Energy Sciences & Technology Dept. Seminar
"UNCERTAINTY QUANTIFICATION FOR ACCIDENT MANAGEMENT USING ACE SURROGATES"
Presented by Athi Varuttamaseni, Ph.D., BNL
Wednesday, April 10, 2013, 10 am
Building 130, Room 2-26
Hosted by: David Diamond, Ph.D.
The alternating conditional expectation (ACE) regression method is used to generate RELAP5 surrogates which are then used to determine the distribution of the peak clad temperature (PCT) during the loss of feedwater accident coupled with a subsequent initiation of the feed and bleed (F&B) operation in the Zion-1 nuclear power plant. The construction of the surrogates assumes conditional independence relations among key reactor parameters. The choice of parameters to model is based on the macroscopic balance statements governing the behavior of the reactor. The peak clad temperature is calculated based on the independent variables that are known to be important in determining the success of the F&B operation. The relationship between these independent variables and the plant parameters such as coolant pressure and temperature is represented by surrogates that are constructed based on 45 RELAP5 cases. The time-dependent PCT for different values of F&B parameters is calculated by sampling the independent variables from their probability distributions and propagating the information through two layers of surrogates. The results of our analysis show that the ACE surrogates are able to satisfactorily reproduce the behavior of the plant parameters even though a quasi-static assumption is primarily used in their construction. The PCT is found to be lower in cases where the F&B operation is initiated, compared to the case without F&B, regardless of the F&B parameters used.