Young Researchers and Elite Club, Miyaneh Branch, Islamic Azad University, Miyaneh, Iran
Abstract: (4415 Views)
Some uncertainties in the field parameters such as dispersion and hydraulic conductivity, unknown boundary conditions and the noise of the measured data are among the main limiting factors in the groundwater flow and contaminant transport modeling. Thus, simulation of contaminant transport can be an important task in hydro-environmental researchs and consequently, it is necessary to develop the robust models which can determine the temporal and spatial forecast of contaminant. For temporal modeling contaminant concentration, several numerical methods, such as finite volume method, finite difference method, boundary element method and finite element method have been used for computional solution of governing advection-dispersion partial differential equation. In this study, a new hybrid model based on adaptive neuro-fuzzy inference system (ANFIS) as an black-box model and radial basis function (RBF) as a meshless method was developed. In fact, the proposed method employed the advantageous of both arthificial intelligence and meshless techniqus for modeling contaminant transport in porous media. In this research, an experimental was done for examining the efficiency of the proposed method. In this way, an acrylic sand tank was made with ten piezometers, one inlet with three adjustors. In order to supply contaminant a submersible pump was used. Also, constant water level was maintained using adjustor valves at both end of the tank. The thickness of acrylic sand tank 10 mm and dimensions 2.00×1.30×0.20 m3 were chosen. The sand sample porosity was measured 0.3. The grid size and time interval were considered 0.1×0.1 m×m and 3-minute, respectively. The constant-head test was employed to meaure the hydraulic conductivity of soil as a standard laboratory test. An UV Spectrophotometer (DR5000, HACH Company, USA) was used for measurement of the AO7 concentration. The maximum wavelength was measured 485 nm for AO7 concentration. Also, an electrical conductivity meter (EC600, A FLIR Company, USA) was used for measurement of the resistivity and electrical conductivity of AO7. In this study, time series of AO7 concentration observed at different piezometers of sand tank were firstly de-noised by the wavelet-based data de-noising approach. Then, the effect of noisy and de-noised data on the performance of ANFIS model was compared. For this end, time series of AO7 concentration observed in 10 different piezometers were trained and verified via ANFIS model to predict the AO7 concentration at one month ahead. Then, considering the predicted AO7 concentration of piezometers as interior conditions, the multiquadric radial basis function as a meshless method which solves partial differential equation of contaminant transport modeling in porous media, was employed to estimate AO7 concentration values at any point within the study area (in the experiment, sand tank) where there is not any piezometer. In this stage, optimal values of dispersion coefficient in advection-dispersion partial differential equation and shape coefficient of MQ-RBF were determined using cross validation approche. The cross validation method was finally applied to verify the performance of the proposed ANFIS-RBF model for two piezometers which were not considered in the calibration stage.In temporal contaminanat transport modeling, de-noised data enhanced the performance of ANFIS methods up to 5 percent in the experimental study. Results showed that the efficiency of ANFIS-RBF model is a reliable thechnique for contaminant transport modeling in porous media.
Article Type:
Original Manuscript |
Subject:
Earthquake Received: 2018/01/9 | Accepted: 2018/11/11 | Published: 2018/11/15