Automated Evaluation Algorithms for Online Defect Detection in Carbon Fibre Tape Laying Processes: A Review

The aerospace industry utilises highly automated tape-laying methods for composite part production, reducing time and costs. Despite the automation, manual quality inspection of carbon fibre tape layups, which is still standard in the industry, consumes 50% of production time. This paper presents an overview of automated inspection systems, focusing on image evaluation software based on expert systems and machine learning methodologies. Expert systems offer explainability, reproducibility, and speed but are limited by specific use cases and potential biases. Machine learning algorithms can map complex image relationships but face challenges in explainability and data requirements. A quantitative comparison of these systems is challenging due to differing use cases and datasets. To address this, we propose a benchmark dataset based on a geometric model incorporating features relevant to various laying processes. This model and dataset should be openly available for future research. We advocate for this standard to enable fair benchmarking and informed decision-making in selecting suitable approaches for manufacturing applications.