| Abstract |
Simav geothermal field is located within the Aegean Graben System in western Anatolia. Rock units in the study area are mainly the formation of Menderes Massive. The Simav geothermal waters were grouped into types, namely four; Eynal, Citgol, and Naşa geothermal water and cold water. With this study, it is aimed to introduce a method for classifying waters in the study area using some parameters such as temperature, pH, electrical conductivity and major ions by means of Artificial Neural Network (ANN) method. According to the data obtained from wells drilled for the drinking and irrigation water purposes, ground water flow is toward the desiccated lake. Cold water analysis gave high CO3+HCO3, Ca, Mg ion values, and low NH4, NO3, Fe, NO2, Al and Mn ion values. Hot water analysis gave a cation trend of Na+K>Ca>Mg and an anion trend of HCO3+CO3>SO4>Cl. While preparing the training data set in ANN method, for input, T (C), EC (S), pH, Na (mg/l), K (mg/l), Ca (mg/l), Mg (mg/l), CO3 (mg/l), HCO3 (mg/l), Cl (mg/l) and SO4 (mg/l) values of 50 water samples from the study area were used. Four output values were used. In each output value, the known water represented by 1 and others by 0. A test data set of 15 samples in which the T, EC, pH, Na, K, Ca, Mg, CO3, HCO3, Cl and SO4 values are known but their group are unknown was prepared. And these input values were run in ANN model in order to see how the waters were grouped. The advantages of artificial neural networks can be exploited to solve this problem. The most common ANN architecture is Multilayered Perceptrons, which was used in this study. For this solution, the first artificial neural network model using Extended Delta-Bar-Delta (EDBD) algorithm has been successfully implemented. Mean Square Error result of these model obtained by EDBD algorithm is 1.3x10–3. These results show that the group in which the waters in the study area fall can be determined with high accuracy by using some parameters of water the ion content of water. |