In the framework of risk assessment in nuclear accident, simulation tools are widely used to under-stand and model physical phenomena. These simulation tools take into account a large numberof uncertain input parameters. We often use Monte-Carlo type methods to explore their range ofvariation: the inputs space is randomly sampled and a code run is performed on each sampledpoint. However, some of these code runs may fail to converge. Analyzing these code failures tounderstand which of the inputs have the most influence on them lead to a better understandingof how the code works. It also intends to improve the robustness of the simulation software andcode computations. For this purpose, we propose two complementary approaches performing astatistical analysis of the code failures. The first approach is based on Goodness-of-fit tests andcompares conditional probability distributions according to code failures to a reference one. Asecond approach, based on a dependence measure named the Hilbert-Schmidt Independence Crite-rion (HSIC), provides another way to measure the global dependence between the inputs and thecode failures. The development of this methodology is carried out in the context of severe nuclearaccidents. More especially, the presented methods are applied for the study of a simulation codethat simulates the fuel-coolant interaction in severe nuclear accident context, MC3D.