Study On Rotating Machinery Fault Diagnosis Based On CEEMDAN And Particle Swarm Optimization Algorithm

The rotating machinery is the machinery which completes the main function of itself by rotating.And it is a kind of common parts which is important to change the speed and torque transmission.In recent years,with the development of modern industry,rotating machinery is widely used in aerospace machinery,transport machinery,agricultural machinery,metallurgical machinery,mining machinery,engineering machinery,cutting machine tools and various production departments.Therefore,the working state of rotating machinery has significant impact on complete sets of equipment and even assembly line.The research and development of more effective fault diagnosis technology has become an urgent demand.This technology has great application in maintaining the safety of equipment and production personnel,improving production efficiency,reducing the equipment loss etc.With the support of Doctoral Fund of Ministry of Education of China(Number:2016M601800),taking common rotating machinery as the object,this paper research the fault diagnosis technology of rotating machinery in the aspect of the fault features extracting,fault pattern' s diagnosis and recognition,signal denoising,to seek more excellent fault diagnosis method.The main research contents of this paper includes:(1)In the light of that rotating machinery signal is too complex to find its effective feature,wavelet packet method is use to extract the energy characteristics data of the signal and lay the foundation for fault diagnosis using neural network method.In the light of the disadvantage,of BP neural network which is commonly used fault diagnosis,that is easy to fall into local optimal value and lead to the lack of diagnostic accuracy,the simple and useful Particle Swarm Optimization algorithm is proposed to optimize it.Through the collected vibration signal data,BP neural network optimized by Particle Swarm Optimization is compared to BP optimization without optimization in the diagnosis accuracy,and prove that this optimization method is helpful for rotation Rotating machinery fault diagnosis.(2)In view of the fact that the general denoising methods(including the wavelet packet)don't do well in denoising effect,CEEMDAN(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise)is try to be based on denoise the signal by the way of decomposition and reconstruction.And compared to EEMD(Ensemble Empirical Mode Decomposiyion)denoising of reconstruction,wavelet packet denoising,this method is proved that it has unique advantages in denoising the noise in the vibration signals of rotating machinery.(3)In the light of many disadvantages of method of empirical mode decomposition and its improved EEMD method which are commonly used to extract the characteristics of rotating machinery fault,the CEEMDAN method which was formerly applied to the biomedical field experiment is used to extract fault characteristic frequency of the rotating machinery,found that the recent improved method based on the empirical mode decomposition method has better effect of feature extraction and can make people extract the fault frequency from the map of decomposition results more easily.(4)Four kinds of diagnosis combination is set up to combine CEEMDAN and Particle Swarm Optimization algorithm,and compared in the instance,to further prove of the advantages of CEEMDAN in the aspect of denoising and advantages of Particle Swarm Optimization algorithm in the aspect of optimizing diagnosis.