Forecasting the productivity of a solar distiller enhanced with an inclined absorber plate using stochastic gradient descent in artificial neural networks

• Published in: Multiscale and Multidisciplinary Modeling, Experiments and Design Vol. 7; no. 3; pp. 1819 - 1829
• Main Authors: Mohammed, Suha A., Al-Haddad, Luttfi A., Alawee, Wissam H., Dhahad, Hayder A., Jaber, Alaa Abdulhady, Al-Haddad, Sinan A.
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.07.2024
• Subjects: Characterization and Evaluation of Materials; Engineering; Mathematical Applications in the Physical Sciences; Mechanical Engineering; Numerical and Computational Physics; Original Paper; Simulation; Solid Mechanics; Solar distillers; Neural network; Stochastic gradient descent; Prediction; Machine learning
• ISSN: 2520-8160; 2520-8179
• DOI: 10.1007/s41939-023-00309-y

Solar distillers are of significant importance in advanced technologies, as they provide a sustainable approach to address the issue of water purification. The utilization of innovative methodologies holds considerable significance in improving solar distiller performance. For that, this study implemented the incorporation of an absorber plate to enhance the operational efficiency of a Conventional Solar Distiller (CSD) in addition to a novel Machine Learning (ML) approach. Two solar distillers were designed, manufactured, and subjected to experimental evaluation to acquire data for 10 h. Hourly measurements were recorded for the water temperature within the distiller, the temperature of the glass covering, and the ambient temperature, in addition to productivity values. By implementing the absorber plate, the productivity was notably enhanced, resulting in a 138.68% increase from 1311.3 to 3129.8 ml/m 2 .h in the Modified Solar Distiller (MSD). In order to forecast productivity values, a two-hidden-layer Neural Network (NN) model was utilized. The model underwent further novel improvement through the implementation of Stochastic Gradient Descent (SGD). Remarkably, the SGD-enhanced neural network (NN-SGD) model exhibited outstanding performance in predicting productivity for conventional and modified solar distillers, yielding a determination coefficient of 1.000 and a variation coefficient of 0.00822. Applying machine learning to forecast solar distiller outputs demonstrates the promising potential for future research endeavors.