Ran Guangtao, Chen Hongtian, Li Chuanjiang, Ma Guangfu, Jiang Bin
IEEE Trans Neural Netw Learn Syst. 2023 Sep;34(9):5244-5254. doi: 10.1109/TNNLS.2022.3174822. Epub 2023 Sep 1.
To ensure the safety of an automation system, fault detection (FD) has become an active research topic. With the development of artificial intelligence, model-free FD strategies have been widely investigated over the past 20 years. In this work, a hybrid FD design approach that combines data-driven and model-based is developed for nonlinear dynamic systems whose information is not known beforehand. With the aid of a Takagi-Sugeno (T-S) fuzzy model, the nonlinear system can be identified through a group of least-squares-based optimization. The associated modeling errors are taken into account when designing residual generators. In addition, statistical learning is adopted to obtain an upper bound of modeling errors, based on which an optimization problem is formulated to determine a reliable FD threshold. In the online FD decision, an event-triggered strategy is also involved in saving computational costs and network resources. The effectiveness and feasibility of the proposed hybrid FD method are illustrated through two simulation studies on nonlinear systems.
为确保自动化系统的安全，故障检测（FD）已成为一个活跃的研究课题。随着人工智能的发展，无模型FD策略在过去20年中得到了广泛研究。在这项工作中，针对事先未知信息的非线性动态系统，开发了一种结合数据驱动和基于模型的混合FD设计方法。借助高木-关野（T-S）模糊模型，可通过一组基于最小二乘法的优化来识别非线性系统。在设计残差发生器时考虑了相关的建模误差。此外，采用统计学习来获得建模误差的上界，在此基础上制定一个优化问题以确定可靠的FD阈值。在在线FD决策中，还采用了事件触发策略以节省计算成本和网络资源。通过对非线性系统的两个仿真研究，验证了所提出的混合FD方法的有效性和可行性。
