Accurate determination of shear wave velocity using LSSVM-GA algorithm based on petrophysical log

Shear wave velocity (Vs) is regarded as one of the most crucial parameters in reservoir geotechnics because of its employment in the determination of other petrophysical parameters. Two main sources of data, including laboratory data extracted from core examination and petrophysical logs, are used for Vs estimation. Petrophysical logs are the most common data to determine Vs, because of their availability and simplicity in recording and analyzing. Thus far, many empirical equations have been proposed to determine Vs applying petrophysical logs. However, these empirical methods remarkably suffer from the low degree of precision delivered when applied to a different field. Artificial intelligence models have been found to be efficient tools in addressing this lack of generalizability problem. Therefore, in this study, by combining least square support vector machine with genetic algorithm, a hybrid artificial intelligence model was developed to accurately predict Vs using six petrophysical logs as input variables. The accuracy of the hybrid model was then compared with five common empirical models previously proposed. The results achieved present that the newly configured model evaluated can make a much more precise estimation of Vs (R2= 0.9813, RMSE=0.411 km/s) when compared to all five empirical models reviewed in the present research.