模型未知非线性系统性能监督的故障检测

Performance-Supervised Fault Detection for Model-Free Nonlinear Systems

摘要: 随着工业快速发展及系统复杂度提升，非线性系统故障检测成为保障复杂工程系统安全性与可靠性的重要技术手段。传统方法大多依赖观测器生成的残差信号构建监测机制，普遍忽视了故障对系统关键性能的影响。针对复杂非线性系统的实时性能监测与维护需求，本文提出一种数据驱动的性能监督故障检测新方法。首先基于Takagi-Sugeno (T-S)模糊逼近技术，构建非线性系统性能残差函数，其次设计数据驱动方案对模型未知系统性能残差中的重要参数进行辨识，最后建立性能残差的评价函数，并设置兼顾故障检测率与漏报率的阈值。以实验室三容水箱系统为对象进行的仿真研究表明，该方法可有效检测水箱管道堵塞与液位传感器故障，验证了其在复杂动态工程系统中的实际适用性。

Abstract: With the rapid development of industry and the increase in system complexity, fault detection of nonlinear systems has become important technical means to ensure the safety and reliability of complex engineering systems. Traditional methods mostly rely on residual signals generated by observers to establish monitoring mechanisms, yet they generally overlook the impact of faults on the key performance of systems. To address the needs of real-time performance monitoring and maintenance for complex nonlinear systems, this paper proposes a novel data-driven approach for performance-supervised fault detection. First, with the aid of Takagi-Sugeno (T-S) fuzzy approximation techniques, a performance residual function is constructed for the nonlinear systems. Then, a data-driven scheme is designed to identify key parameters in the performance residual of the systems with unknown models. Finally, an evaluation function is established for the performance residual, and a proper threshold is set to ensure the trade-off between the fault detection rate and the false alarm rate. Simulation studies conducted on a laboratory three-tank system demonstrate that the proposed method can effectively detect the pipe plugging and level sensor faults, verifying its practical applicability in engineering systems with complex dynamics.