ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM FOR ESTIMATION OF WATER QUALITY INDEX IN DUHOK CAMPS

Abstract

The most appropriate method of communicating water quality situation of water bodies is the Water Quality Index (WQI); while user participation and dealing with uncertainty are required for the evaluation of WQI. The aim of WQI is to convert complicated water quality data to information which can be used and understood by users. This index is vital for users to know the gradation of suitable (fresh) water and unsuitable water which might be poisonous and cause serious diseases sometimes. The index might also be used to test the water quality before drilling water wells which are costly and can be really harmful to the environment; accordingly, costs and risks can be reduced a great deal. Lately, the algorithms of artificial intelligence which are suitable for nonlinear prediction and dealing with uncertain domains have been implemented in different fields of water quality estimation. The purpose of this study is to estimate the water quality index using data sets obtained from 22 camps located in six districts in Duhok city for the period March to August 2018. The data sets contain six water quality parameters which are Nitrates (NO3), Sulphate (SO4), Total Hardness (TH), PH, Total ALkalinity (T. AL) and Calcium (Ca). This paper uses the application of Adaptive Neuro Fuzzy Inference System (ANFIS) for modeling the estimation of water quality index. This model is utilized to train, test and check the index. Statistical criteria such as Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) are used to assess the performance of the ANFIS model. Investigations show that for estimation WQI, the RMSE values are 0.0346, 0.2109 and 0.0403 for training, checking and testing stages, respectively. While, the values of MSE are 0.0012, 0.0445 and 0.0016 for training, checking and testing stages, respectively. Based on the results of the criteria, the ANFIS estimation model has the ability to forecast the water quality index for Duhok camps with reasonable accuracy, and it is useful and valuable for the estimation of WQI.