Dec 10, 2018; Colloquium
Model-based Dependability Analysis for Cyber-Physical Systems
Technische Universität Dresden
Institut für Automatisierungstechnik
01069 Dresden
Abstract:
Modern industrial standards demand high system dependability and resilience. Classical quantitative methods, such as event trees, reliability block diagrams, fault trees, and Markov chain models are recommended for the evaluation of reliability and safety properties. However, the proper application of these methods to complex Cyber-Physical Systems (CPS) with distributed and networked software and hardware parts require thorough comprehension of both dependability properties of components, their interaction as well as structural and behavioral aspects of the complete system. Model-based System Engineering (MBSE) approaches help both to simplify and speed up the development process and provide semi-formal data for earlier dependability analysis. My presentation will cover the latest research results of our team with the key focus on stochastic error propagation analysis of heterogeneous CPS. We develop methods and tools based on probabilistic model checking and statistical simulations. A lot of effort is put into optimizations against state space explosion of underlying formal models and automatic application of our methods using transformation algorithms from common semi-formal MBSE models such as Simulink/Stateflow, UML/SysML, and AADL. The presentation starts with an overview of common MBSE paradigms and state-of-the-art reliability and safety evaluation methods. The main part of the talk addresses the technical details of our analytical methods. Finally, several relevant case studies will be discussed.