结合注意力机制与MSCNN-BiLSTM模型的电梯主动式故障预警

Elevator Active Fault Early Warning Based on Attention Mechanism and MSCNN-BiLSTM Model

主动式故障预警不仅可以辅助电梯的检修与维护,还可以最大程度地降低电梯安全事故。基于某公司现有电梯运行状态的数据集,提出一种电梯健康指数(HI)预测模型。该模型是结合注意力机制的多尺度卷积神经网络(MSCNN)和双向长短时记忆网络(BiLSTM)的融合模型,可以全面提取电梯数据集的深层次特征和时序信息,实现HI预测和主动式故障预警。在与其他常见模型方法的比较中,证实了该文模型具有更好的预测性能。

Active fault early warning can not only assist in elevator's maintenance and repair,but also minimize elevator safety accidents.A predictive model for elevator health index(HI)is proposed based on the existing dataset of elevator operation status in a certain company.This model is a fusion model of Multi-Scale Convolutional neural network(MSCNN)and bidirectional long short memory network(BiLSTM)combining attention mechanism,which can comprehensively extract the deep features and time series information of elevator dataset,and realize HI prediction and active fault early warning.In comparison with other common model methods,it has been confirmed that the model proposed in this paper has better predictive performance.