Optimal PVC reactor temperature and viscosity control for automated manufacturing cells based on machine learning and internet of things

BACKGROUND This study develops an intelligent control framework for optimizing temperature and viscosity during PVC polymerization by integrating machine learning (ML), the Internet of Things (IoT), a Large Separable Quantum Portfolio Convolutional Neural Network (LQCN), and model predictive control (MPC). The proposed approach utilizes real‐time IoT sensor data from PVC reactors to achieve predictive optimization of reactor operating conditions, enhancing process efficiency and product quality. RESULTS The proposed MPC‐LQCN framework demonstrated superior performance compared with existing neural network‐based MPC approaches, achieving faster response times ranging from 0 to 1.6 s. Cross‐validation results showed excellent predictive accuracy, with Mean Squared Error (MSE) values reduced to 0.008 for temperature and 0.0009 for viscosity at Fold 10. Comparative analysis further confirmed that the MPC‐LQCN model outperformed LSTM, PINN, and ANN models in terms of MSE, MAE, MAPE, and RMSE. CONCLUSION The integration of ML, IoT, quantum‐inspired optimization, and MPC significantly improved reactor stability, energy efficiency, and product quality in PVC production. The proposed MPC‐LQCN framework provides a scalable and effective solution for intelligent industrial process control and automated manufacturing applications.