基于贝叶斯方法的高鲁棒性故障检测技术

摘要

Fault detection was applied extendly in control engineering field,and robustness is an important measure of uncertain for a system while detecting faults. In order to improve the performance of fault detection in robust systems,we studied the problem system's fault and uncertain based on Bayesian principle, and proposed a Bayesian based robust fault detection method. Firstly,we defined the problem of fault detection for nonlinear systems,and analyzed their uncertainty of parameter space. Secondly,we analyzed how to infer the parameter space based on Bayesian principle. Finally,while detecting faults of a system,we analyzed the set membership between normal parameter space and faulty parameter space. The experiments showed that,in a quadruple-tank process,the proposed method could detect faults efficiently in a system with uncertain model parameters.%故障检测在控制工程领域具有广泛的应用,而鲁棒性是衡量系统在故障检测时的不确定性的重要指标.为了提高鲁棒系统故障检测的能力,基于贝叶斯的基本原理对系统的故障和不确定性进行研究,提出一种基于贝叶斯方法的高鲁棒性故障检测方法.首先,对非线性系统的故障检测问题进行定义,分析参数空间的不确定性问题.其次,基于贝叶斯基本原理对参数空间的推导进行分析.最后,通过分析系统正常运行的参数空间与发生故障时的参数空间之间的成员关系来进行故障检测.实验表明,在四罐耦合系统中,提出的基于贝叶斯方法的故障检测技术在系统模型参数具有不确定性的条件下可以很好地进行故障识别.