Abstract:
In modern industries and manufacturing techniques, quick and precise maintenance is a vital concept. This study introduces an AI-driven Remote Expert System that transforms industrial maintenance through the integration of artificial intelligence (AI), mixed reality (MR), and natural language processing (NLP) technologies. Our implementation utilizes YOLOv8s on Microsoft HoloLens 2 for real-time object recognition and LED status indicator detection on PLCnext controllers, enabling on-site operators to receive immediate diagnostic information. Experimental
validation and field trials demonstrate robust system performance with an mAP50 of 0.993 and mAP50-95 of 0.844, while optimizing CPU utilization and battery consumption during video streaming operations. The system automatically generates NLP-based maintenance recommendations corresponding to identified issues. This innovative approach establishes a new paradigm for industrial maintenance that significantly reduces equipment downtime, enhances
operational efficiency, and minimizes maintenance costs.