基于多层注意力融合的Transformer故障诊断解释方法及系统
By using a Transformer model with multi-layer attention fusion, the problem of incomplete interpretation of fault diagnosis models in high-reliability scenarios in existing technologies is solved, and a clear revelation of fault characteristics and decision-making logic is achieved, thereby improving the interpretability and diagnostic accuracy of the model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2026-05-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing Transformer-based fault diagnosis models fail to fully reveal the intrinsic relationship between fault characteristics and decision-making logic in scenarios with high reliability requirements. Furthermore, the multi-head attention weight fusion strategy obscures the attention of key heads, resulting in an incomplete explanation of fault decisions.
A multi-layer attention fusion approach is adopted, which combines multi-head importance adaptive attention blocks and multi-level attention flow fusion modules with convolutional embedding layers and posterior interpretability modules to achieve weighted fusion and gradient integration of attention weights, thereby revealing the fault diagnosis decision-making process.
It improves the interpretability and user trust of the model, ensures the focus of key features and the complete delivery of information, and is suitable for scenarios with high reliability requirements such as aircraft engines and nuclear power plants.
Smart Images
Figure CN122196703B_ABST