基于数据驱动的过程监测系统设计方法研究

Data-driven Design of Process Monitoring Systems

为保障工业系统高效安全地运行,主要研究对于一般工业系统基于数据驱动的过程监测系统设计方法。首先,借助互质分解技术学习故障诊断观测器的参数化形式,在已有数据驱动设计方法的基础上,将单维的基于观测器的残差产生器扩展至多维,研究基于多维残差产生器的过程监测系统设计。其次,针对工业系统中普遍存在的非线性、系统不确定性和工作点变化等问题,基于自适应观测技术还为多维残差产生器设计了自适应更新策略。最后通过在三水箱仿真系统中的案例学习验证了所提出的数据驱动设计方法。

To ensure the safe and efficient operation of modern industrial systems,this paper focuses on the data-driven design of process monitoring systems.First,based on the coprime factorization technique the parameterization forms of fault diagnosis observers are studied.Motivated by the study on existing data-driven design techniques,a multi-dimensional data-driven observer-based residual generator has been proposed.And then,in order to tackle the common existing nonlinearity,uncertainty and the operating point change of the industrial system,an adaptive method is proposed for the multi-dimensional data-driven residual generator.At last,the effectiveness of the proposed design schemes are demonstrated using a three-tank benchmark system.