Predicting the residual aluminum in traditional drinking water purification plants using artificial intelligence techniques

Abstract

Alum is widely used in water purification as a chemical coagulant.  A high dose of it causes at least a slight health risk, and some evidence suggests that aluminum can increase the risk of developing Alzheimer's disease.  Hence, it is important to reduce the amount of residual aluminum in drinking water.

In this study, the relation between residual aluminum and various parameters related to the treatment process was studied, by using a hybrid neural network models.

The actual data of the water purification plant in Al Qusayr in Homs over a period of three years was used, and the residual aluminum in drinking water was analyzed by artificial-genetic neural network (GA-ANN) models.

To create simple and reliable prediction models that can be used in the Early Warning System (EWS).

Models were constructed using the data of raw water: turbidity of the raw water, pH, temperature and coagulant dose in order to predict the residual aluminum values ​​for the water leaving the plant.  And the effect of preprocessing data was studied by comparing results with previous studies.  

Several models were built, and a network with a higher fitness was selected.

The results showed that it is possible to predict the residual aluminum values ​​with a high accuracy using the structured (4-7-1) with (LM) algorithm.  The variables with the greatest influence on the residual amount of aluminum were turbidity of raw water, coagulant dose, respectively.

The results of the selected network were able to predict the aluminum of the output with high accuracy with a correlation coefficient (R) (0. 93) and absolute errors (AE) 0. 024.