基于长短期记忆神经网络的采煤机摇臂轴承剩余寿命预测

Prediction of Residual Life of Coal Mining Machine Rocker Arm Bearings Based on Long Short-Term Memory Neural Network

为解决采煤机摇臂关键零部件的失效问题,基于长短期记忆神经网络(Long Short-Term Memory,LSTM)提出了一种创新性的方法,以预测采煤机摇臂轴承的剩余寿命。基于长短期记忆神经网络理论,通过建立轴承寿命退化指标,对轴承剩余寿命进行预测,同构利用分层抽样方法对数据集进行划分;通过引入粒子群算法优化LSTM,解决LSTM算法选择最优超参数的问题,提高轴承剩余寿命预测精度。研究结果表明,基于LSTM的轴承剩余寿命预测结果与实际轴承寿命变化情况基本一致,预测结果均在置信区间内,可以为轴承维修保养工作提供参考。

To solve the failure problem of key components of coal mining machine rocker arms,an innovative method based on Long Short-Term Memory(LSTM)neural network is proposed to predict the residual life of the coal mining machine rocker arm bearing.Based on the theory of Long Short-Term Memory neural network,by establishing a degradation index for bearing life,the residual life of the bearing is conducted to predict.Isomorphism uses stratified sampling method to partition data sets;By introducing particle swarm algorithm to optimize LSTM,the problem of selecting the optimal hyperparameter in LSTM algorithm is solved,and the accuracy of predicting the residual life of the bearing is improved.The research results indicate that the residual life predicted results of the bearing based on LSTM is basically consistent with the actual changes situations in bearing life,and the predicted results are all within the confidence interval,which can provide reference for bearing maintenance and upkeep work.