AWWA MTC57563

An artificial neural network model is derived and validated for predicting contaminant removal during nanofiltration of ground and surface waters under conditions typical of drinking water treatment. The network was trained using operating conditions such as permeate flux, feed water recovery, and element recovery (crossflow velocity), and feed water quality parameters including pH, total dissolved solids concentration (surrogate for ionic strength), target contaminant concentration, and where possible the diffusion coefficient as inputs to predict the permeate concentration. Deterministic and pseudo stochastic simulations showed that artificial neural networks closely predicted permeate concentrations of several organic and inorganic contaminants in experiments using source waters from seven different locations by two commercial thin film composite membranes operating in a wide range of permeate fluxes and feed water recoveries. Hence, neural networks can predict transport of heterogeneous water treatment contaminants such as natural organic matter and disinfection byproduct precursors, whose physicochemical properties are unknown. Includes 36 references, figure.