Synthetic Share Wave Velocity Employing Multiple Regression and ANN Techniques for the Shale and Sandstone Formations

ABSTRACT An extensive understanding of the mechanical characteristics of rocks is essential for addressing wellbore-related problems, including wellbore instability, the inflow of sand, and reservoir collapse. Thus, to perform any activity, it is necessary to have continual profiles of the rock’s mechanical characteristics. Obtaining rock specimens in the reservoir at various depths and conducting lab testing is highly costly and time-consuming. These rock characteristics can thus be determined using the shear wave velocity measured from welllogging. However, this sonic log is frequently unrecorded due to several challenges, such as time and cost savings. For this reason, a field study was implemented in southern Iraq to address the objectives of this research. Simple and reliable mathematical models, i.e., multiple regression analysis (MRA) and artificial neural network (ANN), were developed for the clastic (shale and sandstone) formations to predict the shear wave velocity using well logging data, thence the profiles of the elastic rock parameters can be determined. The findings indicate that MRA and ANN are highly compatible with well-log data in accurately forecasting the shear wave velocity of clastic layers. ANN demonstrates superior accuracy compared to MRA, as evidenced by a better coefficient of determination (R2) of 0.92 and a lower mean squared error (MSE) of 0.06 for the shale formation, while 0.87 and 0.12, respectively, for the sandstone formation. Finally, this research proposed precise and costeffective techniques that can be applied to synthetic share wave velocity for clastic reservoirs, assisting geomechanical experts in constructing an accurate mechanical earth model.