Multivariate Gaussian Process Regression for Evaluating Electromagnetic Profile in Screening Process of Seabed Logging Application

• Main Authors: Mohd Aris, M.N., Daud, H., Mohd Noh, K.A., Dass, S.C.
• Format: Conference or Workshop Item
• Published: Springer Science and Business Media B.V. 2021
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85123299195&doi=10.1007%2f978-981-16-4513-6_43&partnerID=40&md5=0d1719fe4a2071ea2f25b0bda0d981e8; http://eprints.utp.edu.my/29270/

Computer simulation is an important task for reservoir modeling in screening process of seabed logging. Information acquired from the computer simulation could provide reliable information of electromagnetic (EM) profile and subsurface underneath the seabed. However, the computer simulation could be a time-consuming task in the screening process due to its intricate mathematical equations. In this paper, a predictive model based on Gaussian process regression (GPR) is used to provide information of EM profile at various observations with low time consumption. Multivariate GPR model is developed based on computer simulation outputs. Normalized magnitude versus offset plots are analyzed to eliminate data from any undesired wave interaction. Root mean square error and coefficient of variation between the GPR model and the computer simulation outputs at untried observations are computed. On average, the resulting error was 0.0352 and the coefficient of variation was less than 0.5. This indicates the multivariate GPR model is well-fitted and capable of evaluating EM profile with low processing-time. © 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.