基于多元统计理论的间歇过程性能监控、质量预测和质量调控方法研究

项目名称： 基于多元统计理论的间歇过程性能监控、质量预测和质量调控方法研究

项目编号： No.61304116

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 罗利佳

作者单位： 浙江工业大学

项目金额： 24万元

中文摘要： 间歇过程作为一类重要的工业生产方式，广泛应用于生物制药、精细化工等领域。间歇过程的过程安全和产品质量一直是人们关注的焦点。然而，由于间歇过程的过程特性极其复杂，对其进行有效的过程监控和质量控制是亟待解决的难题。本项目以间歇过程为研究对象，利用海量过程数据，借助多元统计理论，结合流形学习和张量分析等方法，围绕过程监控、质量预测和质量调控等关键问题展开研究。具体内容包括：（1）研究数据全局-局部结构分析方法，提出基于张量数据结构分析的间歇过程监控方法；（2）研究基于多向并行偏最小二乘法的间歇过程质量相关故障监测、质量预测和保质调控方法；（3）针对间歇过程的多时段特性，研究基于张量模糊聚类技术的时段软划分方法以及基于相关分析的时段定位和关键时段辨识方法，提出基于时段的过程监控、质量预测和质量调控方法。本项目旨在形成一套面向实际需要的间歇过程综合性能监控、质量预测和质量调控理论方法体系。

中文关键词： 间歇过程；过程监控；质量预测；故障诊断；多元统计分析

英文摘要： As an important industrial mode of production ，the batch process has been widely applied in biopharmaceutical and chemical industries etc. The batch process safety and product quality always is a focus of people's attention. However, due to the complicated process dynamical features of batch processes, the process monitoring and quality control become urgent problems need to be solved. In this work,the data-based batch process monitoring, quality prediction and quality control techniques are studied by combining the multivariate statistical theory with other methods such as manifold learning,tensor analysis, and so on. The specific content includes: (1) studying the data global-local structure analysis method, and proposing the tensor data structure anlysis based batch process monitoring method,(2) studying quality-relevant fault monitoring, quality prediction and quality control techniques based on the multi-way concurrent partial least squares, (3) for multiphase batch processes, studying a soft phase partition algorithm based on the tensor fuzzy clustering technique, phase positioning and critical-to-quality phase identification methods based on the correlation analysis, and proposing phase-based process monitoring, quality prediction and quality control techniques. This work aims at forming an applicable the

英文关键词： Batch process；Process monitoring；Quality prediction；Fault diagnosis；Multivariate statistical analysis