Implementation of a direct perturbation method in MCNP and application to SCALE verification

• Published in: Annals of nuclear energy 2013-12, Vol.62, p.291-297
• Main Authors: Trottier, A., Blomeley, L., Chow, J.C., Colton, A., Masala, E., Shukhman, B., Watts, D., Wilkin, B.
• Format: Article
• Language: English
• Subjects: Monte Carlo; Nuclear data; Reactor physics; Sensitivity and uncertainty
• ISSN: 0306-4549; 1873-2100
• DOI: 10.1016/j.anucene.2013.06.020

•77 Heavy-water moderated critical ZED-2 experiments analyzed with MCNP and KENO.•SCALE used to find nuclear data adjustments minimizing MCNP a priori biases.•MCNP patch used to repeat analysis using nuclear data adjustments from SCALE.•Present a detailed code-to-code comparison of of a priori and a postiori keff biases.•SCALE results consistently underestimate impact of proposed corrections by 1–2mk. Many nuclear data uncertainty propagation methods are implemented on the basis of first-order perturbation theory. These methods are complex and integral verification using direct perturbation of the nuclear cross section data is difficult due to the structure of nuclear data files. We present a new implementation of the direct perturbation method, which eliminates the need to modify these files. The method was implemented in DPERT, a patched version of MCNP5. Using the DPERT patch, we present an integral verification of TSURFER based on 77 heavy water moderated ZED-2 critical experiments. TSURFER is a module of the SCALE code suite and applies first-order perturbation theory to propagate nuclear data uncertainties. The experiments were modeled using the standard MCNP5 code to establish the a priori keff calculation biases. TSURFER was used to minimize these biases by adjusting the underlying nuclear data. The proposed cross section alterations were then applied to the experiment models, and the DPERT patch was used to verify TSURFER’s evaluation of the a posteriori keff biases. The study confirmed the TSURFER bias reduction prediction, but suggests TSURFER may underestimate the impact of the nuclear data corrections by 1.35±0.05mk on average.