The Study Of Intelligent Diagnosis And Life Prediction For Idler Bearing Fault Of Belt Conveyor

With the combination of big data and industry,the coal industry and the productivity is gradually improved.The development of the belt conveyor with large-scale,high-speed and intelligence has been driven by the shortage of coal transportation capacity.However,this leads to the belt conveyor bearing work more intensity,the probability of failure is also more higher.If the roller shell bearing lose efficacy,it would cause tremendous losses.Therefore,this paper aims to improve the reliability and safety of intelligent operation and maintenance of belt conveyor of idler bearing and make the research on intelligent diagnosis and life prediction of bearing fault of idler belt conveyor.According to the problem that the early weak fault signal of idler bearing is not obvious with severe working conditions,it is really hard to accurately clarify the fault category and predict the remaining life,a resonance sparse decomposition feature extraction method with the fusion of genetic simulated annealing algorithm and particle swarm optimization algorithm is proposed.The vibration signal of bearings with complex composition and modal aliasing is analyzed.The optimal quality factor is quickly and accurately searched for by using the fast global optimization characteristics of the fusion optimization algorithm to achieve the perfect separation of bearing fault signals.And the superiority of the proposed method is verified by simulation and experimental methods.According to the resonance sparse decomposition method of fusion optimization algorithm,the failure training set is determined by inverting the failure mode probability by using the rolling bearing fault tree model.Then,with the intelligent diagnosis method of bearing failure,the VGG-19 network model is used to diagnose the problem of low accuracy under the variable condition sample imbalance data set,and the cyclic generation adversarial network combined with the VGG-19 optimized by AdaBN to obtain the variable condition sample imbalance fault diagnosis network model.A countermeasure network is used to balance the sample data set,and AdaBN optimization algorithm is used to solve the fault diagnosis problem under off-design conditions.The countermeasure training method is used to train the diagnosis model,so that it has high accuracy and robustness of fault pattern recognition when the samples under off-design conditions are unbalanced.Based on simulation and experimental results,it is proved that the proposed method is superior to other network models in fault intelligent diagnosis under off-design conditions and unbalanced samples.Finally,the rolling bearing life prediction method combining convolutional neural network and bi-directional long and short term memory neural network is proposed with the resonant sparse decomposition feature extraction by the fusion optimization algorithm.According to the three properties of time dependence,monotonicity and robustness,the fault feature parameters are extracted from the separated low resonance components,and six parameter data sets that can characterize the bearing degradation features are identified as the input.Particle Swarm Optimization(PSO)is used to optimize the initial training parameters,and then the PHM-2012 bearing degeneration data set is used to confirm this method,which verifies that this method has higher prediction accuracy than other prediction models.There are 95 figures,32 tables and 101 references in this paper.