Vibration Measuring Points Optimal Configuration And Intelligent Fault Diagnosis For Shearer Rocker

As the main comprehensive mining equipment for coal mining,normal operation of the shearer has a direct impact on the production efficiency and economic benefits of the enterprise.As an important but vulnerable part of the shearer,the rocker directly affects the working status and life cycle of the entire shearer equipment.The internal transmission system of the rocker is huge and complicated,whether it can work normally affects the working state of the entire rocker.Therefore,it is very important to accurately and effectively monitor the internal transmission system of the shearer's rocker and perform accurate fault diagnosis on it.Aiming at the above problems,an improved simulated annealing algorithm to obtain vibration measuring points configuration scheme of rocker and a fault diagnosis method based on deep learning is proposed.This method can not only achieve accurate monitoring of the internal transmission system of the rocker,but also intelligently identify the detected signals,which greatly reduces labor costs and improves the efficiency of fault diagnosis.For the question of monitorring the internal transmission system of the shearer's rocker accurately,a method of vibration measuring points configuration based on improved simulated annealing algorithm is proposed.The corresponding objective function is established based on the theory of optimal configuration of vibration measuring points,and the modal displacement data corresponding to the position of initial measuring points is obtained by modal analysis of the threedimensional simplified model of the shearer rocker shell,which is used as the input data of optimal configuration calculation of vibration measuring points.Improved simulated annealing algorithm is used to optimize calculation of the objective function built.Finally,result of the optimal calculation and the actual working environment of the shearer are considered together,a reasonable rocker vibration point configuration scheme is obtained.At the same time,the convergence and stability of the improved simulated annealing algorithm is verified.The method of fault diagnosis is researched,and a method of fault diagnosis based on one dimensional deep convolutional neural network is proposed to achieve intelligent and efficient fault diagnosis.The shearer's rocker loading test bench is used to simulate the five working states of the cutting two axes gear,which include normal working,tooth surface wear,tooth surface scratch,tooth surface damage due to gluing and bearing outer ring wear and the common fault types in the high-fault area of the shearer's rocker,and corresponding vibration signal data was obtained.For the data set of cutting two axes,the maximum prediction accuracy of the one dimensional deep convolutional neural network on the training and test sets was 100%;For the data set of the high-fault area of the shearer's rocker,the accuracy of the one dimensional deep convolutional neural network model on the training set is 100%,and the accuracy of the fault diagnosis on the test set is 99.697%,which verifies effectiveness and feasibility of the one dimensional deep convolutional neural network.In order to improve accuracy and stablity a fault diagnosis,a fault diagnosis method based on one dimensional deep convolutional neural network and ensemble naive bayes classifier is proposed.The Naive Bayes classifiers are integrated by using the ensemble learning Bagging algorithm and the output layer Softmax of the traditional one dimensional deep convolutional neural network is replaced with the ensemble Naive Bayes classifiers to obtain a new model of fault diagnosis.The data set of the five working states of the cutting two axes gear,which include normal working,tooth surface wear,tooth surface scratch,tooth surface damage due to gluing and bearing outer ring wear and the common fault types in the high-fault area of the shearer's rocker is used to train and test hybrid neural network model built,which shows the stability and accuracy of the hybrid neural network model are superior to traditional one dimensional deep convolutional neural networks.