基于数据驱动的复杂工况过程监测方法研究进展

Research Progress on Data-driven Monitoring Methods for Complex Process

随着过程工业系统结构日益复杂,系统安全及产品质量的在线监控显得尤为重要。该文首先介绍了过程监测与诊断技术的研究背景和意义,对现有基于数据驱动的过程监测方法研究现状作了综述。重点阐述了针对单一约束条件,尤其是过程的非线性或动态性的多变量统计过程监测方法。由于非线性和动态性这两种特性在实际工业过程中普遍同时存在,研究此类复杂工况下的过程监测问题具有重要意义。最后,分析了采用流形学习方法解决非线性动态过程监测问题的可行性和挑战性,并指出非线性动态过程、故障检测与诊断以及机器学习新算法的研究将是过程监测与建模技术的近期发展趋势。

As industrial processes become more complex, plant safety and product quality are two important issues that should be paid great attention. The research background and significance of process monitoring and fault diagnosis technologies are introduced and the current development on data--based process monitoring methods is summarized in this paper. Then the multivariate statistical process control （MSPC） methods for process with single constraint, especially for single nonlinear process or nonlinear process are elaborated. As the modern industrial process data always show strong nonlinear and dynamic behaviors, study on monitoring tech- nologies for nonlinear and dynamic process is of great importance. In the end, the feasibility and challenging of manifold learning algorithms based process monitoring methods for nonlinear and dynamic process are analyzed. Some challenges such as nonlinear and dynamic process, fault detection and diagnosis as well as new machine learning methods are indicated.