Feed Forward Artificial Neural Network Model to Estimate the TPH Removal Efficiency in Soil Washing Process

Background & Aims of the Study: A feed forward artificial neural network (FFANN) wasdeveloped to predict the efficiency of total petroleum hydrocarbon (TPH) removal from acontaminated soil, using soil washing process with Tween 80. The main objective of thisstudy was to assess the performance of developed FFANN model for the estimation ofTPH removal.Materials and Methods: Several independent repressors including pH, shaking speed,surfactant concentration and contact time were used to describe the removal of TPH as adependent variable in a FFANN model. 85% of data set observations were used for trainingthe model and remaining 15% were used for model testing, approximately. Theperformance of the model was compared with linear regression and assessed, using Root ofMean Square Error (RMSE) as goodness-of-fit measureResults: For the prediction of TPH removal efficiency, a FANN model with a threehidden-layer structure of 4-3-1 and a learning rate of 0.01 showed the best predictiveresults. The RMSE and R2 for the training and testing steps of the model were obtained tobe 2.596, 0.966, 10.70 and 0.78, respectively.Conclusion: For about 80% of the TPH removal efficiency can be described by theassessed regressors the developed model. Thus, focusing on the optimization of soilwashing process regarding to shaking speed, contact time, surfactant concentration and pHcan improve the TPH removal performance from polluted soils. The results of this studycould be the basis for the application of FANN for the assessment of soil washing processand the control of petroleum hydrocarbon emission into the environments.