Robot fault recognition method and system, model training method, device and medium
By constructing physical state diagrams and control state diagrams of the robot during operation, and combining them with the cross-domain connection matrix of the deep learning model, the problem of poor fault analysis in multi-source information fusion is solved, and the accuracy of robot fault identification is improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHONGQING JINKANG NEW ENERGY VEHICLE CO LTD
- Filing Date
- 2026-05-22
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies for robot fault identification, the fault analysis results are poor due to the feature splicing method when fusing multi-source information, and it is difficult to effectively learn cross-domain coupled features, which affects the accuracy of fault identification.
By synchronously collecting physical state sensor data and control state sensor data during robot operation, physical state diagrams and control state diagrams are constructed. Based on a deep learning model, a cross-domain connection matrix is built to explicitly model the coupling relationship between the control command domain and the physical execution domain, and feature aggregation and fault probability determination are performed.
It improves the accuracy of robot fault identification, avoids damage to the physical internal structure and the introduction of noise and redundancy, and can explicitly learn cross-domain coupling features, thus improving the accuracy of fault identification.
Smart Images
Figure CN122401366A_ABST