Nuclear Magnetic Resonance (NMR) technique is currently used in well logging to provide important informations so that the interpreter can analyze the physical properties of oil reservoirs. This tool provides informations, in a more effective way than the conventional logs on free and irreducible fluids, total and effective porosity, and permeability, which allows a better reservoir interpretation and the consequent hydrocarbon evaluation. One of the ways to make this is using artificial intelligence approaches as Genetic Algorithm, Fuzzy Logic, Neural Network, etc., to synthesize porosity and permeability parameters derived from NMR log, using as input only conventional logs of petroleum industry, such as natural gamma ray, spontaneous potential, caliper and resistivity. This process is justified because NMR log is currently expensive and, for this reason, it is not done at all wells and can not be obtained in cased holes, but provides physical characteristics of the reservoirs which are hardly obtained by other means, either through conventional geophysical logs or formation evaluation techniques. Initially, we developed an algorithm, using Fuzzy Logic Tool of MATLAB software, with log data of East Texas, aiming to see the influence of the number of membership functions in clustering. When three linguistic classes were selected, the results were not satisfactory because the synthetized logs showed a box form. Then, when we removed the middle linguistic class, an improvement was observed in the synthetized logs, however, still having a box pattern. Finally, using the ISODATA algorithm of MATLAB, a better fit with the real porosity and permeability curves was obtained, eliminating box shape and promoting a more careful division of clusters. In this form, we demonstrated the feasibility of the simulation of parameters of NMR log through Fuzzy Logic.