An intelligent defect detection method for automatic fiber placement of composite materials
By improving the feature extraction and fusion module of the YOLO12 model and combining it with the FM-IoU loss function, the problem of insufficient detection accuracy in the automatic fiber placement process of composite materials was solved, and efficient detection and accurate bounding box regression of defects of different scales were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QUANZHOU INST OF EQUIP MFG
- Filing Date
- 2026-04-20
- Publication Date
- 2026-07-21
AI Technical Summary
Existing methods based on general target detection networks have problems in the automatic fiber placement process of composite materials, such as insufficient adaptation of backbone features to frequency information, inadequate characterization of the contribution differences of multi-scale feature fusion at different levels, and insensitivity of bounding box regression to boundary extrema and the localization of long and thin targets, resulting in insufficient detection accuracy.
An improved YOLO12 model is adopted, replacing the C3k2 module with a multi-scale feature fusion convolution module, replacing the A2C2f module with a region attention function module, and adding a multi-scale feature weighted fusion module and an attention secondary fusion module. Combined with the FM-IoU loss function, the feature extraction and fusion process is optimized.
It improves the accuracy and engineering applicability of automatic fiber placement defect detection in composite materials, enhances the ability to extract defect features at different scales, improves the perception of weakly significant defects, and achieves accurate regression of bounding boxes.
Smart Images

Figure CN122090039B_ABST