Research And Application Of Bearing Condition-based Maintenance In Production Line

With the continuous development of the railway equipment assembly line to automation and specialization,companies are paying more and more attention to production efficiency and economic benefits.And due to the integrity of the assembly line,if a key equipment component breaks down,it will cause the entire production line to run out of order and cause incalculable losses.Therefore,while ensuring the high-speed and stable operation of the enterprise's automated production line,it has become a growing concern for the enterprise to control the resources and cost effectively.On this basis,it is obvious that the traditional maintenance strategy,which is based on ex post maintenance and fixed maintenance,can not meet the needs of enterprises,and the concept of "condition-based maintenance" is more and more studied and applied by the enterprise.The condition-based maintenance is state-based maintenance,which is to know the equipment fault in advance,locate the problem,make the judgment according to the fault evaluation and fault trend,so as to carry on the more effective maintenance measure.Based on the condition-based maintenance,this paper puts forward the fault feature value prediction and fault classification algorithm of the bearing,and carries on the realization of the two algorithms in the enterprise production management system,analyzes the output result of the model,and carries on the visual maintenance management to the bearing.The specific research contents are as follows:(1)the fault types of bearing are classified by using the SOM(Self-Organizing Feature Map)neural network clustering model.However,the traditional SOM algorithm randomly selects the initial weight vector,which seriously affects the training effect of SOM network.In this paper,PSO-SOM algorithm is proposed.The PSO(Particle Swarm Optimization)particle swarm optimization algorithm is used to find the clustering center of all kinds of samples in the input samples.The clustering center vector is used as the initial weight vector of the SOM network.(2)For the prediction model of bearing eigenvalue,this paper uses wavelet neural network algorithm to predict the eigenvalue,but the traditional wavelet network weight correction process uses gradient descent method,and adopts a single direction search strategy.There are great defects in global search and optimization performance.In this paper,the FA-wavelet network algorithm is proposed.Using the individual moving search method of the FA(Firefly Algorithm)firefly algorithm,the weight value of the wavelet network is continuously optimized and the global search and optimization ability is improved.(3)Based on the validation and usability analysis of the two algorithms,this paper designs and implements the two algorithms,and applies them to the real state monitoring platform of the enterprise.According to the output result of the model,a maintenance strategy is proposed,which is mainly based on condition maintenance and supplemented by scheduled maintenance.And in the maintenance management module to provide maintenance log information feedback,more comprehensive maintenance of bearings according to the situation.