Study On The Processing Methods For Transient Electromagnetic Data Of Central Loop

Refering to literature about the progress of researches for central loop of transient electromagnetic, analysising to one-dimensional forward and effective data processing methods, the main methods of data processing is following:(1) One dimensional modeling is the basis of iterative fitting inversion, the forward calculation accuracy can produce certain influence to the inversion result,forward modeling computing of time domain response uses the sine and cosine transform algorithm, found that computing the vertical magnetic field by cosine has high precision and computing induction electromotive force by sine has high precision, it lays a foundation for improving the precision of inversion.(2) Apparent resistivity depth section is the basis of the electric data explanation,study about two ways of calculation to the all time apparent resistivity of central loop,the all time apparent resistivity curve of the induction electromotive force kernel function is more sensitive to the electrical layered interface, but sometimes produce false phenomenon of thin layer; the all time apparent resistivity of vertical magnetic field kernel function is more smooth, and no false thin layer phenomenon, experience depth formula of smoke ring inversion sometimes is not discontinuously increasing,therefore, making change to it, the effect that directly transformation time into depth is better in this paper.(3) I study the preprocessing methods based on the characteristics of transient electromagnetic signal, three points autocorrelation and three points index approximation smoothing when noise is weak, when have a negative effect is poorer.When interference noise is bigger, appropriate wavelet threshold denoising method,wavelet decomposition layer number 3 to 5 best, SureShrink and MinimaxShrink threshold filtering effect is better.(4) Using H and K type models as examples, studying the resolution of apparent longitudinal conductance and its differential curve with apparent resistivity, the high and low resistance thin layers, when longitudinal conductance and apparent resistivity is very difficult to distinguish, the first longitudinal conductance differential curve can amplifies the original curve’s curvature variation characteristics, clearly distinguish the thin layer, low resistance thin layer is more distinguish than high resistance, but the apparent depth error is bigger, we can do depth correction to the depth of inversion based on forward model, but when the measured data lacks geological infomations, the error of depth correction still will be large. When calculating the longitudinal conductance second differential curve, using the method of cubic B spline five ponts center difference, but the calculation effect is poor. Depth correctionand second derivative problems remain to be further research.(5) Fitting inversion method, damped least square method that combines the advantages of gauss-Newton method has good convergence, and it is simply and easily programming implementation, based on Fortran language to write the inversion procedure, introducting the method of feasible direction modification parameter,inversion accuracy of A, Q, H, K four kinds of typical model is higher, because the calculation of Jacobi matrix among inversion, computing speed is slow. At the begin of inversion,we need to give the initial model, although the damping factor can make the inversion convergence, but its still strongly depends on initial model, if assuming the initial model value is more defferent to true model, the inversion may go to the local minimum, we can using the human-computer interaction to improve the inversion accuracy, but it needs higher experience and knowledge of geological for the inversion work. with the increasing of the model layers, inversion time increases obviously, so we need to continue to study the adaptive regularized least squares inversion method that work experience and the initial model requires lower.(6) The measured data based on the interference noise power degree choose different methods for pretreatment, pretreatment effect is good, smoke ring inversion depth is greater than the apparent vertical conductance depending on the depth and the depth of the least squares inversion, direct deep conversion depth minimum, inverse depth is about 1.25?2 times the depth of direct transformation deepth.