modeling of sorkheh reverse osmosis (ro) water treatment plant by artificial neural network (ann) with genetic algorithm (ga)

The reverse osmosis (ro) process is one of the most widely used technologies in the water treatment industry to reduce water hardness. in this paper, the reverse osmosis treatment plant of sorkheh city (semnan province, iran) was modeled by artificial neural network (ann) model. the ann model parameters were optimized with the genetic algorithm (ga) method to increase ann model accuracy. the optimization was done by updating the weight and bias, the number of layer neurons, activator functions, and the ann training equation. the mean relative error of optimized ann model results with respect to industrial data was obtained about 0.73% for water outlet flow rate and 0.47% for water ph. while the mean relative errors for water outlet flow rate and water ph in the non-optimized ann model were evaluated 26.24% and 4.76%, respectively. also, the results showed that the regression coefficient for the optimized neural network is equal to 0.995.