Intelligent Autonomous Inspection Framework for Chemical Process Plants

Chemical process plants present complex safety challenges requiring comprehensive inspection protocols while minimizing human exposure to hazardous environments. This paper introduces a novel intelligent autonomous inspection framework (IAIF) that synergistically integrates deep learningenabled hazard detection with risk-aware autonomous navigation for chemical plant safety management. The proposed framework combines a YOLOv3-based multi-modal object detection system with QR code-guided autonomous navigation, enhanced by real-time HIRARC (Hazard Identification, Risk Assessment, and Risk Control) risk evaluation. Our approach represents the first comprehensive integration of autonomous navigation, real-time hazard detection, and formal risk assessment methodologies in a unified AI-powered inspection system. Experimental validation in a three-level Hydro Treating Unit (HTU) pilot plant demonstrates exceptional performance with 1.0 mean average precision (mAP) for stationary hazard detection, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 8 8}$</tex> precision for dynamic navigation scenarios, and successful autonomous completion of complex multi-level inspection routes. The framework achieved 100 % route completion success rate while maintaining sub- 40 ms real-time processing capabilities. This work establishes a new paradigm for AI-driven industrial safety systems, providing a scalable, intelligent solution that significantly reduces human risk exposure while enhancing inspection comprehensiveness and consistency