A novel method to improve dose assessment due to severe NPP accidents based on field measurements and particle swarm optimization Cláudio M.N.A. Pereira b,⇑ , Andre Przewodowski Filho a , Roberto Schirru a a Universidade Federal do Rio de Janeiro – PEN/COPPE/UFRJ, Ilha do Fundão, 21941-901, Centro de Tecnologia, Rio de Janeiro, Brazil b Comissão Nacional de Energia Nuclear – IEN/CNEN, Rua Helio de Almeida, 75 Ilha do Fundão, 21941- 906 Rio de Janeiro, Brazil article info Article history: Received 3 March 2017 Received in revised form 14 June 2017 Accepted 16 June 2017 Keywords: Dose assessment Source term estimation PSO Swarm optimization Field measurements Dose estimation correction abstract Severe nuclear power plant (NPP) accidents are those which involve significant core degradation and lead the plant to conditions more severe than a design basis accident. Under such conditions the accident pro- gression might become unpredictable and the source term estimation, imprecise by orders of magnitude. The consequence is a dose assessment very far from the reality and a deficient decision making support. This work presents a novel approach to improve accuracy of dose estimation, based on field measure- ments and particle swarm optimization (PSO) algorithm. The main idea is to determine a correction matrix, which once applied to the originally estimated (incorrect) dose distribution map, generates a cor- rected one, which better fits to the field measurements. The proposed correction matrix is the result of a concatenation of geometric transformations and an amplification/attenuation factor, aimed to fit the shape of the original map and radiation intensities in order to match the field measurements. Finding the optimum transformations (correction matrix) is, however a complex nonlinear optimization problem, which has been successfully solved by using a PSO algorithm. Results demonstrate that PSO was able to find good correction transformations, which can be used to better project future dose distributions and, consequently, improve decision making support. Ó 2017 Elsevier Ltd. All rights reserved. 1. Introduction According to the International Atomic Energy Agency (IAEA) classification, the ‘‘beyond design basis accidents” are those which comprises conditions more severe than a ‘‘design basis accident”, and may or may not involve core degradation (IAEA, 2007). When the ‘‘beyond design basis accident” involves significant core degra- dation, it is classified as a ‘‘severe accident” (SA). Due to the very nature of SA, the source term (release material and its characteristics) estimation under such circumstances becomes a very difficult task. The nuclear power plant (NPP) is beyond design conditions; instrumentation might be affected becoming unreliable; quick and unexpected changes in the plant statuses may occur, making the accident progression unpredictable (McKenna and Giitter, 1988). Therefore, great imprecision is expected in source term estimation and instrumentation readings, leading to a poor dose assessment. The projection of doses received by the people in the area affected by the atmospheric dispersion of the radioactive material will be, probably, very far from reflecting the reality and decision making would be compromised. In order to improve accuracy of dose predictions, several approaches that use field measurements to better estimate (or cor- rect) the source term have been investigated. Some of them are very simple, such as considering a linear correlation between field measurements and the source term (Athey et al., 2013). In this case a proportionality factor is applied. The problem of this method is that, in real situations, the correlation is often very non-linear. Others more complex and time consuming approaches propose solving the inverse problem aimed to characterize the actual source term based on field measurements. Most of them are applied to air pollution models in general (Chow et al., 2008; Zheng and Chen, 2010). Chow et al. (2008) proposed to solve an inverse problem to esti- mate plume dispersion in urban environments using downwind concentration measurements and Building-Resolving Simulations. The method is very time consuming and needs hours to be exe- cuted in a computer cluster. Zheng and Chen (2010) proposed a method also based on down- wind measurements. They formulate an optimization problem to be solved by a pattern search method in order to predict strength http://dx.doi.org/10.1016/j.anucene.2017.06.027 0306-4549/Ó 2017 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail addresses: cmnap@ien.gov.br (C.M.N.A. Pereira), schirru@lmp.ufrj.br (R. Schirru). Annals of Nuclear Energy 110 (2017) 148–159 Contents lists available at ScienceDirect Annals of Nuclear Energy journal homepage: www.elsevier.com/locate/anucene