Integrated LLM-Powered Framework for Real-Time Monitoring in Machining using Smartphone and AI Agent Orchestration

Reliable tool condition monitoring is essential for maintaining machining quality, reducing production downtime, and preventing unexpected tool failures. Conventional tool condition monitoring systems can accurately detect tool wear but typically provide limited support for automated diagnosis and maintenance decision-making. To address this limitation, this paper proposes an economical AI agent-driven maintenance framework for milling operations that integrates smartphone-based vibration sensing, one-dimensional Convolutional Neural Networks (1D-CNNs), and Large Language Model (LLM)-powered AI agents. Vibration signals acquired using a smartphone-embedded accelerometer are preprocessed and analyzed using a 1D-CNN to classify tool conditions into fresh and worn states. The classification results are then transmitted to an n8n-based workflow, where LLM-powered AI agents automatically interpret the monitoring results, generate diagnostic reports, and recommend appropriate maintenance actions. Experiments conducted on Aluminum 6061 milling demonstrate accurate binary classification of tool conditions and validate the feasibility of integrating AI agents into the maintenance workflow. The proposed framework provides a low-cost approach to combining intelligent tool condition monitoring with automated maintenance support, making it a promising solution for small and medium-sized manufacturing enterprises.