时间 : 2019年12月12日 09时00分
地点 : 能源与动力工程学院303
主办单位 : 工程学部
协办单位 : 能源与动力工程学院
Uncertainty and Sensitivity Methodologies in Best Estimate Safety Analyses
主讲人 : Prof.Rafael Macian-Juán
The safety analysis of nuclear Systems is made today mostly with the use of sophisticate system analysis codes that integrate various physical aspects of their behavior, such a thermal-hydraulics, neutronics and fuel behavior. These computer codes solve the differential equations that describe fluid flow, neutron transport and fuel material behavior with the help of advanced numerical methods and a set of physical models to simulate the physical processes that drive the system’s behavior.
In the past, a conservative approach was usually employed, that sought to present the most challenging conditions for the simulation of nuclear transients and accident situations in a manner in which, if the plant was able to cope with such situations in a safe manner, then any realistic scenario would be as well overcome. Even though today conservative analysis is still required by the licensing and regulatory authorities, a more physically realistic type of analysis has become more and more preferred, because of its fidelity in the representation of the actual plant’s behavior. It is d on the use of the sophisticate codes mentioned above, which include the best available physical modelling, thus producing the best estimates for the relevant safety variables of the plant: it is the so-called Best Estimate Analysis. It can provide plant with increased operational margins and, thus, make them more efficient and in a sense safer to use, than conventional conservative calculations. As was shown in the case of SBLOCAS, conservative approaches are not always “safer”.
Relying on state-of-the-art and ever improving physical and numerical methods requires also a quantification of their capability to accurately models what they are supposed to model. Because of lack of knowledge or inaccuracies in the necessary data to build the computer models of reactors and plants, uncertainty is introduced in the estimates of the plant variables as predicted by Best Estimate codes. Soon after the adoption of the Best Estimate methodology, this issue was recognized and addressed with the development of methodologies for the propagation of uncertainties in the calculations and the determination of the sensitivities of the safety variables of interest to the uncertainties in the codes’ physical models and system de ion parameters.
This talk will present an overview of the Best Estimate approach and a de ion of uncertainty propagation and sensitivity methodologies d on a Monte-Carlo approach supported by non-parametric statistics and surrogate model development. These methodologies can be successfully applied to Multiphysics coupled calculations and can propagate many sources of uncertainties in the simulation results which can shed light on quality of the simulations, determine margins for operation with a desired statistical significance, and contribute to the evaluation of the suitability of models and data for the simulation of physical processes in the nuclear reactor and that plant. Sensitivity measures can also contribute to identifying important core or plant parameters or physical processes that contribute significantly to the progression and severity of accidents.
Prof. Rafael Macián-Juan, MSc Energy Systems (1990), University Polytechnic of Valencia, Spain, MSc in Nuclear Engineering (1993), and PhD in Nuclear Engineering (1996), Penn State University, USA, is the head of the Lehrstuhl für Nucleartechnik (Chair of Nuclear Engineering) at the Technical University Munich (TUM). He has more than 25 years of experience in nuclear safety and in the development and application of computer- d methodologies for the analysis of nuclear systems. Prof. Macian-Juan worked for ten years at the Paul Scherrer Institute (PSI), Switzerland, carrying out research and development in the areas of reactor thermal-hydraulics and coupled neutronics calculations, and performing safety evaluations for the Swiss Nuclear Power Plants. In 2007 he joined the faculty of mechanical engineering at TUM to take charge of the new Lehrstuhl für Nukleartechnik established thanks to the initial support from E.On-Kernkraft. Prof. Macián-Juan has published more than 100 articles in refereed journals and in peer-reviewed conferences on nuclear engineering related subjects. His main areas of research are d on nuclear safety with Multiphysics and multiscale simulation codes, uncertainty and sensitivity methodologies, experimental and numerical thermal-hydraulics, medical application of radiation transport, and the safety analysis and development of future nuclear reactors (mainly molten salt d) and nuclear fuel designs. Recently a new line of research on the application of machine learning to nuclear safety has been initiated. He is currently also a Guest Professor at Harbin Engineering University in China.