Model-based development (MBD) is increasingly being used for system-level development of safety-critical systems. This approach allows safety engineers to leverage the system model created during the MBD process to assess the system's resilience to component failure. In particular, one fundamental activity is the identification of minimal cut sets (MCSs), i.e, minimal sets of faults that lead to the violation of a safety requirement. Although the construction of a formal system model enables safety engineers to automate the generation of MCSs, this is usually a computationally expensive task for complex enough systems. We present a method that leverages Max-SMT solvers to efficiently obtain a small set of faults based on a local optimization of the cut set cardinality. Initial experimental results show the effectiveness of the method in generating cut sets that are close or equal to globally optimal solutions (smallest cut sets) while providing an answer 5.6 times faster on average than the standard method to find a smallest cut set.