Carbon fiber reinforced polymers (CFRP) have significant applications in aerospace and automotive manufacturing. However, due to the complexity of CFRP structures, manufacturing defects are challenging to avoid and even affect the mechanical properties. Timely detection and repair are essential to ensure product quality. In this study, we propose a robotized framework for real-time detection and in-situ repair of manufacturing defects in CFRP patch placement. First, the influence of three typical defects (delamination, wrinkle and impurity) on mechanical properties is analyzed through numerical analysis. Then, a defect detection model is improved using the channel attention mechanism and decoupling head module, which enhances detection accuracy and the ability to identify small and deep defects. Based on the identification result, a corresponding repair strategy is generated, which considers the effects of force, path, heating and repair modes. The experimental results demonstrate that the tensile stiffness and bending strength of the repaired material are improved by 12.34% and 230.92%, respectively.

中文翻译：

---

用于实时检测和原位修复 CFRP 补片放置中制造缺陷的机器人框架


碳纤维增强聚合物（CFRP）在航空航天和汽车制造领域有着重要的应用。然而，由于CFRP结构的复杂性，制造缺陷很难避免，甚至影响其力学性能。及时检测和修复对于保证产品质量至关重要。在这项研究中，我们提出了一种机器人框架，用于实时检测和原位修复 CFRP 补片放置中的制造缺陷。首先，通过数值分析分析了三种典型缺陷（分层、起皱和杂质）对力学性能的影响。然后，利用通道注意力机制和解耦头模块改进了缺陷检测模型，提高了检测精度以及识别小而深的缺陷的能力。根据识别结果，生成相应的修复策略，该策略考虑了力、路径、加热和修复模式的影响。实验结果表明，修复后材料的拉伸刚度和弯曲强度分别提高了12.34%和230.92%。