Induced Electrical Polarization in Geophysical Prospecting for Economic Minerals

Abstract

The electrical properties of earth materials, as described usually by the Cole­ Cole model, can be used to discriminate earth materials in electrical and electromagnetic exploration methods. The dispersion of the electrical properties with frequency may lead in the transient electromagnetic (TEM) coincident or central loop systems to the appearance of a characteristic transient late time negative voltage called negative response (NR) phenomena. A main goal of the present work is to investigate the possibility of inverting for the TEM response of the coincident loop above layered polarizable ground to obtain the Cole-Cole parameters of the polarizable material which represents an economic mineral. The Neural Networks (NN) and the Genetic Algorithms (GA) inversion techniques and combinations of them are investigated for making the inversion. The NN method is investigated in more detail due to its efficiency since once the network is properly trained for certain ranges of the Cole-Cole parameters, inversion can be made quickly within these ranges of parameters. The different factors affecting the learning error of the NN are •investigated such as the number of inversion parameters, the ranges of the parameters and the number of samples in each parameter range. In addition, the effect of the amplitude and the duration of the NR in the transient pattern is investigated, which in turn, depends on the sampling period of the TEM response and its starting time, the loop radius, the layering structure of the ground, the conductivity values and contrasts and the polarization parameters. Accurate inversions are performed for the TEM response above polarizable ground in the form of a half­ space or a two-layer model with the upper or the lower layer being polarizable. The possibility of improving the inversion using a second NN inversion or GA inversion in narrow ranges of parameters with higher resolutions are explored. The effect of random natural noise in the response on the inversion is discussed. The possibility of estimating the uncertainty of the inversion due to the ambiguity or equivalence of the TEM response at different values of the model parameters is investigated. The effect of the sensitivity of the TEM response to the Cole-Cole parameters and its effect on the inversion are discussed. Successful inversion for the Cole-Cole parameters of a case study is performed.