Preliminary Benchmarking of SAPHSOLVE, XFTA, and SCRAM using Synthetically Generated Fault Trees with Common Cause Failures

Nowadays, probabilistic risk assessment (PRA) has contributed prominently to many technological fields, assisting decision-makers and engineers in assessing and mitigating risks inherent in their systems and designs. In the nuclear industry, PRA has become crucial in all nuclear power plant (NPP) design stages, informing engineers and decision-makers to efficiently ensure the reliability and safety of their NPPs. However, significant computational resources are required to perform PRA with current legacy PRA tools, and their enhancement is indispensable. Event tree and fault tree analyses are the most utilized methods to perform PRA in various technological industries, including the nuclear and aerospace industries. These methods comprise different modeling and analysis techniques implemented by many commercial and research PRA tools, such as SAPHIRE with SAPHSOLVE executable engine, CAFTA with PRAQuant and FTREX engines, XFTA and SCRAM engines, and OpenPRA engine. In this study, we benchmark the accuracy and computational efficiency of SAPHSOLVE, XFTA, and SCRAM engines by developing synthetically generated fault trees with various specifications and levels of complexity to challenge these codes. We compare their accuracy in terms of failure probabilities and the number of minimal cut sets, their computational speed, and memory usage. Additionally, the study's results will be utilized to enhance the quantification approach and algorithm used in the open-source PRA tool (OpenPRA) that is under development by the PRA group at North Carolina State University (NCSU).