| When the motor fault is occurred, several comparatively independent working systems will be involved in it. Therefore, the motor fault diagnosis is more complicated than other normal mechanism diagnosis, which covers multiple technology fields, including Electric Machinery, thermodynamics and heat transfer, high voltage technology, material engineering, machine diagnosis, electronically measure, information engineering, computer technology, and so on. Because of several interactive systems working together, the fault reason and symptom for motor is multiform, that one trouble may cause another one and result in multi-fault coupling finally. The inter-coupling effect in kinds of fault types increases the difficulty of motor fault diagnosis. With the development of computer technology, data fusion and artificial intelligence theories, there has new solution for motor fault diagnosis. According to the motor fault characteristics, integrated diagnosis is implemented to solve this problem in this dissertation, which adopts modern calculation methods, such as wavelet analysis, fuzzy algorithm, artificial immune theory, multi-sensor information data fusion and so on.Main contributions of this dissertation are stated as below:(1) The typical fault types, fault mechanism, fault characteristics and fault diagnosis methods of motor has been analyzed and summarized in this paper. It's also be researched about the influence of motor electrical fault on motor vibration signals and the influence of motor vibration fault on motor electrical signals. The application of integrated multi-sensor information is discussed to improve the accuracy of motor fault diagnosis significantly.(2) Put forward an multi-sensor characteristic level fusion diagnosis method based on the artificial immune algorithm. The frequency bond energy information in current and vibration signal is extracted as motor fault feature, diagnosed the motor fault through Threshold method. The diagnosis results verify that the diagnosis method can obtain precise diagnosis result in single fault mode, at the same time the diagnosis method can quantificational describe the matching degree between motor signal and fault type, it can be apply to the decision fusion diagnosis. (3) According to the shortage that incompatible rules can't be disposed in traditional D-S evident decision rule, dynamic fuzzy inference method is introduced to decision fusion in the fault diagnosis field. Through the competition principle, majority rules accord with sample will be reserved and other incompatible rules will be deleted. Meanwhile, the redundant similar rules will be omitted when be judged by the threshold method.(4) Put forward an optimization method which applies decision level fusion to motor fault diagnosis based on immune algorithm. Initializing the fuzzy degree of membership function based on the distributing of matching degree in its value range, optimization the function with the evaluation function which constituted by output error and punished item; the redundant fuzzy divided elements have been deleted through immune and Threshold method; Optimization the rules in fusion system by multiple hump immune algorithm, because optimizing the rules in several different groups simultaneous, the optimizing speed is improved, and finding out all of the part optimization result.(5) Put forward a motor fault diagnosis method with multi-sensor mingles fusion method. First, fusing the congener sensors data by immune diagnosis method, and educing the part diagnosis result; second, fusing the uncongenial sensor data in decision level fusion by dynamic fuzzy inference. That diagnosis method not only can reducing the difficulty of fusion system optimization, but also can simulating the effect between different faults in coupling fault mode, obtain precise diagnosis result.
|
PDF Full Text Request