Abstract: Aiming at the missing detection and erroneous detection in fault detection, a method based on residual and fault estimation is proposed. Firstly, with the application of Schur complementary lemma, the method of sliding-mode observer gain matrix is transformed into a multi-objective convex optimization problem under the restrict of LMI,and a reasonable gain is designed to solve the robustness problem. Secondly, according to the residual randomness, statistical theory is introduced to determine the fault detection threshold, which makes the fault detection results more credible. Then, in order to reduce missing detection and erroneous detection, the sliding mode observer is designed to generate the residual and realize the first fault detection, the fault estimation is conducted by equivalent output injection theory to realize the second fault detection. Finally, simulation instance proves the effectiveness of the proposed methods.

Key words: fault detection, missing detection and erroneous detection, sliding mode observer, residual, fault estimation