Modeling and risk analysis of hardware-software systems

Hardware-software systems are widely used now and must be safe and reliable. Manual analysis of risks for structural complex systems is very expensive, so formal automated methods are required. The most important aspect here is the possibility to describe safety requirements in terms used in safety theory, such as Markov chains or logic-probabilistic functions, since for the decades of development of the theory, a large number of very useful results have been accumulated. Different approaches to assessing safety of systems do not compete, but complement each other, so having some universality in describing safety requirements is a very valuable quality. In this article, we demonstrate the advisability of using the AADL modeling language and its extension Error Model Annex to describe safety requirements of a system under design. First, we describe a mathematical model of safety requirements expressible in AADL Error Model Annex. Next, we present algorithms to perform the following automated risk analysis on the base of AADL models: Fault Tree Analysis (including calculation of minimal cut sets and ranking of primary events with respect to different relevant importance measures), Failure Mode and Effects Analysis, and Markovian Analysis. At last, we consider an example of a real avionic system. We present an architecture of an AADL model of the system under design and describe how to develop Error Model Annex specifications for the model. With the help of risk analysis, we show how one can identify, localize and fix a bug in the architecture of the system on the design stage of the system development. All presented algorithms are implemented in MASIW framework for design of modern avionics systems.