Research On Fault Modes Classification Of Key Components Of Complex Equipment Based On Data-driven

With the advancement of the industrial level,the production equipment of manufacturing enterprises is becoming more and more complicated,intelligent and large-scale,which increases the possibility of failure and increases maintenance costs.Therefore,it is necessary to classify the fault modes of the key components of the mechanical equipment.And determine the fault type and the severity of the fault in time so that perform corresponding repair or replacement work and reduce maintenance costs,improve equipment reliability,and ensure normal operation of the equipment.Among the key components of mechanical equipment,rolling bearings are the most important and most prone to failure.Therefore,this paper takes the rolling bearing as the research object and establishes a data-driven rolling bearing fault mode classification model.Based on the data-driven method,it can be divided into multivariate statistical learning related methods and machine learning related methods.The Mahalanobis-Taguchi system(MTS)is a typical multivariate statistical learning method,and the support vector machine(SVM)is a typical machine learning method.In this paper,the original signal collected by the vibration sensor is processed by EEMD method and a series of Instrinsic Mode Functions(IMF)are obtained.The IMF sensitive discriminant algorithm is constructed to screen out the IMFs with high correlation with the fault information.Then calculate the time domain and frequency domain characteristic parameters of each IMF and the energy entropy parameters of the original signal.The initial multidimensional feature space is constructed.Since MTS and SVM are essentially two classifiers,this paper constructs a partial binary tree structure,and uses the Euclidean distance of the inter-class distance method to determine the order of the binary tree,thus the fault mode multi-classification model of the MTS and SVM are constructed respectively.By analyzing and comparing the results of the experimental data of two models,the corresponding conclusions are obtained.In this paper,the experimental data of the Case Western Reserve University rolling bearing experimental platform is used to verify the validity of the model.The failure modes of rolling bearings are classified into fault type and fault severity.The fault types are divided into four categories,which are normal,inner raceway,outer raceway and rolling body.Each fault type is diviede into minor,moderate and severe,so there are ten categories.In this paper,four sets of control experiments are set up.The results of the first three sets of experiments verify the effectiveness of the proposed method.The results of the group four show that the accuracy of the MTS fault types classification of rolling bearing is basically the same as that of the SVM,but the efficiency of the model is higher than the latter.However,the SVM has a much better effect on the severity classification of rolling bearing faults.At the same time,when the dimension of the feature space is high,the MTS is no longer applicable,but SVM can easily process high-dimensional data.