Results 1 - 10 of 3420
Results 1 - 10 of 3420. Search took: 0.022 seconds
|Sort by: date | relevance|
[en] Highlights: • Core Damage Frequency and Large Early Release Frequency. • Multi –Unit Risk Metrics. • Aggregation of CDF of NPP through Mean Values. • Aggregation of CDF of NPP as Random Variable. - Abstract: The nuclear generating sites around the world are mostly twin unit and multi-unit sites. The PSA risk metrics Core Damage Frequency (CDF) and Large Early Release Frequency (LERF) currently are based on per reactor reference. The models for level 1 and level 2 PSA have been developed based on single unit. The Fukushima accident has spawned the need to address the issue of site base risk metrics, Site Core Damage Frequency (SCDF) and Site Large Early Release Frequency (SLERF), on the site years rather than reactor years. It is required to develop a holistic framework for risk assessment of a site. In the context of current study, the holistic framework refers to integration of risk from all units, dependencies due to external events and operation time of individual units. There is currently no general consensus on how to arrive at site-specific risk metrics. Some documents provide suggestions for site CDF and site LERF. This paper proposes a new method of aggregation of risk metric from the consideration of operating time of individual units under certain assumptions with a purpose to provide a new conceptual aspect for multi-unit PSA. The result of a case study on hypothetical data shows that site level CDF is not sum of CDF of all units but around 18% higher than unit level CDF. When the CDF is considered to be a random variable then, the new methodology produces site CDF as 50% higher than single unit CDF. These two approaches have been detailed in the paper. For a general data set of CDF for individual units, site CDF would more than individual unit CDF however, it would not be multiples of a single unit value.
[en] The multiple safety systems including high pressure safety injection (HPSI), low pressure safety injection (LPSI), safety injection tank (SIT), and etc. have been designed to protect the core under the accidents in NPPs. If the decay heat from reactor is not removed due to the failure of safety systems under the accidents, the core can be melted. Therefore, the monitoring of reactor internal phenomenon is very important to prevent core meltdown. The deep learning model can be simulated for reactor internal phenomena without knowledge of physical. Deep learning is one of the most active research fields in recent years because computer’s performance has been improved. It has been widely used not only in science but also in various industries such as medicine, advertising, and finance. Deep learning is a technology that applies information processing methods of human brain to machines. The basic structure of Deep learning has a multilayer perceptron (MLP) structure consisting of three or more hidden layers. The MLP is a neural network composed of several nodes and layers. The location of data is as follows: The thermal distribution of core cell, core baffle, bypass, support barrel, down comer, and vessel cylinder. The data was obtained by using MELCOR which is the severe accident analysis code. The operators can maintain the integrity of the reactor when an unexpected severe accident is occurred in the nuclear power plant because the developed model can predict the reactor internal phenomena by the thermal distribution of vessel cylinder.