| Summary: | This research investigates the application of artificial neural networks (ANNs) to enhance the efficiency and predictability of metal oxide photocatalytic systems for dye removal in wastewater treatment. We implemented various machine learning algorithms, including Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG), and explored different network architectures and activation functions. Our findings show that the optimized ANN architecture (4-10-1) with a learning algorithm Levenberg-Marquardt (LM) achieved a high prediction accuracy with an R2 value of 0.9133 and a mean squared error (MSE) of 14.6659. Using MATLAB for model development and analysis, we trained the ANN with around 100 input data sets, divided into training, validation, and testing processes. The research findings significantly contribute to advancing the accuracy and reliability of predictions in metal oxide photocatalytic systems, facilitating optimized dye removal processes. This study offers valuable insights into optimizing ANN architectures, demonstrating their superiority over traditional methods. It lays the groundwork for the practical application of ANN-based models in improving the effectiveness of metal oxide photocatalytic systems, offering a promising avenue for sustainable and efficient environmental remediation. The limitation of this study is that lab-scale findings may not directly apply to industrial scales, requiring adjustments to the ANN model that could impact its predictive accuracy. © 2024 IEEE.
|
|---|