ESTIMATION OF NUCLEAR MAGNETIC RESONANCE LOG PARAMETERS FROM WELL LOG DATA USING A COMMITTEE MACHINE WITH INTELLIGENT SYSTEMS

摘要

A Nuclear Magnetic Resonance (NMR) log providesrnuseful information for petrophysical studies ofrnhydrocarbon bearing intervals. Free fluid porosityrn(effective porosity), rock permeability and boundrnfluid volume (BFV) can be obtained by processingrnand interpreting NMR data. The present studyrnproposes an improved strategy to make a quantitativerncorrelation between the NMR log parameters andrnwell logs by integrating the different intelligentrnsystems using the concept of committee machine.rnThe committee machine with intelligent systemsrn(CMIS) combines the results of Fuzzy Logic (FL),rnNeuro-Fuzzy (NF), with the Neural Network (NN)rnalgorithms for the overall estimation of the NMR logrnparameters from well log data. It assigns a weightrnfactor to each of the individual intelligent algorithms,rnshowing its contribution in the overall prediction.rnThe weight factors are derived in two ways: simplernaveraging and weighted averaging. In the weightedrnaveraging method, a genetic algorithm (GA) isrnemployed to obtain the optimal contribution of eachrnalgorithm in the construction of the CMIS. Thernpetrophysical logs from two wells are used forrnconstructing the intelligent models and a third wellrnfrom the field is used to evaluate the reliability of therndeveloped models. The results indicate that thernhigher performance of the GA can optimized thernmodel and is better than the individual intelligentrnsystems performing alone.