Faults Diagnosis Of Motor Based On EMD And Feature-fusion

In recent decades, disastrous accidents caused by motor faults often occurred, resulting in huge economic loss. In order to fundamentally avoid disastrous accidents, faults diagnosis of motor must be taken.Through the diagnosis of motor faults, the faults can be found earlier and prevented further deterioration, which not only can reduce enormous economic loss caused by disastrous accidents but also can provide some useful information on motor for motor manufacturers to improve performance and reliability of motor.In this paper, the diagnosis methods of motor faults are analyzed and studied, and finally the acoustic analysis method is adopted, which mainly include four aspects that are acquisition the motor acoustic signal, analysis and processing the motor acoustic signal, construction feature vector and faults pattern recognition. So the research and exploration in this paper is based on the above four aspects. The following findings have been achieved.1) A acoustic collection system based on two microphones to acquire motor faults information is designed and implemented. The hardware mainly includes microphones, preamplifier circuit, anti-aliasing filter circuit, post-amplifier circuit and data acquisition card; The software is developed based on the platform of LabVIEW with favorable operation interface. Compared with the traditional acoustic collection system based on a single microphone, the collection system in this paper can obtain more information of motor faults, which is contributed to improve the accuracy of diagnosis.2) A method of processing motor acoustic signal based on empirical mode decomposition is proposed. Firstly, the method, obtains natural harmonics of motor acoustic signal by characteristic time scale of motor acoustic signal, and then decomposes the motor acoustic signal based on natural harmonics. The basis fuction used to decompose motor acoustic signal is from the motor acoustic signal itself. The experimental results show that the method can effectively decompose acoustic signal.3) A feature vector construction method based on the combination of the energy of intrinsic mode function and feature-fusion is proposed for the first time. Firstly, the method take the energy of instrinsic mode function as feature parameter, and then fuse the two groups of feature parameters set into future vector by using feature-fusion technology, which not only can retain more information about motor faults but also can avoid the high dimensionality and redundancy of feature vector. The method provides a new way to construct the feature vector.4) Pattern recognition of motor faults is carried out with the feature vector. In this paper, the neural network and support vector machine are both applied to the faults diagnosis of motor. The final results of recognition are compared and analyzed, which show that the support vector machine is more suitable for the feature vector constructed in this paper and verify the validity of the motor faults diagnosis method based on the feature vector constructed in this paper.