SMV model-based safety analysis of software requirements

Fault tree analysis (FTA) is one of the most frequently applied safety analysis techniques when developing safety-critical industrial systems such as software-based emergency shutdown systems of nuclear power plants and has been used for safety analysis of software requirements in the nuclear industry. However, the conventional method for safety analysis of software requirements has several problems in terms of correctness and efficiency; the fault tree generated from natural language specifications may contain flaws or errors while the manual work of safety verification is very labor-intensive and time-consuming. In this paper, we propose a new approach to resolve problems of the conventional method; we generate a fault tree from a symbolic model verifier (SMV) model, not from natural language specifications, and verify safety properties automatically, not manually, by a model checker SMV. To demonstrate the feasibility of this approach, we applied it to shutdown system 2 (SDS2) of Wolsong nuclear power plant (NPP). In spite of subtle ambiguities present in the approach, the results of this case study demonstrate its overall feasibility and effectiveness.