Prediction of Drying Efficiency in Cabinet Solar Dryers for Medicinal Plants Using Artificial Neural Networks

This research presents a neural network-based predictive model to evaluate the operational performance of a cabinet-type solar dryer utilized for dehydrating Plantago major leaves under natural climatic conditions. While solar drying is known for its sustainability and energy efficiency, its performance is highly influenced by complex, nonlinear factors such as solar irradiance, internal chamber temperature, and ambient humidity. To overcome the constraints of conventional statistical modeling, a multilayer feedforward artificial neural network (ANN) was constructed using MATLAB’s Neural Network Toolbox. The architecture consisted of an input layer with three neurons, two hidden layers each containing ten neurons, and a single output neuron. Input parameters included solar radiation (W/m²), drying chamber temperature (°C), and relative humidity (%), while the output was defined as drying efficiency, calculated from weight loss and initial/final moisture content. Experimental data were gathered during the summer season under varying irradiance levels (650–900 W/m²) and ambient temperatures (30–42°C), yielding 120 samples. The dataset was normalized and partitioned into training (70%), validation (15%), and testing (15%) subsets. The network was trained using the Levenberg–Marquardt optimization algorithm to minimize the mean squared error. The final model achieved a strong predictive correlation (R² = 0.97) and low error (MSE < 0.0015), demonstrating a 22–25% improvement over traditional models. This methodology enables real-time monitoring and supports the integration of intelligent control systems in solar drying processes, enhancing both efficiency and consistency.