Research Of Fault Diagnosis Based On Iterative Learning Algorithm

As an important guarantee for the safety of control systems, fault diagnosis technology has been a popular research topic for experts and scholars at home and abroad, and it has formed a comprehensive discipline with numerous branches after decades of development. Model-based approach is one of the major research subject in fault diagnosis, it has the features of clear logic and high diagnostic accuracy because the mathematical model of the system is used as its diagnostic tool. It is a new method of model based fault diagnosis that combining iterative learning algorithm with observer theory to construct fault diagnosis filter so that the advantages of iterative learning algorithm such as simple structure and high control precision can be applied to the fault estimation, which provides a new inspiration to the research of fault diagnosis. The main contributions are as follows:1. A fault reconstruction approach based on filter is proposed for nonlinear systems with actuator and sensor fault. To make the algorithm applicable to both state and output side, a new state equation is constructed to transform and extend the system, which could convert nonlinear terms and fault of original system output to state equation of the extended system. Afterwards, a fault diagnosis filter is designed, iterative learning algorithm is chosen to update virtual fault to make it approximate to actual fault. This algorithm can detect and estimate system fault and has certain adaptability for faults of different types. Simulation results illustrate the feasibility and effectiveness of this algorithm.2. For a class of linear discrete systems with actuator faults and bounded disturbances, an improved discrete iterative learning diagnosis filter is proposed, in which the sequence of diagnosis time domain and iteration domain are exchanged so that the system faults can be diagnosed by sampling time sequence. Afterwards, a sliding average filtering principle based initial value estimation algorithm for virtual fault is proposed, in which fault information of previous sampling points are used to estimate the fault value of current sampling point before diagnosis. At last, the fault diagnosis is started with the estimated fault initial value. Simulation results show that the algorithm proposed can reduce the total number of iterations effectively and improve the efficiency of the fault diagnosis.3. For a class of non-uniform sampling system with actuator faults and bounded disturbances, an iterative learning fault diagnosis algorithm is proposed. Firstly, the actual fault function is transformed to an equivalent fault model by using the integral mean value theorem, then the sampling system is converted to a continuous systems based on the output delay method. Afterwards, an observer-based fault diagnosis filter is designed to estimate the equivalent fault, and the iterative learning regulation algorithm is chosen to update the virtual fault repeatedly to make it approximate the actual equivalent fault. Simulation results of an electro-mechanical control system model illustrate the effectiveness of this algorithm for fault estimation of the non-uniform sampling system.