Estimating NMR T2 distribution data from well log data with the use of a committee machine approach: A case study from the Asmari formation in the Zagros Basin, Iran

Abstract

Abstract The Nuclear Magnetic Resonance (NMR) log is one of the most valuable logs in petroleum exploration which is used to precisely evaluate the reservoir and non-reservoir horizons. Along with porosity logs (neutron, density, sonic), NMR log is used to estimate the porosity and permeability of the hydrocarbon bearing intervals. The current study focuses on estimating NMR T 2 distribution data from conventional well log data with the use of artificial intelligent systems. The eight bin porosities of the combinable magnetic resonance (CMR) T 2 distribution alongside with the T 2 logarithmic mean (T 2 LM) values are predicted using the intelligent models developed in this study. The methodology applied here combines the results of the individual models in a committee machine with intelligent systems (CMIS) for estimating the NMR T 2 distribution and T 2 logarithmic mean data. The Fuzzy logic (FL), the adaptive neuro fuzzy system (ANFIS) and artificial neural networks (ANNs) are utilized as intelligent experts of the CMIS. The NN models are developed with four different training algorithms (Levenberg–Marquardt (LM), scaled conjugate gradient (SCG), one step secant (OSS) and resilient back-propagation (RP)) and the best one is chosen as the optimal NN expert of the CMIS. The CMIS assigns a weight factor to each individual expert by the simple averaging and weighted averaging methods. A genetic algorithm (GA) optimization technique is used to derive the weighted averaging coefficients. The results indicate that the GA optimized CMIS performs better than the individual experts acting alone for synthesizing the NMR T 2 curve and T 2 LM data from one specific set of conventional well logs.