Robust fault diagnosis with parameter uncertainties based on bond graph

Abstract

The hybrid system consists of discrete subsystems and continuous subsystems. The variable conversion in the system is complex and the parameters are uncertain, which leads to the high false alarm rate of fault diagnosis. For this problem, a single phase full bridge inverter is the research object, and a bond graph in linear fractional transformation(BG-LFT) is proposed to establish a diagnostic hybrid bond graph(DHBG) model of the system parameter uncertainty and to generate the adaptive threshold according to the model. Based on causal and structural properties of the hybrid bond graph, robust analytical redundancy relations for all valid modes are then derived from the DHBG, and the residual is evaluated by the adaptive threshold to realize the robust fault diagnosis of the hybrid system. Modeling and simulating in 20-sim, the simulation results verify the effectiveness of the proposed method.

Keywords

• bond graph
• parameter uncertainty
• linear fractional transformation
• adaptive threshold
• robust fault diagnosis