Smart manufacturing 4.0: Integration of artificial intelligence for real-time process control and productivity enhancement

Abstract

Abstract. The proliferation of interlinked industrial devices at an exponential rate in the Industry 4.0 paradigm has produced volumes of real-time manufacturing information not seen before, creating a need and opportunity to establish intelligent process control. This article suggests a new hybrid artificial intelligence system of smart manufacturing, combining Long Short-Term Memory (LSTM) networks, Convolutional Neural Networks (CNN), ensemble tree-based classifiers, and a Proximal Policy Optimization (PPO)-based Reinforcement Learning (RL) agent in a four-layer system that includes data acquisition, AI processing, decision control, and feedback actuation. The framework is tested on the SECOM semiconductor manufacturing data with added synthetic CNC machining data, where a fault detection accuracy of 96.7, an Overall Equipment Effectiveness (OEE) of 91.8 and a defect rate is reduced by 6.8 to 1.1 compared to traditional Statistical Process Control (SPC) baselines. Latency of inference 44 ms meets hard real-time requirements in manufacturing. The superiority and generalizability of the proposed approach are supported by the results of comparative analysis against six state-of-the-art methods. The findings indicate that predictive modeling, computer vision and adaptive closed-loop control in a synergistic combination is a viable scalable route to intelligent manufacturing excellence.