Robust Fault Diagnosis And Its Application In Nonlinear Electromechanical Systems

Electromechanical systems are widely used in production and life.In the process of development,it has the characteristics of complicated structure,scale expansion and function diversification,so the possibility of the fault in the electromechanical systems is increased greatly.The fault in the electromechanical system can bring huge impacts on production efficiency to cause disastrous economic losses.Therefore,it is of great significance to study the robust fault diagnosis method for electromechanical systems and the improvement of the fault detectability and isolability.This thesis is focused on uncertainties model of nonlinear electromechanical system based on bond graph(BG),and the adaptive threshold and multiple fault isolation methods.The main studies of the thesis is as follows.(1)The principle of BG method is introduced,which shows that the BG has an advantage in the modeling of multi-energy domain system.(2)The experimental platform of electromechanical system is discussed.The nonlinear BG model for electromechanical system with Stribeck friction is built.The BG based fault diagnosis which derives the analytical redundancy relations(ARR)of the system is developed.The resultant ARRs are used to analyze the detectability and isolability of the components.(3)The principle of linear fractional transformation(LFT)is introduced.The BG-LFT model is established to separate the uncertainty of the components from the nominal components.The adaptive threshold is derived from the BG-LFT.Based on the adaptive threshold,the BG-LFT model is proposed to improve the adaptive threshold,where the fault detection performance can be improved.(4)A multiple fault isolation method and fault isolation signature matrix are proposed.The component isolability is obtained by analyzing the fault isolation signature matrix.The multiple fault isolation method is used to isolate the fault by calculating the residual of ARR online,by which the fault isolability can be improved.(5)The effectiveness of adaptive threshold and multi-fault isolation algorithm is verified by simulation and experiment results.