Study On The Fault Diagnosis Of Rotating Machinery Based On Noise Signal

Rotating machinery is one of the most basic machinery and it is widely used in mechanical equipment.It is directly related to the work efficiency,economic benefits and the safety of the workers.So the study of rolling rotating machinery has important practical significance.However,the commonly used fault diagnosis methods of vibration signals can not meet the requirements because of their own characteristics.The text combined with advantages of the noise signal monitoring(non-contact measurement,simple equipment,easy signal measurement,easy to find fault,without prior paste sensor,the on-line monitoring,and so on)and research on the method of fault diagnosis for noise signal.Firstly,in this paper,the common fault types and fault mechanism of rotating machinery are introduced in detail.Based on the theory of acoustics,this paper explains the feasibility and representation of fault diagnosis based on noise signal.And the paper introduces the method of fault diagnosis by using the noise signal,and the principle and characteristic of each method.At the same time,the characteristics of the noise signal of the rotating machinery are analyzed.Secondly,according to the research purpose and research background,combined with the existing experimental equipment design experiment.The experiment is divided into two parts,the one is the fixed-point data acquisition,the other one is the mobile data acquisition.The main experimental objects include rolling bearing fault data,rotor shaft failure data,gearbox gear fault data and mixed fault data.Thirdly,research on fault feature extraction based on acoustic signal using acquired fault signal.Short-time Fourier transform(STFT)are introduced into the feature extraction to verify the fault information contained in the noise signal.According to the results,it is found that the noise signal contains more complex and powerful background noise.So,consider using EMD denoising,empirical mode decomposition and kurtosis or correlation analysis are combined to realize the decomposition and reconstruction of signals,and then realize the signal denoising.Compared with the frequency spectrum before and after denoising,it can be seen that the denoising method has a better effect on the signal.Because of the existence of frequency aliasing problem,the fault feature extraction method based on ensemble empirical mode decomposition(EEMD)is studied.Then,based on the characteristics of the collected noise signals,a method of sound data processing based on energy is proposed.Finally,the paper studies the classification method of fault signal of rotating machinery.Support vector machine is applied to fault feature classification.Optimal model parameters of the classifier are searched through cross-validation method.The fixed point data and mobile data are identified,and then the mixed fault signals are identified and classified.The classification accuracy of noise signal is observed by support vector machine.