基于自适应时间窗的设备剩余寿命实时预测研究

Research on the Real-Time Prediction of Remaining Useful Life of Equipment Based on Adaptive Time Window

为适应智能工厂设备剩余寿命预测和维护决策的实时性与准确性要求,针对设备的多样化健康状态与独立退化特性,综合考虑设备在不同作业环境中的个体差异与同类设备在重要指标上的共同依赖,设计了智能工厂MES中基于数据驱动的剩余寿命预测流程,该流程旨在实现独立退化特性设备的实时性剩余寿命预测;随后结合广义回归神经网络,提出一种基于实时状态的剩余寿命预测方法,该方法不仅采用自适应时间窗,提高了预测的精度,还进一步采用动态步长策略与相空间重构技术,降低了时序特征波动与训练样本较少带来的误差风险;最后利用轴承全生命周期数据,运用仿真验证了该方法的有效性。

In order to meet the real-time and accuracy requirements for the remaining useful life(RUL)prediction and mainte-nance decisions of equipment in intelligent factories,theremaining useful life prediction process based on data drive in manufacturing execution system(MES)of intelligent factory is designed after comprehensively considering the equipment’s ind-ividual differences in different operating environments and the common dependence of similar equipment on important indicators in terms of equipment’s diversified health status and independent degradation characteristics.Afterwards,combined with generalized regression neural network(GRNN),a method for the prediction of remaining useful life based on real-time st-atus is proposed.This method not only uses adaptive time window to improve the accuracy of prediction,but also uses the dynamic step strategy and phase space reconstruction to reduce the risk of deviation caused by the fluctuation of timing characteristics and a smallamount of training samples.Finally,the effectiveness of the method is verified by simulation with the life cycle data of the bearing.