Research On The Detection Technology Of Electric Motor Abnormal Noise

The detection technology of electric motor abnormal noise is the key process of motor production,and the traditional motor manufacturing industry relies on the method of manual listening to identify the existence of abnormal noise motor,which has low efficiency and poor efficiency consistency.It has long been difficult to replace with automated detection equipment.In this paper,the application of signal feature extraction and machine learning methods for air-conditioner plastic motor noise detection technology to conduct research.The main contents include.The first chapter introduces the background of the subject and the purpose and significance of the study,and analyzes the research status of the acoustic diagnosis of mechanical equipment and the application of machine learning in fault diagnosis.After summarizing the main difficulties that motor abnormal noise detection is facing,we gives the content of this paperIn the second chapter,the method of collecting and preprocessing the motor audio signal is studied.Firstly,the signal acquisition system is designed,which includes the selection of the sensor and the setup of the experimental device.Then eliminating the trend of the motor audio signal,pre-emphasis,frequency noise elimination and windowing preprocessing technology are adopted to reduce the error of the acquisition process audio signal interference.In the third chapter,the feature extraction technology of motor audio signal is studied.Firstly,the types of motor abnormal noise are analyzed and classified into two kinds of faults:electromagnetic noise and mechanical noise.Then,three methods,Mel Frequency Cepstrum Coefficient,Hilbert-Huang Transform and Wavelet Packet Transform,are respectively applied to get the characteristics of motor audio signal to and get the characteristic curves.The typical electromagnetic and mechanical abnormal noise samples were used to verify the recognition ability of the three methods for fault samples and qualified samples.The fourth chapter,based on the support vector data description,realizes the automatic recognition of motor abnormal noise.Firstly,the learning model of supportive vector data description is introduced,and the method and evaluation standard of training model are established.Aiming at the three feature extraction techniques in the third chapter,a method to describe the characteristic curves by means of average and peak-peak is proposed.Three methods of eigenvector are used to trains the learning model respectively and realize automatic recognition of motor noise.According to the characteristics of the type of motor abnormal noise,a hybrid motor fault identification method based on Mel Frequency Cepstrum Coefficient and wavelet packet energy spectrum is proposed,which can effectively improve the recognition rate of fault samples and qualified samples.The fifth chapter introduces the development of a prototype motor abnormal noise detection system,including training modules and prediction module.The task of the training module is to adjust the model parameter optimization model by collecting or importing the existing sample data in the database,training the learning model and verifying whether the model meets the requirements.The task of the prediction module is to import the optimal model obtained from the training module and judge whether there is abnormal noise in the new motor sample.