(A) study on the method of combining empirical data and a deterministic model for fuel failure prediction

Abstract

Some difficulties are encountered when the behavior (i.e., fuel failure probability) of complex systems which have unreliable deterministic models is predicted. For more realistic prediction of the behavior of complex systems with limited observational data, the present study was undertaken to devise an approach of combining predictions from a deterministic model and actual observational data. Predictions by this method of combining were believed to be of higher reliability than separate predictions made by either model taken independently. A systematic method of hierarchical pattern discovery based on the method developed in the SPEAR was used for systematic search of weighting factors and pattern boundaries for the present method. This method of hierarchical pattern discovery was found to be effective in two or three dimensional feature space models. A sample calculation was performed for prediction of CANDU fuel failures which had occurred due to power ramp during refuelling process. The PROFIT model which is a failure model for estimating the probability of failure in transient increase in power based on a standard statistical method and a mechanistic analysis was adopted as a deterministic model, and the ramp-related fuel failure data of the Pickering 8-bundle shifting process was used as an empirical data set. A region of feature space in which failure probabilities from the PROFIT model nearly agree with those from observational data was demonstrated to exist by the sample calculation.