithm, which is the key problem how to improve the accuracy of the feature extraction algorithm and reduce the characteristic dimensions. Further, for the extractive features in different anomaly patterns, the more than one classifiers are designed based on the anomaly detection methods by wavelet transformaton, which are the wavelet transformation is combinationed with the nearest neighbor and neural network as well as support vector machine. The adaptive feature classification algorithm is set up.Finally, by means of the matching degrees of various quality abnormal patterns and the trained child classifiers, the dynamic process quality diagnosis criteria are determined for on-line quality control. This projec

英文关键词： dynamic process；quality abnormal pattern；wavelet transform；neural network；supprot vector machine