A committee machine with intelligent experts (CMIE) for estimation of fast and slow shear wave velocities utilizing petrophysical logs. (August 2022)

• Record Type: Journal Article
• Authors: Gholami Vijouyeh, Ali; Kadkhodaie, Ali; Sedghi, Mohammad Hassanpour; Gholami Vijouyeh, Hamed
• Is Part Of: Computers & geosciences. Volume 165(2022)
• Publication Date: 2022-08
• Subjects: Fuzzy logic -- Neuro-fuzzy -- Neural network -- Genetic algorithm -- Simulated annealing -- Committee machine; Environmental policy -- Periodicals; 550.5
• DOI: 10.1016/j.cageo.2022.105149
• Languages: English
• ISSNs: 0098-3004

Abstract

Abstract: Petrophysical properties determination of a reservoir such as shear wave velocity plays a crucial role in exploration and production management. The purpose of this study is to develop a robust committee machine model for estimating fast and slow shear wave velocities from petrophysical logs. To this end, appropriate inputs were determined and different algorithms including artificial neural network, fuzzy logic and neuro-fuzzy were applied. Afterward, the obtained outputs were merged by utilizing optimization methods consisting of Ant Colony Optimization for Continuous Domain, Genetic and Simulated Annealing approaches. The construction of the committee machine was done using a case study including 2000 data samples from the Sarvak Formation in one of the southwestern Iran oilfields. The utilized dataset was divided into training and testing data which contain 1600 and 400 samples, respectively. Mean Squared Error was presented as a determinative factor for evaluating system performance. To obtain the best system performance with the lowest Mean Squared Error, the parameters setting of each algorithm were coded in different ranges. A total 52416, 76896 and 29400 cases were run to optimize the parameter settings of genetic algorithms, simulated annealing and continues ant colony optimization models, respectively. A weight factor is assigned to each method indicating their contribution in the final estimation. The desirable weights' combination is acquired using optimization methods. Among the individual systems, fuzzy logic and artificial neural network obtained the best performance for estimating fast and slow shear wave velocities, respectively. Also, the results of committee machine algorithms indicate a well optimization on individual systems and have superior performance over them. The most efficient optimization algorithms in the structure of committee machines for fast shear wave velocity are Genetic and Simulated Annealing and for slow shear wave velocity is Genetic Algorithm, in terms of Mean Squared Error criteria. Highlights: Simultaneous prediction of fast and slow shear wave velocities using a CMIE. Integrating the outputs of individual experts in three committee machines. A total of 52416, 76896 and 29400 cases for GA, SA and ACOR to optimize CMIE, respectively. Achieving the most efficient shear velocities estimation through CMGA. Applicability of anisotropy calculation through simultaneous fast and slow Vs estimation.