Fault Prediction Of Rolling Bearing Based On General Regression Neural Network Optimized By Drosophila Algorithm

Rotating machinery plays an important role in complex equipment systems such as aviation,spaceflight and shipping.Rolling bearing is one of the most important parts of rotating machinery whose health state has a great influence on the whole equipment.Most of the current research is the fault diagnosis and evaluation of the current state of the machine,however,the research on its health management and fault prediction is less.Therefore,the failure prediction of the rolling bearing can be more helpful for its pre-known maintenance and health management.First,we deal with vibration signal for rolling bearings for noise reduction.In order to solve the modal aliasing problem in EMD and EEMD,the CEEMD method is used to reduce noise of the vibration signal data and decompose the IMF component.The effectiveness of the method is proved by being compared with the noise reduction effect of EMD and EEMD.Secondly,because the multi-scale entropy algorithm of coarse graining is easy to lose information and other shortcomings,we use the feature extraction method of CEEMD and multiscale entropy to extract the feature of IMF component after screening.The superiority of the improved method is verified by comparing the feature extraction results with the common multi-scale entropy method.Thirdly,aiming at the problem of low accuracy of fault prediction and difficult establishment of prediction model,a smoothing regression algorithm based on Drosophila algorithm is used to optimize the smoothing factor of generalized regression neural network(GRNN),so as to establish a generalized regression neural network(FOA GRNN)optimized by fruit fly algorithm.Finally,the vibration data collected from the rolling bearing fault simulation test platform are verified by an example.The signal denoising feature extraction and fault prediction are carried out to prove the superiority of the prediction method proposed in this paper.