The EnviroPiNet (Environmental Buckingham Pi Neural Network), developed by researchers from Glasgow University’s James Watt School of Engineering, utilizes advanced machine learning and physical modeling to predict biofilter performance in wastewater treatment, achieving up to 90% accuracy. Biofilters remove organic carbon compounds, commonly found in contaminants from human waste, agriculture, and industry, before water is returned as drinking water. As biofilters degrade, monitoring their efficacy is crucial for maintaining water quality. The innovation addresses limited data challenges in wastewater treatment by combining the Buckingham Pi theorem with machine learning. After training on an extensive dataset, EnviroPiNet outperformed traditional models, exceeding the accuracy of PCA-based methods and autoencoders. The authors emphasize the importance of enhancing data strategies for better accuracy in environmental predictions. Currently, they are collaborating with industry partners to apply EnviroPiNet in real-world conditions, potentially expanding its use to healthcare settings, all while being available for free online.
