Research On Performance Degradation Evaluation Of Rolling Bearing Based On Optical Fiber Sensor

With the rapid improvement of modern industrial level,mechanical equipment is gradually developing in the direction of intelligence.At present,in order to meet the production requirements,the production pressure is gradually increasing,but the working environment of the mechanical equipment is harsh and changeable.During the work,the equipment will gradually age and the remaining service life will continue to decrease.Once a failure occurs,it will cause incalculable losses.Rolling bearings are important transmission parts in mechanical equipment,so it is essential to evaluate their performance degradation.In recent years,because of its unique sensing characteristics and material advantages,fiber Bragg gratings are currently being researched in the field of performance degradation of transmission parts.The use of FBG to build a rolling bearing detection system has higher sensitivity and implementability than traditional vibration sensors.Therefore,designing a bearing fault diagnosis and performance degradation evaluation system based on FBG sensors is of great significance for FBG research in this field.Aiming at the problems that traditional rolling bearing fault diagnosis methods are susceptible to environmental interference and poor recognition accuracy,a compound fault diagnosis method for rolling bearing with improved convolutional neural network is proposed.First,the bearing vibration signals of different faults are obtained through the FBG rolling bearing detection platform,the signal is decomposed using the empirical modal decomposition(EMD)algorithm,and the comprehensive index composed of Pearson correlation coefficient and kurtosis value is used to screen for useful indicators.Intrinsic Mode Function(IMF)component to ensure the correlation between the component signal and the original signal.Next,the structural characteristics of the IMF component are used to expand the component in the form of sampling points,extract the amplitude of each sampling point,form a two-dimensional matrix,and input to the Convolutional Neural Network(CNN)for feature Extraction,output high-dimensional vectors after convolutional layer and pooling layer,and then use decoupling classification algorithm to classify and identify single faults and composite faults of bearings.In view of the low accuracy of current rolling bearing performance degradation prediction methods,an empirical mode decomposition and Support Vector DataDescription(SVDD)rolling bearing performance degradation prediction method is proposed.This method first obtains the full life data samples of rolling bearings,uses EMD algorithm to decompose the data,filter useful IMF components,extract various indicators as features,compose feature vectors to input to SVDD for training,and obtain the samples and initial values for each time period The Euclidean distance between the samples is used to evaluate the degradation performance of the bearing.The FBG rolling bearing test experimental platform was built to verify the effectiveness of the above method.The experimental results show that the proposed method can effectively identify the normal,outer ring fault,inner ring fault,roller body fault and composite fault types,and the recognition accuracy rate is more than 90%.And use the above rolling bearing performance degradation prediction method to verify the bearing life in the FBG rolling bearing detection experimental platform.The results show that this method can successfully detect the degradation process of the bearing and can effectively predict the remaining service life.