A study of uncertainties in the prediction of neutron porosity logging-tool responses

The prediction of logging-tool responses suffers from both systematic and random sources of error of which only the latter can be treated in a scientifically objective manner. The use of a point-energy Monte Carlo code with a versatile modeling package essentially eliminates systematic error. The authors present a detailed discussion of all sources of random uncertainty, with particular reference to the responses of a dual-spaced neutron porosity tool (WILT, Winfrith Logging Tool) in both water and sandstone. Sensitivity analysis is used to identify the principal contributions to the total calculational uncertainty, and the statistical credibility of the comparison between measurements and calculations is discussed. Uncertainty in nuclear cross-sections is shown to be the dominant single source of uncertainty in calculated reaction rates. This applies with particular force as source-counter separation increases. The accuracy of the counting ratio calculation depends largely on the accuracy of the far counter calculation