Application of Soft Computing Method for predicting rock Stiffness and Lame parameter by seismic attributes

Adib Amini¹ Esmael Makarian² Ahmad Taheri³

1) MSc. of Civil Engineering of Sahand University of Technology:
2) MSc. of Mining Engineering of Sahand University of Technology:
3) MSc. of Computer Engineering, of Yasouj University:

محل انتشار : کنفرانس بین المللی پیشرفت های اخیر در مهندسی، نوآوری و تکنولوژی - بلژیک(eitconf.com/1st)
Abstract :
Geomechanical characterization is one of the significant steps in subsurface studies. Stiffness (M) and lambda parameters are two critical factors commonly used to evaluate rocks. There are some ways to masseur them which they are classified by two main methods, including direct and indirect methods. Direct methods are done on coring samples by laboratory test; however, some problems limit these methods. For example, obtaining cores in some situations is difficult or impossible. In this paper, using Deep Artificial Neural Network (DANN) based on seismic velocities, we predict stiffness and lame parameters. Finally, all results are evaluated by (R), Root Mean Square Error (RMSE), and Mean Square Error (MSE). The results prove that the DANN method should be considered a suitable tool for predicting target parameters with R=0.97 and 0.98 for stiffness and lambda parameters, respectively. Furthermore, RMSE and MSE for lame prediction are less than that of stiffness.
Keywords : Soft computing approaches, Deep learning neural network, rock stiffness, lame parameters, seismic velocities, porous media.