Fault diagnosis method based on time-decay position encoding, electronic device
By employing temporal decay position encoding and a multi-scale time-aware attention mechanism, the problem of deep learning models being unable to distinguish between short-term and long-term faults in elevator fault diagnosis is solved, achieving higher fault detection accuracy and early fault warning.
CN122241444APending Publication Date: 2026-06-19CHONGQING SPECIAL EQUIP TESTING & RES INST (CHONGQING SPECIAL EQUIP ACCIDENT EMERGENCY INVESTIGATION & PROCESSING CENT)
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHONGQING SPECIAL EQUIP TESTING & RES INST (CHONGQING SPECIAL EQUIP ACCIDENT EMERGENCY INVESTIGATION & PROCESSING CENT)
- Filing Date
- 2026-05-22
- Publication Date
- 2026-06-19
AI Technical Summary
Technical Problem
Existing deep learning models cannot adaptively capture the temporal characteristics of elevator faults in elevator fault diagnosis, and it is difficult to distinguish between short-term and long-term faults.
Method used
A temporal decay position coding method is adopted. By calculating the adaptive decay coefficient and multi-scale time-aware attention mechanism, combined with a temporal Transformer network, short-term and long-term fault features in elevator operation data are extracted.
Benefits of technology
It improves the accuracy of elevator fault differentiation and detection, can adapt to different working conditions, effectively captures short-term and long-term fault characteristics, and provides early fault diagnosis and maintenance suggestions.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122241444A_ABST
Abstract
This invention relates to the field of fault detection technology, providing a fault diagnosis method and electronic device based on time-series decay location coding. The method includes: acquiring time-series data during elevator operation; preprocessing the time-series data based on a preset strategy to obtain a time-series tensor; calculating an adaptive decay coefficient based on the time-series tensor using a time-series decay model, and obtaining a time-series decay location code based on the adaptive decay coefficient; inputting the time-series tensor and the time-series decay location code into a deep learning model, whereby the deep learning model processes the time-series tensor and the time-series decay location code into a fused time-series feature, and obtaining a fault judgment result based on the fused time-series feature. This invention can effectively capture both short-term and long-term fault features, thereby improving the accuracy of elevator fault differentiation and detection.
Need to check novelty before this filing date? Find Prior Art