Multivariate correlations developed via radial-basis function (RBF) network analysis of a modern suite of logs are used to estimate crossplot density-neutron porosity and pseudo-deep resistivity from 1950's vintage well logs. Historically, porosity has been estimated with the logarithm of the neutron count rate standardized with a 2% porosity tight spot and a maximum porosity of ~40% in a shale stringer.Despite a frequently observed visual correlation of the gamma ray track with the neutron count rate, the gamma ray log information is not included in conventional porosity estimates.

The visual comparison of the openhole gamma ray and the neutron porosity to the shape of the crossplot porosity from a new well, Foster 4, is similar to the pattern seen between the old gamma ray and the neutron-count rate logs. The new log data were investigated with various RBF networks to correlate the gamma ray and neutron-count rate with the crossplot porosity. An optimal RBF network was then used to estimate the crossplot porosity of the Penrose sand for 22 wells in the Reed Sanderson Unit where only cased-hole gamma ray and neutron-count rate logs are available. An equation converts the modern neutron log to the old style neutron count rate while providing a normalized input to the network training. The "goodness" of the RBF network predictions was validated via exclusion testing.

A similar procedure was used to estimate a pseudo-deep resistivity in old wells. The study indicated that a carefully trained RBF network could identify the high resistivity intervals (shale stringers) with only gamma ray and neutron count rate used as the network inputs. The methodology presented should prove useful to others faced with characterizing old fields.


The lack of adequate logs was a difficulty encountered during the Reed-Sanderson Unit (Fig. 1) reservoir characterization study. This is frequently a problem when re-engineering 1950-vintage oil fields, where only cased gamma ray (GR) and neutron count rate (NR) are available. The Reed-Sanderson Unit was acquired by Lynx Petroleum Consultant, Inc. from Conoco Inc. in 1992. Lynx entered into an agreement with Los Alamos National Laboratory under the ACTI Advanced Reservoir Management CRADA No. LA95C102371,2 to demonstrate the application of computer modelling/communications technology as a means of reducing costs and increasing the effectiveness of producing oil and gas from the Penrose Sand zone in the unit.

As a primary objective, the estimation of porosity, permeability, and initial water saturation distributions plays a crucial issue in accurately evaluating the reservoir and the simulation model. However, the lack of density and resistivity logs in the unit makes it difficult to assess the reservoir properties. In this work, a new approach was investigated that uses radial-basis function (RBF) networks to generate pseudo-crossplot density-neutron porosity font=symbol charset=fontspecific code=70TeX='\Phi ' descr='[PHgr]'X) and pseudo-deep laterolog (LLD) from old GR and NR readings.

This content is only available via PDF.
You can access this article if you purchase or spend a download.