Data-Driven Design of Fault Detection Approach for Dynamic Systems Based on Subspace Identification Method

In the context of continuous advancements in modern industry and the escalating demands for the safety and reliability of control systems, timely fault diagnosis of control systems becomes paramount. Driven by this imperative, this article devotes to data-driven design of fault detection systems based on subspace identification method. For this purpose, a fault detection approach based on Subspace State Space System Identification and model-based full-dimensional observer design is firstly designed. It is followed by a data-driven fault detection approach based on one-step identification of stable kernel representation and direct design of the residual generator. Finally, the effectiveness and performance of the two proposed approaches have been verified through a case study on the laboratory threetank system.