Development of a hybrid integrated moisture content measurement method based on near-infrared (NIR) spectroscopy

Abstract

To measure the moisture content of the product in the fluidized bed drying (FBD) process in real time, a hybrid integrated soft measurement method based on near-infrared (NIR) spectroscopy is proposed in this paper. The training set for the proposed method is developed using historically acquired spectral data paired with their corresponding moisture content measurements. This dataset is subsequently integrated with several widely used machine learning models, including Partial Least Squares (PLS), Support Vector Regression (SVR), and Decision Trees (DT), and further optimized through the Particle Swarm Optimization (PSO) algorithm. By leveraging diverse data perspectives, the method constructs an optimal hybrid integrated model, ensuring robustness and enhanced predictive performance. To determine the optimal parameters for each base learning, a 5-fold cross-validation approach was employed. Spectral knowledge validation showed that the optimized hybrid integration weights by PSO conformed to the characteristic peaks of moisture, demonstrating that the hybrid integration soft measurement model could reflect the physical knowledge. To validate the effectiveness of the proposed method, three sets of comparative experiments were conducted: (1) predictions using standalone PLS, SVR, and DT models; (2) predictions employing Bagging-integrated PLS, SVR, and DT models; and (3) predictions utilizing the hybrid integrated soft sensing method proposed in this study. The experimental results demonstrate that the proposed hybrid integrated soft sensing method achieves superior prediction accuracy compared to the other approaches, underscoring its efficacy for real-time moisture content measurement in the fluidized bed drying process.