Behavioral Fault Modeling for Model-based Safety Analysis

Abstract

Recent work in the area of Model-based Safety Analysis has demonstrated key advantages of this methodology over traditional approaches, for example, the capability of automatic generation of safety artifacts. Since safety analysis requires knowledge of the component faults and failure modes, one also needs to formalize and incorporate the system fault behavior into the nominal system model. Fault behaviors typically tend to be quite varied and complex, and incorporating them directly into the nominal system model can clutter it severely. This manual process is error-prone and also makes model evolution difficult. These issues can be resolved by separating the fault behavior from the nominal system model in the form of a "fault model", and providing a mechanism for automatically combining the two for analysis. Towards implementing this approach we identify key requirements for a flexible behavioral fault modeling notation. We formalize it as a domain-specific language based on Lustre, a textual synchronous dataflow language. The fault modeling extensions are designed to be amenable for automatic composition into the nominal system model.