A Knowledge-Driven LLM-Based Decision-Support System for Explainable Defect Analysis and Mitigation Guidance in Laser Powder Bed Fusion
Ontology-guided AI analyzes 27 defect types with explainable reasoning.
A knowledge-driven LLM-based decision-support system integrates structured defect knowledge with LLM reasoning for explainable defect diagnosis and mitigation guidance in laser powder bed fusion (LPBF). Built on a knowledge base of 27 LPBF defect types, the system supports fuzzy natural language queries, literature-supported explanations, and multimodal image assessment via foundation models. The fully integrated configuration achieved a macro-average F1 score of 0.808, with inter-rater reliability analysis showing substantial agreement with literature-derived reference labels, enabling consistent, interpretable defect analysis.
- System uses 27 hierarchical LPBF defect types with causal relationships for explainable diagnosis
- Supports fuzzy natural language queries and multimodal image assessment via semantic alignment scoring
- Fully integrated configuration achieved 80.8% macro F1 score with substantial inter-rater reliability
Why It Matters
Brings explainable, ontology-guided AI to additive manufacturing quality control, improving defect detection consistency and safety.