Uncertainty and regression analysis of the MSLB accident in PWR based on unscented transformation and low rank approximation

•The singular value decomposition (SVD) and the Unscented Transform (UT) algorithms are combined (SVD-UT).•The computational time is drastically reduced by revealing the active subspace thanks the to Low Rank Approximation (LRA) algorithm.•Both SVD-UT and LRA-UT results are compared with the random sampling technique.•The uncertainty quantification is performed during MSLB in PWR, where the response variables are computed by ATHLET thermal-hydraulic code.•The SCALE 6.2 code is used for calculating the reactivity coefficients and the covariance matrix. The present studies focus on the quantification of uncertainty during the main steam line break accident scenario (MSLB) in PWR, assuming that there is a failure on the feed-water regulating valve of the broken steam generator. The scenario is characterized by the associated positive Doppler and coolant density reactivities which bring the core back to critical (return-to-power). Accordingly, the input uncertainty parameters are the Doppler and coolant density reactivities taking into account the correlation matrix among the input parameters, which is calculated by SCALE 6.2 code. The main safety parameters are the maximum cladding surface temperatures and average core power during the accident which are computed by ATHLET thermal-hydraulic code. The sampling-based uncertainty technique is considered to be the most dependable technique which can be applicable to any code, however it is computationally expensive. Therefore, it is important to develop efficient techniques which are capable of reducing the calculation time. The first approach is the SVD-UT where the Unscented Transform (UT) algorithm and singular value decomposition (SVD) are combined to generate a minimal sample points. In addition, due to the strong correlation between the input reactivities, the computational time can be further reduced by implementing the Low Rank Approximation (LRA) and revealing the active subspace.