Identifying high-risk scenarios of complex systems using input domain partitioning

Scenario based dynamic analysis is an important technique used in the verification of specification models for complex real time systems. One of the important problems facing developers of these systems is conducting risk analysis at early stages of development. Our methodology for risk assessment uses colored Petri net (CPN) models for predicting risk factors of system components, based on severity and complexity measures. CPN models are developed from system requirements specifications, and risk analysis provides guidance for identifying high risk components prior to their actual design and implementation. The analysis of the specification models is performed through scenario based simulations. Even though the set of scenarios used for simulation is very important for the success of risk analysis, the scenarios are chosen in an ad hoc fashion, usually guided by the experience of domain experts. Therefore, it is likely that some important scenarios are overlooked, due to the complexity of the system. We propose a technique that increases the likelihood that high risk scenarios are identified. The technique is based on input domain partitioning. Partitions can be determined from the given CPN model automatically. Predicates, which describe subdomains of the input space, assist users in revealing interesting scenarios. This methodology is applied to the assessment of a commanding component of NASA´s Earth Observing System (EOS)