Research On Multiple-Model Estimation And Parameter Identification Methods For Typical Rotor Fault Diagnosis

Vibration signal characteristics have being basically used for rotorfault diagnosis. However, this method involves human experience andjudgment. Recent years, intelligent diagnosis methods have been invented,but the accuracy dependents on fault accident and fault data, whichbecomes the limitations of practical application. This paper focuses onrotor fault diagnosis methods based on multi-model estimation andparameter estimation, to rich rotor fault diagnosis means and improve thediagnosis capabilities. The main research work is as follows:1. Analyzed the mechanism and kinetics of a variety of typical fault(misalignment, rubbing, cracks and bending) to the single disk rotorsystem with rigid support respectively. Obtained the rotor differentialequations of motion under different fault type. Summarized single faultmodel and build multiple fault kinetic equation.2. Applied the multi-model estimation method to diagnosis of rotorsystem common faults, such as rubbing, misalignment and cracks. Thestudy showed that the multi-model estimation method has a goodapplication for single unknown parameters model. But when it comes to multi-parameter unknown model, the parameter permutations greatlyincrease the number of the Kalman filters, thus making unnecessaryfiltering and reduce the calculation speed. Therefore, this paper putforward a new method of fault diagnosis of a multi-parameter unknownmodel based on the combination of extreme value range search methodwith multi-model estimation theory.3. Applied the extended Kalman filter method to the bending fault,misalignment fault and crack fault diagnosis, introduced EKF-WGIalgorithm. As EKF may fail to a strongly nonlinear model, this paper putforward a new diagnosis method, EKF parameter estimation withmultiple model estimation joint method. Simulation confirmed theeffectiveness and practicality of the method.4. Applied particle filtering algorithm to the fault diagnosis ofrotating machinery. Diagnosed single fault, especially multiple faults ofthe rotor system with particle filter algorithm. Identified crack locationusing particle filter algorithm to single span double-disc rotor system.5. Simulated crack fault, misalignment fault, bending fault andbending crack multiple fault on rotor experimental stage. Fault diagnosisusing multi-model estimation method, EKF-GI method and particle filteralgorithm with fault vibration signal. Analyze the advantages anddisadvantages of each of the three methods in several ways.