Real-time internal state estimation in composite photocuring via coupled heat-transfer modeling and Bayesian-optimized adaptive extended Kalman filtering

Abstract

Ultraviolet (UV) photocuring of composite materials involves strongly coupled optical, thermal, and chemical processes, making transient internal temperature and degree of cure difficult to measure directly during processing. This study develops an online internal-state estimation framework using surface thermal observations. A process-specific nonlinear state-space model is formulated by coupling UV irradiance attenuation, curing kinetics, reaction heat generation, and three-dimensional heat conduction. The measured incident irradiance is used as the model input, while the ROI-averaged surface temperature measured by infrared thermography serves as the sole online observation. To reduce dependence on empirically selected noise covariances, a two-stage Bayesian optimization-based adaptive extended Kalman filter (BO-AEKF) is developed, combining offline Bayesian optimization of the initial process and measurement noise covariances with residual-based adaptive updating during online estimation. The framework recursively estimates internal temperature and degree-of-cure fields. Under a representative irradiance condition, BO-AEKF achieved a surface-temperature RMSE of 1.9909 °C, reducing RMSE by 44.1% and 45.3% relative to EKF and AEKF, respectively. Independent validation against embedded temperature sensors yielded an average internal-temperature RMSE of 4.5684 °C, while DSC-derived degree-of-cure values yielded an RMSE of 0.0478. Additional experiments at different irradiance levels confirmed that the estimator retained surface-temperature tracking capability across the tested conditions. These results demonstrate the feasibility of process-specific online estimation of internal thermal and curing states during composite photocuring.