Autonomously Detecting Defects in Circular Production Using Multimodal Large Language Models and Retrieval Augmentation
Abstract
Automated defect detection is a key enabler for scalable inspection in circular production, where varying component conditions and the scarcity of large-scale annotated datasets challenge conventional supervised vision models. This paper investigates whether multimodal large language models (MLLMs) can serve as a flexible alternative for industrial defect detection without task specific training. To bridge the domain gap of general-purpose models, several retrieval-augmented generation (RAG) architectures are designed to leverage ex pert-defined examples for visual defect detection on used gears in a circular pro duction context. The study compares baseline prompting with text-based, image based, and self-reflective retrieval strategies that dynamically provide context relevant examples and descriptions during inference. The results show that MLLMs already achieve good defect detection accuracy out-of-the-box. Integrat ing RAG increases defect detection accuracy to 95.83% and further improves fine-grained defect classification, with gains of up to 63.94% for specific rare defect types. While the models remain weaker in precise localization and still lack deeper object-level understanding, the results indicate that RAG-enhanced MLLMs are a viable, low-barrier solution for inspection tasks in remanufacturing scenarios with limited expert supervision and diverse object classes. These find ings establish a reproducible benchmark for RAG-augmented VLMs in defect classification…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.