Research On Uncertainty Analysis Method Based On Latin Hypercube Sampling In The Probabilistic Safety Assessment

With the frequent occurrence of nuclear secure events in recent years, to adopt effective methods for reasonable management and assessment of the nuclear security is becoming an important topic. PSA is a method widely used in the field of safety assessment. It is a common way to build the probabilistic safety model in field of nuclear security. The uncertainty analysis based on sample method is widely used in the PSA for its universal applicability. Improved distributed hypercube sampling (IHS) method is an optimized Latin Hypercube Sampling (LHS) method, which overcomes the space non-uniformity in LHS method by putting forward the idea of optimal distance. However, the optimal distance could only achieve ideally, which can bring in error in practical situation and limits its utilization. Then based on the method of IHS, how to break the limits of IHS and introduce some new ways is an important problem in our work.In this thesis, we first had a review of the uncertainty analysis theory and the common sampling methods. Then we focused on the shortcomings of IHS method and came up with the concept of modificatory factor to optimize the method, which used to modify the optimal distance. The experimental results showed the sampling effectiveness and it could be improved significantly in the situation of low sample numbers and low sample spatial dimension. As the modificatory factor varied in different sample numbers and sample spatial dimension, we obtained the factor function by analyzing large amount of sample data. What’s more, based on the factor function, the α-IHS method had been elaborated. The experiment proved the sample distribution of the new method improves nearly10%.In this thesis, it’s well combination of methods and applications that our proposed method was applied into the reliability and probabilistic safety assessment program RiskA which was developed by FDS Team, Institute of Nuclear Energy Safety Technology, Chinese Academy of Sciences, while the sampling method of RiskA initially was simple random sample. The sampling number of the new method was less than50%of the old method; it increased the sampling efficiency by reducing the unnecessary sampling times, and will lay a good foundation for developing more efficient uncertainty methods.