一种基于机器学习的表面裂纹缺陷识别方法及装置
By introducing a surface parameter field and array element pose table onto the surface of a large-curvature metal object, combined with multi-frequency demodulation and velocity correction, a physically consistent feature vector is generated. A machine learning model is then used for crack identification, solving the signal distortion problem caused by incomplete bonding of the array-type AC electromagnetic field probe and achieving high-precision crack identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- YUNNAN SPECIAL EQUIP SAFETY TESTING RES INST
- Filing Date
- 2025-10-14
- Publication Date
- 2026-07-17
AI Technical Summary
On the surface of a metal object with high curvature, it is difficult for an array of AC electromagnetic field probes to fit completely, resulting in uneven current distribution. This leads to atypical distortion of signals in the crack length and depth directions, affecting the accuracy of crack identification.
By introducing a surface parameter field and an array element pose table, the triaxial magnetic induction intensity vector is subjected to direction cosine projection and lift compensation. Combined with multi-frequency demodulation and velocity correlation correction, a physical consistency feature vector is generated, and a machine learning model is used to accurately identify the crack geometry parameters.
It significantly improves the accuracy of surface crack identification under complex curvature conditions, eliminates signal amplitude distortion, reduces measurement errors, enhances the stability and robustness of feature parameters, and strengthens the model's generalization ability to complex scenarios.
Smart Images
Figure CN121365279B_ABST