基于核方法的冷轧带钢产品质量监控与诊断研究

项目名称： 基于核方法的冷轧带钢产品质量监控与诊断研究

项目编号： No.51204018

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 冶金与矿业学科

项目作者： 何飞

作者单位： 北京科技大学

项目金额： 25万元

中文摘要： 冷轧带钢产品广泛应用于建筑、汽车、家电、机电产品等行业。我国每年生产8000万吨冷轧带钢产品，但由于生产工艺复杂，高端产品质量与国外同类产品相比有一定差距。本课题针对冷轧带钢生产中多变量强耦合、变量类型混杂、过程非线性等特点，采用非线性统计建模中的核函数方法，研究生产工艺参数与产品质量之间的内在关系，建立产品质量监控模型，实现冷轧带钢产品质量诊断，定位异常工艺参数，缩小人工分析范围。主要研究内容包括：研究基于核熵成分分析的建模方法，定位引起带钢宽度和厚度质量异常的工艺参数；研究基于流形距离核熵聚类和核统计特征量模式分析的诊断方法，解决带钢板形识别与异常诊断问题；研究基于多核相关向量机的规则抽取建模方法，解决带钢表面质量异常下工艺参数的定位和调整问题。本课题的预期研究成果可以推广到连铸、热轧等冶金生产过程，而且在石油、化工、机械制造等领域有着广泛的应用前景，具有重要的理论价值和实际工程意义。

中文关键词： 核函数；带钢；质量监控；质量诊断；

英文摘要： The cold rolled steel strip is widely used in construction, automotive, household appliance manufacturing, electromechanical products, and so on. In China, 80 million tons of cold rolled steel strip products are manufactured every year. Due to the complex production process, there is some discrepancy in the top products compared with similar foreign products. Because the cold strip rolling is the multi-variable, strong-coupled, complex variable types and nonlinear system, the kernel method as the nonlinear statistical model is used in this project to mine the essential relationship between the process parameters and product quality. Then, the quality monitoring model is built to diagnosis product quality, fix on the abnorm process parameters or reduce the manual analysis range. The main research topics include: Research on the kernel entropy component analysis to orient the abnormal process parameters that result in the strip thickness or width abnormity. Reseach on the kernel entropy clustering based on the manifold distance and the kernel statisical pattern analysis to resolve the shape defect recognition and diagnosis. Research on rule extraction based on multiple relevant vector machine to orient and adjust the abnormal parameters when there are strip surface defects. The desired research results will be use

英文关键词： Kernel function；Strip；Quality monitoring；Quality diagnisis；