Research On Malfunction Classification Monitoring Of Bearing Equipment Based On Adaboost Algorithm

As years go by,the application of artificial intelligence,big data,and various algorithms in machine learning has gradually penetrated all walks of life.It is reflected in the fields of agriculture,military affairs,manufacturing,and the chemical industry.Especially in the industrial field,the application is more remarkable.For example,machine vision inspection can be applied to the shell integrity inspection of mobile phone manufacturers,or the location,recognition,and classification of objects on the assembly line,as well as the fault diagnosis of industrial equipment,etc.In the era of large machines,artificial intelligence is needed to improve our operational efficiency.The main body of this paper is bearing equipment,as a very important component in modern large machinery and equipment,it can play an important role in production and operation,it mainly uses bearings to support the rotating part of large machinery,which can reduce the friction force of the machine in the whole use process,thus ensuring the rotating accuracy of the machine.Therefore,the working state of bearings determines the overall efficiency of the whole operation process,so it is necessary to study the bearing state.The research of bearing equipment status is mainly realized by the following steps:1 、 Relying on machine learning related algorithms,enumerate the possible problems in the existing bearing operation data and deal with them pertinently,and then carry out feature processing,including feature selection and feature generation,then,the data that has been pre-processed are constructed with multi-classification samples,and the samples in the data set are labeled with classification labels,which have four states: fault-free,outer ring fault,inner ring fault,and ball fault.At the same time,the data imbalance problem existing in the multi-classification samples is efficiently processed,and a data set with balanced sample distribution is constructed.2、After processing the data problems,start to build a multi-classification model to deeply explore the performance of bearing equipment,thereby realizing the classified monitoring of bearing equipment faults,through a variety of algorithms,that is,logistic regression,artificial neural network,random forest,Adaboost and other algorithms to build a multi-classification model.At the same time,several evaluation indexes are introduced to score and evaluate each model,In this paper,we first use Kappa coefficient,Macro＿F1＿Score,Weighted＿Macro＿F1＿Score and Jecard similarity coefficient to score the four models.On the basis of considering the importance of the above four evaluation indexes,we continue to consider the weighted sum of Kappa coefficient,Weighted＿Macro＿F1＿Score and Jecard similarity coefficient,and introduce a comprehensive evaluation index to evaluate the bearing fault classification model simply and intuitively,which makes the comprehensive cognition of the model more direct.After various comparisons,Adaboost model gets a high score of 0.89 under the comprehensive evaluation index,which can determine the fault types of bearings with high accuracy and perform very well.It can further prescribe the right medicine for the detected faults and treat them accurately,thus ensuring the high efficiency and smoothness of the whole operation process.