ScholarWorks@경상국립대학교

초록

Water distribution networks play a crucial role in ensuring a reliable water supply, yet they encounter challenges such as corrosion, scale formation, and biofilm growth due to interactions with environmental elements. Biofilms and corrosion layers are significant contaminants in water pipes, formed by complex interactions with pipe materials. As the structure of these contamination layers varies depending on the pipe material, it is essential to investigate the contamination layer for each material individually. Specifically, biofilm growth is typically investigated concerning organic sources, while the growth of humus layers is examined in relation to inorganic elements such as manganese (Mn), iron (Fe), and aluminum (Al), which are major elements and organic substances found in water pipes. Real-time imaging of recently contaminated layers can provide important insights to improve system performance by optimizing operations and cleaning processes. In this study, cast iron (7.10 ± 0.78 nm) exhibits greater surface roughness compared to PVC (5.60 ± 0.14 nm) and provides favorable conditions for biofilm formation due to its positive charge. Over a period of 425 h, the fouling layer on cast iron and PVC surfaces gradually increased in fouling thickness, porosity, roughness, and density, reaching maximum value of 29.72 ± 3.6 μm, 11.44 ± 1.1%, 41673 ± 1025.6 pixels, and 0.80 ± 0.3 fouling layer pixel/layer pixel for cast iron, and 8.15 ± 0.4 μm, 20.64 ± 0.9%, 35916.6 ± 755.7 pixels, and 0.58 ± 0.1 fouling layer pixel/layer pixel, respectively. Within the scope of the current research, CNN model demonstrates high correlation coefficients (0.98 and 0.91) in predicting biofilm thickness for cast iron and PVC. The model also presented high accuracy in predicting porosity for both materials (over 0.91 for cast iron and 0.96 for PVC). While the model accurately predicted biofilm roughness and density for cast iron (correlation coefficients 0.98 and 0.94, respectively), it had lower accuracy for PVC (correlation coefficients 0.92 for both parameters). © 2024 Elsevier Ltd

키워드