An improved formal failure analysis approach for safety-critical system based on MBSA

Abstract

Abstract It is important to analyze the failure in safety-critical system because a disaster may occur once any type of failure mode and/or failure effect is neglected or misjudged. In order to conduct the failure analysis more effectively and efficiently, the concept of formal modeling is introduced. This paper improved the model-based safety analysis (MBSA) working process to optimize the formal failure analysis approach of safety-critical system. As the core works of MBSA process, the formal modeling and model extension aim to build an integrated system model which can be used for analyzing the failure behaviors in the system by model checking. However, in order to automatically check if there are any potential failures in the structured system model and whether the model satisfies the specified system properties and requirements using model checker, model transformation is normally needed, which can introduced potential errors during the transformation. Moreover, different model checkers generally require the system models to be expressed in a particular input language, which increases the difficulty of modeling as well. In order to avoid these problems and improve the efficiency of failure analysis work, this paper focused on how to build an unified model of safety-critical system quickly and accurately using symbolic language SMV, and conduct automatic verification using the corresponding open-source model checker NuSMV soon afterwards. After the model checking, the formal verification results such as counter-examples generated by model checking need to be transformed into traditional failure analysis artifacts, such as FMEA and/or FTA, to guide the iterative improvement of system development conveniently. Therefore, to solve the transformation from formal verification conclusions to traditional failure analysis results is another key point of this paper. Finally, a case study about airborne equipment is provided to validate the proposed method.