In the present study, a physically-based hydraulic modeling tool and a data-driven approach using artificial neural networks （ANNs） were evaluated for their ability to simulate the fate and transport of microorganis...In the present study, a physically-based hydraulic modeling tool and a data-driven approach using artificial neural networks （ANNs） were evaluated for their ability to simulate the fate and transport of microorganisms in a water system. To produce reliable data, a pipe network was constructed and a series of experiments using a fecal coliform indicator （Escherichia coli 15597） was conducted. For the physically-based model, morphological （pipe size, link length, slope, etc.） and hydraulic （flow rate） conditions were used as input variables, and for ANNs, water quality parameters （conductivity, pH, and turbidity） were used. Both approaches accurately described the fate and transport of microorganisms （physically-based model： correlation coefficient （R） in the range of 0.914 - 0.977 and ANNs： R in the range of 0.949 - 0.980）, with the exception of one case at a low flow rate （q = 31.56 cm^3/sec）. This study also indicated that these approaches could be complementarily utilized to assess the vulnerability of water facilities and to establish emergency plans based on hypothetical scenarios.展开更多
文摘In the present study, a physically-based hydraulic modeling tool and a data-driven approach using artificial neural networks （ANNs） were evaluated for their ability to simulate the fate and transport of microorganisms in a water system. To produce reliable data, a pipe network was constructed and a series of experiments using a fecal coliform indicator （Escherichia coli 15597） was conducted. For the physically-based model, morphological （pipe size, link length, slope, etc.） and hydraulic （flow rate） conditions were used as input variables, and for ANNs, water quality parameters （conductivity, pH, and turbidity） were used. Both approaches accurately described the fate and transport of microorganisms （physically-based model： correlation coefficient （R） in the range of 0.914 - 0.977 and ANNs： R in the range of 0.949 - 0.980）, with the exception of one case at a low flow rate （q = 31.56 cm^3/sec）. This study also indicated that these approaches could be complementarily utilized to assess the vulnerability of water facilities and to establish emergency plans based on hypothetical scenarios.
