Combined approaches are now applied in safety analyses of design-basis accidents. They consist of using best-estimate models in the computer codes together with their estimated uncertainties, and require the most unfavorable initial and boundary conditions (IBCs) to be found with respect to the plant operating conditions. This implies determining first the worst-case scenarios, then predicting the figures of merit (FOMs) that must fulfill safety criteria. Such scenarios can be identified by sensitivity studies on IBCs resulting in an input vector of fixed values to realize a deterministic bounding calculation. However, it is a difficult and time-consuming iterative task especially for complex transients with interactions between parameters. Alternatively, the RIPS (Reduction of the Interval of variation of the Parameters of the Scenario) method has been developed in a best-estimate plus uncertainty approach to find the worst IBCs as a set of reduced ranges of variation of the related inputs, rather than by a vector of discrete values. It defines a critical zone for which the FOM is maximized (or minimized). To this end the RIPS method provides quantitative and graphical outcomes enabling identification of the detrimental (or favorable) ranges of variation of the preponderant IBC parameters. This is done through a statistical analysis of a large set of calculations in which all the input parameters and code model uncertainties are randomly sampled. The RIPS method analyzes the higher (or lower) quantiles of the FOM cumulative density function and determines for each input parameter the critical zone within its variation interval, i.e., where it is the most influential. Correlations between parameters are also detected. This paper describes the RIPS method and demonstrates with several examples its ability to adequately identify the critical zone of the IBC configuration space.