Research on Elevator Fault Prediction Based on Multi-scale Feature Fusion
SUN Ruixin
CHEN Liang
Abstract：The safe operation of elevators is essential to ensure the safety of production and personnel.Predicting failures of el-evators improves their safety and reliability.Aiming at the problem that there are too many factors affecting the normal operation of the elevator to accurately obtain useful information,this paper proposes an elevator fault prediction model that combines multi-scale feature fusion and Informer.Firstly,the elevator operation data are collected to construct feature vectors,which are inputted into the multi-scale feature fusion module to extract multidimensional feature information such as space and time,and then Informer is used to embed the location-encoded information into the input information,which undergoes a sparse probabilistic self-attention mecha-nism in order to reduce the computational complexity while capturing the long-term temporal dependence and obtaining the high-er-level temporal feature information.Finally,the future fault prediction results are output through the fully connected layer.The re-sults show that compared with existing models,the proposed method reduces 20%～40%in both MSE and MAE values,presenting higher prediction accuracy.It shows that the proposed method has significant advantages in elevator fault prediction and provides strong support for the safe operation of elevators.
Keywords：time series analysisfault predictionmulti-scale feature fusionInformer
Publication Date：2025-10-20
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：6( 43-48 )