Bridging the Knowledge Gap in NDE 4.0: AI-Augmented Insights Through Retrieval-Augmented Generation (RAG)

Abstract

The rapid digitization of nondestructive evaluation (NDE) has led to fragmented, siloed knowledge across formats (manuals, reports, standards) and a generation gap among practitioners. In this work we present an AI-augmented knowledge assistant for NDE 4.0 that uses Retrieval-Augmented Generation (RAG) to provide real-time technical guidance. Our system ingests diverse NDE documents, chunks and semantically indexes their content (using MiniLM-L6-v2 and FAISS), and at query time retrieves relevant context before querying a language model (Google Gemini). We describe the system architecture and implementation, report metrics (high retrieval precision, sub-second latency, minimal API cost), and demonstrate a sample NDT use case (e.g. explaining ultrasonic testing and magnetic particle inspection steps). This AI-RAG assistant can help inspectors access “know-how and know-why” on demand, enabling continuous learning and upskilling in line with NDE 4.0. Future work includes multilingual support, voice interfaces, and LMS integration.