Developing Synthetic Magnetic Resonance Logs (CMR logs) from Conventional Well Logs

NMR logs use nuclear magnetic resonance to estimate high resolution total porosity, free-fluid volume, bound-fluid volume and permeability. In this study twomethodologies to generate synthetic CMR logs are presented. In this approach ANN is used as the main tool. Neural network model is developed using CMR logs and conventional logs such as gamma ray, neutron, resistivity and etc. obtained from three wells of South Pars Field, Iran. Then the model is applied to generate synthetic MRI logs such as Free Fluid porosity (CMFF), Bound Fluid porosity (BFV) and permeability (KTIM) by using just the conventional logs in another well. The Synthetic logs are generated through two different methods of Backpropagationand Generalregression. The results presented that MRI log can be generated with a high degree of accuracy. And thebest performance was obtained for Generalregression architecture with approximately 93%- 98% accuracy. By using this method the synthetic MR logs can be predict for all the wells in the field and a much better reservoir characterization can be achieved at a much lower cost . Also Generalregression is a more powerful method to generate synthetic logs compared to Backpropagation especially in case of lack of the data.