一种工业检测金属零件内部微小缺陷质量评价方法
By collecting multimodal data and combining local signal-to-noise ratio analysis with the Peephole-LSTM model, a high-confidence defect feature map is generated to guide local CT scans and total variational regularization. Convolutional neural network scoring is used to solve the problem of quality evaluation results being out of sync with actual performance caused by neglecting defect types in existing technologies, thus achieving high-precision comprehensive quality assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIANKE PRECISION COMPONENTS (KUNSHAN) CO LTD
- Filing Date
- 2025-09-02
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for evaluating the quality of metal parts ignore the fundamental differences in mechanical behavior and failure risk among different defect types, leading to a disconnect between evaluation results and actual service performance.
Multimodal data (ultrasound A-scan waveform, eddy current impedance signal and X-ray image) are acquired. Defect feature maps are generated through local signal-to-noise ratio analysis and dynamic weight allocation of the Peephole-LSTM model. Combined with local CT scan and total variational regularization, defect quantification parameters are calculated and scored using a convolutional neural network.
It enables precise location and characterization of minute internal defects in metal parts, improving the accuracy and scientific validity of test results and ensuring the safety and reliability of high-precision equipment.
Smart Images
Figure CN121190399B_ABST