The XXI International Scientific Conference of Young Scientists and Specialists (AYSS-2017)

The impact of the covariance matrix on the accuracy of unfolding process using Bayesian inference

Not scheduled

Oral Mathematical Modeling and Computational Physics

Speaker

Mr Michal Orlinski (AGH University of Science and Technology)

Description

The most important information of various nuclear systems is knowledge of accurate neutron flux spectrum. However it is impossible to directly measure the neutron radiation and moreover, it is extremely difficult to take measurements for complex nuclear systems over the unusually high level of radiation. The above factors led us to develop indirect methods of neutron energy distribution assessment. The activation foils method is one of the examples of the indirect method, however, this case usually generates the undetermined problem, which is not trivial to solve. There are several methods to solve the undetermined problem, which can be divided into two main groups: probabilistic and deterministic[3,4,5]. One of the probabilistic approaches proposed in this article using the Monte Carlo sampling from multivariate normal distribution (MVN) of prior distribution with Bayesian inference method [1,2]. The above methodology allows getting satisfactory results of neutron spectrum unfolding. Moreover using the algorithm presents above it is possible to obtain complex information about the error estimation of every energy group by computing the posterior covariance matrix. However, there are many factors that affect the accuracy of the neutron unfolding results, which still demand further investigation. This work focus on the impact of the covariance matrix of prior distribution to the accuracy and physically correct results of the unfolding process. [1] J.Cetnar, I.Królikowski, L.Ottaviani, A.Lyoussi2, Characteristics of SiC neutron sensor spectrum unfolding process based on Bayesian inference, ANIMMA 2015: the fourth international conference on Advancements in Nuclear Instrumentation Measurement Methods and their Applications: Lisboa, 20–24 April 2015. [2] J.Cetnar, I.Królikowski, Report on treatment and analysis of the outcome signal of a combined detector system, I_SMART (Innovative Sensor for Material Ageing and Radiation Testing), Work Package 4: Tools for signal recognition, InnoEnergy Innovation Project Proposal for the topic “Nuclear Energy“, 1 Dec 2014. [3] G. D’Agostini, A Multidimensional unfolding method based on Bayes’ theorem, Nucl. Instrum. Meth. A 362 (1995) 487, 1995 [4] M. Matzke, "Propagation of Uncertainties in Unfolding Procedures", Nucl. Instr. and Methods A476 (2002) 230.241, 2002 [5] S.A. Hosseini, Neutron spectrum unfolding using artificial neural network and modified least square method, Radiation Physics and Chemistry,126,75-84, 2016