An Optimized ANFIS Framework for Intelligent Prediction and Control of Agricultural Product Drying Processes in Indirect Solar Dryers

Agricultural product drying in indirect solar dryers is characterized by strong nonlinearities, time-varying environmental disturbances, and complex interactions between temperature, humidity, and airflow, making accurate prediction and intelligent control challenging using conventional approaches. This study presents an optimized Adaptive Neuro-Fuzzy Inference System (ANFIS) framework for modeling and intelligent control of the drying process in an indirect natural convection solar dryer. A dataset consisting of 200 experimentally collected samples was obtained from a laboratory-scale solar drying system under varying operating conditions. Initially, a Neural Network Predictive Controller (NNPC) was developed using the Levenberg–Marquardt (LM, trainlm) algorithm to identify the dynamic behavior of the drying process. Subsequently, ANFIS models were synthesized using three different training strategies, namely Grid Partition with Backpropagation, Grid Partition with Hybrid learning, and Subtractive Clustering combined with Hybrid learning. The models were trained and evaluated at different epoch numbers (150, 250, and 350 epochs) to investigate their convergence characteristics and prediction accuracy. The results demonstrated that the Subtractive Clustering–Hybrid ANFIS approach significantly improved model performance by automatically generating an efficient rule base and optimizing membership function parameters. Progressive reductions in prediction errors and improvements in generalization capability were observed throughout the training process. The optimized model achieved a predictive performance of RMSE = 0.002, MAE = 0.0091, and R² = 0.9268, indicating excellent agreement between experimental and predicted values. The findings demonstrate that integrating neural predictive control with optimized ANFIS architectures provides a robust soft-computing framework capable of accurately capturing the nonlinear dynamics of agricultural drying processes. The proposed intelligent model offers a reliable foundation for real-time monitoring, adaptive control, and energy-efficient operation of indirect solar drying systems, thereby contributing to improved product quality and sustainable agricultural processing technologies.