Monte Carlo simulation of dual-spaced neutron porosity well-logging tool responses

It is difficult to correct porosity measurements from dual-spaced neutron logs for lithology, fluid properties, temperature, pressure, and other borehole environmental conditions by using only laboratory test pit data. Mathematical models, particularly those based on Monte Carlo simulation, are being used to study these effects and augment these corrections. However, typical general purpose Monte Carlo computational codes require several hours on large super computers.; This work presents a new Monte Carlo code (McDNL) which has been developed specifically for dual-spaced neutron porosity tool simulation. The code features two improved variance reduction techniques that increase the computing efficiency by orders of magnitude over analog simulation. They are the direction biasing and statistical estimation methods. These methods are used in lieu of the more common method of geometric splitting.; In the direction biasing scheme, the neutron direction is biased in the preferred direction (detector direction) and its weight is properly modified. As a result the neutron can reach the detector in fewer collisions and the associated variance in the detected fraction is reduced.; Instead of using the analog estimation method which scores only the neutrons that physically reach the detector, the statistical estimation method is used in which an estimate of the score following each scattering event is determined in the Monte Carlo history.; Other commonly used variance reduction schemes including implicit capture, Russian roulette, and the exponential transform are also used to obtain maximum computing efficiency. Correlated sampling is also included as an option within the code to evaluate environmental corrections to tool response at little extra cost in the computing time. The code is easy to use and unlike the common general purpose Monte Carlo codes, it requires little expertise in the Monte Carlo method.; The code has been benchmarked against five sets of laboratory test pit data and proved to be valid and at least as fast as the general purpose Monte Carlo codes. For a typical 6" open bore hole the current version of the code takes on the average 24 hours on a DEC Microvax II computer to yield a 5% error in the near-to-far detector counting rate ratios.