AI-Driven Predictive Maintenance in Smart Manufacturing: Enhancing Efficiency through Deep Learning Models

Tool wear monitoring and predictive maintenance are critical in manufacturing, where traditional methods often struggle to adapt to changing conditions. This research presents an Adaptive Reinforcement Learning Framework for Real-Time Tool Wear Optimization and Predictive Maintenance (ARTOM). The framework integrates reinforcement learning with real-time sensor feedback to optimize machining parameters and maintenance schedules dynamically. Proximal Policy Optimization (PPO) is used to guide decision-making by balancing tool life, product quality, and operational costs. Multi-agent reinforcement learning divides tasks among agents to handle diverse machining scenarios, while sliding window techniques and dimensionality reduction ensure efficient data processing. The study has used the benchmark dataset, which include time-series sensor data and machining parameters. Metrics potential metrics have been used to evaluate prediction accuracy, while runtime and memory usage assess computational efficiency. Results has shown that ARTOM consistently achieves lower prediction errors and faster execution times than contemporary baseline models. These findings demonstrate ARTOM’s ability to adapt to different tool conditions and improve operational decision-making.