基于PSO优化LSTM的滚动轴承剩余寿命预测

Predicting the Remaining Service Life of Rolling Bearings Based on PSO-LSTM Network Model

为了提高滚动轴承退化特征关于时间序列的预测精度,使预测模型更加适用于滚动轴承的运行退化数据,采用粒子群算法对长短期记忆网络的参数进行优化,构建PSO-LSTM滚动轴承寿命预测模型,根据模型拟合出轴承的剩余寿命曲线。经过实验发现,PSO-LSTM网络模型可以较好地拟合复杂工况下轴承的寿命退化趋势,且与其他模型相比拟合效果更好,预测结果更为准确。

For the sake of improving the degree of accuracy about time prediction of rolling bearing degradation characteristics,and make the prediction model more suitable for rolling bearing operational degradation data,the particle swarm optimization(PSO)was used to optimize parameters of LSTM so as to build PSO-LSTM rolling bearing’s service life prediction model and then to have it based to fit the residual life curve of the rolling bearings.Experimental results showed that,the PSO-LSTM network model can better fit life degradation trend of bearings under complex operating mode,and it outperforms other models in both matching result and the prediction accuracy.