Application Of Artificial Neural Network Models For Predicting Total Dissolved Solids In Marsh Water

Abstract

In this paper an Artificial Neural Networks (ANNS) model is designed to predict the
Total Dissolved Solids (TDS) concentration in marsh water. A previous data set are selected
from previous studies which done on analysis of marsh water quality, these data are arranged
in a format of five input parameters to feed forword back-propagation including the acidity
(pH), calcium concentration (C), Magnesium Concentration (M) , Chloride Concentration
(Cl) and Sulphate Concentration (S), and one output parameter as Total Dissolved Solids
concentration. Artificial Neural Network used to study the effect of each parameter on TDS
concentration in marsh water. Several structures of ANNs model is examined with different
transfer functions, activation functions, number of neurons in each hidden layer and number
of hidden layers. Results show that the two hidden layer network with transfer function
(trainscg) with (12 & 10) neurons in the first and second hidden layer respectively and
(tansig-tansig-purelin) gives the best performance (Mean Square Error: 3.05e-5) network for
this prediction.