Gboyega O. Ayeni and M. O. Olorunfemi
Obafemi Awolowo University, Ile-Ife, Nigeria
A software for interpretation of Schlumberger Resistivity Sounding data was developed using Koefoed's indirect technique and an Artificial Intelligence principle called Evolutionary Programming. The program (GOA_RES 1.0) was built using combined programming approach that entailed FORTRAN 77, FORTRAN 90 and VISUAL BASIC for different components of the software.
The 'forward' module computes resistivity transforms T(m) using the input model parameters and then generates apparent resistivity data through convolution of digitized values of the transform (sampled at 8 points per decade) with specified inverse linear filters. The inversion process utilizes Darwinian evolutionary principles of natural selection (population generation, fitness computation and mutation) to create several models, making comparisons between the field and generated data sets until a desired match is obtained.
The accuracy of the software was tested using six data sets including one synthetic and five real field data. The results were compared with those from a standard interpretation software (RESIST 1.0) using the same input parameters. Predicted layer resistivities from the two programs were similar. Models from GOA_RES have relatively low root-mean-square errors (0.24-3.49) compared with RESIST (1.30-3.20) and conventional partial curve matching (1.53-57.99). Correlation of depths with available driller's log showed that depths predicted by GOA_RES are within +/-4% ofthe actual values recorded during drilling.
Thus GOA_RES can be used for interpretation of Schlumberger Resistivity Sounding data in groundwater, engineering geophysical and other geoelectric applications. The algorithm may be modified for other electrode arrays as well as borehole resistivity data interpretation.