Artificial Neural Network (ANN) modelling of elemental concentration in soil using XRay Fluorescence (XRF) counts data
Loading...
Date
2021-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand)
Abstract
Generally, fundamental parameter approach is used for the analysis of the XRF data for which various parameters needs to be calculated. To make the process less strenuous, Artificial Neural Network (ANN) has been used in the current work. The aim of the study is to develop a model of elemental concentration, after obtaining the Energy Dispersive X-Ray Fluorescence (EDXRF) intensities and counts data of soil samples. An ANN model is designed to obtain the concentrations of elements without using fundamental parameters. The designed network was the feed-forward neural network having two hidden layers with five and seven neurons in each layer, respectively. Quick propagation algorithm has been applied to the network. The network has been trained using supervised learning. A total number of 12 elements in soil samples, namely Mg, Al, Si, K, Ca, V, Mn, Fe, Cu, Rb, Sr and Pb, has been used for the training of the model. A wide range of concentration (7.66 ppm for Cu to 228360 ppm for Si), obtained using EDXRF results, were modelled by ANN. The correlation coefficient and R-square values for the training set came out to be greater than or equals to 0.97 for most of the elements. The R-square values for the prediction set were also satisfactory. Overall, good results were obtained by the ANN model for almost entire range of concentration except for a few elements. So, it can be concluded that ANN works well for the elemental concentration modelling avoiding the rigorous fundamental parameters approach. Multiple regression has also been applied and the results were satisfactory.