Research On Intelligent Fault Diagnosis Algorithms For Rotating Machinery Bearings

Rolling bearings,as the key driving components of rotating machinery,play an important role in the operation of the equipment.When it fails,it directly affects the safe and reliable operation of the equipment.Therefore,the intelligent fault diagnosis method for rolling bearings can be used to identify the running state of rolling bearings accurately and efficiently in real time,which has far-reaching significance and industrial value for ensuring the safe operation of equipment.The vibration signal of rolling bearing can reflect a large number of bearing status information.Therefore,this paper takes the rolling bearing vibration signal as the analysis object,combines information mining and artificial intelligence pattern recognition method to deeply research the intelligent fault diagnosis algorithm model of rolling bearings.Firstly,the traditional fault diagnosis model based on statistical features is studied.The statistical features of the three kinds of analysis fields in the time domain,frequency domain and time-frequency domain of the rolling bearing vibration signal are extracted.And the neural network and support vector machine(SVM)fault diagnosis model are respectively trained based on the statistical feature set to complete the fault pattern recognition.The validity of the two diagnostic models is verified by experiments.According to the experimental results,the SVM fault diagnosis model is more advantageous.Secondly,it is difficult to extract effective statistical features for the lack of prior knowledge,and the intelligent fault diagnosis model based on automatic feature extraction is studied.An intelligent fault diagnosis algorithm based on Frequency Spectrum Sparse Auto-Encoder(FSAE)and SVM is proposed.The automatic feature extraction of rolling bearing vibration signal is carried out by FSAE,and the fault diagnosis model is constructed by SVM classifier.The effectiveness of the algorithm is verified by experiments,and intelligent fault diagnosis can be completed without expert experience and has more accurate diagnosis results.Finally,the problem of the generalization performance of the fault diagnosis model is caused by the small amount of faulty sample data marked by the rolling bearing,and the fault diagnosis algorithm based on semi-supervised learning is studied.A fault diagnosis method based on K-nearest neighbor and SVM cooperative training is proposed.The algorithm takes full account of the influence of unmarked data on the diagnosis model.It uses the K-nearest neighbor algorithm to learn the spatial distribution of a large number of unmarked data,and assists the SVM to construct the classification of the entire data set.The fault diagnosis model training is completed by means of mutual correction of the two classifiers.The experimental results show that the proposed algorithm effectively improves the generalization performance of the model and has better accuracy and stability than the semi-supervised SVM classifier.