AIoT-Driven Closed-Loop Thermal Regulation for Photopolymerization: A CNN–TCN Edge-Computing Framework

<title>Abstract</title> <p> Effective thermal management in photopolymerization processes is essential for ensuring material integrity, product quality, and modern nail-gel curing applications. Conventional UV-LED curing systems operate in an open-loop mode, lacking thermal feedback, resulting in uncontrolled temperature fluctuations and inconsistent curing outcomes. This study introduces a Nail Curing Intelligent Testing System (NCITS) that integrates a 1D CNN and a TCN system for real-time thermal prediction and regulation. Although validated through gel nail curing experiments, the NCITS is designed as a representative demonstrator for controlling exothermic photopolymerization in laboratory-scale prototype contexts, including additive coatings and composite curing. The system interprets thermal dynamics via learned derivative features, enabling predictive intervention before thresholds are exceeded. Experimental results demonstrate peak temperature reductions of up to 35% while maintaining curing efficacy/quality. Compared with PID control, NCITS reduces overshoot by 39.7%, improves settling time by 12.5%, and lowers <italic>RMSE</italic> by 46.6%. Deployment on edge devices with 8-bit quantization achieves 4.2× model compression and inference latencies of 23.4–45.6 ms, confirming its suitability for real-time environments. This research demonstrates potential applicability in small-scale intelligent thermal regulation systems, providing a scalable pathway for the cosmetics industry. </p>