Mohsen Niazi; Ali Naghizadeh; Mansour Baziar
Abstract
AbstractBackground and purposeThe turbidity of treated water is measured as an important parameter in determining the quality of drinking or industrial water in all treatment plants. Due to the importance of the prevalence of pathogens such as Giardia and Cryptosporidium, which cause dangerous diseases ...
Read More
AbstractBackground and purposeThe turbidity of treated water is measured as an important parameter in determining the quality of drinking or industrial water in all treatment plants. Due to the importance of the prevalence of pathogens such as Giardia and Cryptosporidium, which cause dangerous diseases such as dysentery, the relationship between reducing turbidity and increasing the elimination of these microorganisms has been proven in studies.Materials and methodsIn this study, an artificial neural network (ANN) model and multiple linear regression(MLR) were developed and their performance was compared to predict the turbidity of treated water of Tabas water treatment plant. Total dissolved solids, pH, temperature and input turbidity of raw water were used as input parameters of the models in the predictions. The best backpropagation algorithm and number of neurons were determined to optimize the model architecture.ResultsThe results showed that the Levenberg–Marquardt algorithm was selected as the best algorithm and the number of optimal neurons was determined to be 16.Also, the results of the sensitivity analysis of the neural network model showed that the input turbidity with a value of 29% is the most important parameter in the development of the ANN model.ConclusionThe results of correlation coefficient of MLR and ANN models were obtained for training data 0.63 and 0.8921 and for testing data 0.60 and 0.8571, respectively, which show the superiority of ANN model in Predicting the turbidity of the output of Tabas water treatment plant.