Residual Life Prediction Of Key Parts Of Shearer Rocker Arm Based On LSTM Network

As one of the modern mechanical equipment of coal mining,the healthy and stable operation of shearer is the basis of mine production efficiency.In coal mining,adverse factors such as poor working environment,high-intensity workload and lagging maintenance measures make frequent failures of shearer equipment parts,which pose a potential threat to the life safety of coal mine workers.The rocker arm transmission system of shearer is the key actuator to drive the drum to rotate and cut and realize the coal dropping function.As the main support in the transmission system,the rolling bearing plays an important role in reducing the friction resistance of rotating parts and improving the transmission stability.Its healthy state directly affects the reliability of the whole shearer.Therefore,understanding the degradation mode of bearing and accurately predicting the remaining life can provide accurate guidance for the implementation of predictive maintenance strategy of shearer,which is of great engineering practical significance for the health management of shearer.In the environment of increasingly mature sensor technology and artificial intelligence technology,intelligent algorithms based on deep learning have attracted extensive attention in health management.This paper takes the rolling bearing,the key component of the rocker arm transmission system of the shearer,as the research object,and uses the data-driven research method to predict its remaining life,so as to provide technical support for the health management of coal mine machinery and equipment.The main research contents are as follows:(1)The structure of the rocker arm transmission system with high failure rate of the shearer is analyzed,and the rolling bearing,one of the key components of the rocker arm transmission system of the shearer,is taken as the main monitoring and research object;The basic composition,failure form and failure mechanism of rolling bearing are analyzed.(2)The discrete wavelet transform is used to decompose the original signal,suppress the noise component,extract the features from the decomposed signal sub components,and obtain the trend degradation features with better monotonicity.By inputting the high-dimensional feature parameter set into the variational self encoder for fusion dimensionality reduction,the effective expression of low-dimensional features can be realized,and the comprehensive degradation trend index reflecting the whole life cycle of bearing can be obtained.(3)The evolution law of bearing degradation is explored,and based on this,a good bearing degradation index is constructed to effectively improve the accuracy of life prediction.At the same time,the principle and method of stratified sampling are used for reference to realize the reasonable division of data sets.The long-term and short-term memory network with strong time series processing ability is selected as the remaining life prediction model algorithm,and the super parameters are optimized by particle swarm optimization algorithm to realize the remaining life prediction.The effectiveness of the proposed method is verified by bearing data set.(4)With mature asp Net system as the development framework,using the visual studio system development platform,SQL Server database software and deep learning algorithm collaborative programming,combining the deep learning data mining ability with the graphical visual interface,a set of system with the functions of condition monitoring,fault early warning and life prediction is developed,which enriches the predictive maintenance and health management means of the shearer.