Safety Analysis of Computer based system using Stochastic Modeling

Abstract

A safety-critical system executes the critical tasks, the failure of which may jeopardize human life, lead to considerable nancial misfortune, or cause extensive environmental damage. Therefore, safety is considered as one of the most critical areas of research, while dealing with safety-critical systems. Traditional safety-critical systems are being converted into digital systems for several bene ts. To support various features of a safety-critical system, the digital systems are becoming more complex in functional behaviors. Complex digital systems comprised of many software and hardware components mostly of them are heterogeneous in nature. A safety-critical system comprises of a large no. of heterogeneous components, there is higher risk always associated with the safety-critical systems due to possible failures in Hardware/Software involved there in. Several techniques are available to perform the safety analysis of such systems. An extensive literature survey was carried out to identify the available methods for safety analysis of safety-critical digital systems. Most of them works on qualitative assessment rather than quantitative assessment. However, quantitative assessment has several bene ts over qualitative assessment such as { (i) risks are sorted by their adversity impact, and (ii) security levels can be better determined/de ned through consideration of three elements that are availability, integrity, and con dentiality. Further, safety analysis during the early phases of system development life cycle has many signi cant bene ts such as { (i) help in taking decisions to select most suitable design (ii) cost minimization (iii) analyzing the sensitivity of the system safety to its component safety (iv)identify safety bottlenecks. The proposed work deals with a new probabilistic approach to quantify safety of safety-critical systems during the design phase of the systems that is based on the probabilistic safety assessment to deal with the shortcomings of the existing techniques using state-space models. Further, it is a challenging task to capture all the requirements including safety requirements through state space models. Also, verifying that the constructed model has captured all the requirements is again a problem in itself, because of all the understanding of stakeholders may not get captured during development of the state-space model. Failing to model all the requirements will give inaccurate safety assessment. UML is a well-known and successful way of modelling which is used for specifying requirements. UML can capture all the requirements and be easily understood by all the stakeholders. This motivates us to propose a methodology to convert the UML model into a state space model that can be used for quantitative assessment of a safety-critical systems under consideration. A framework is proposed and introduced in this thesis to transform UML model into a state-space model in the form of a Petri net, which is a reliable graphical and mathematical tool to perform several static and dynamic analysis. All the above proposed approaches are validated considering a real-time safety-critical system of Nuclear Power Plant along with some noticeable ndings.