A Model-Based Framework for Analyzing the Safety of System Architectures
Panagiotis Manolios, Kit Siu, Michael Noorman and Hongwei Liao.
RAMS, 2019 © IEEE
Abstract
We introduce a compositional, model-based framework for modeling, visualizing and analyzing the safety of system architectures for safety-critical cyber-physical systems. Our work provides a unified, end-to-end, framework that encompasses high-level models, fault trees and qualitative and quantitative safety analyses in one semantically coherent framework. Our framework enables the rapid development, modification and evaluation of architectures for complex systems.
Our framework includes a modeling language for defining libraries of component models that include information on component reliability, connectivity and fault propagation. System architectures consist of a sequence of component instantiations, component connections, and the identification of top-level faults. Our framework includes algorithms for automatically synthesizing and reducing fault trees from architectures and library models. The generated fault trees are then automatically analyzed to determine cutsets and the probability of top-level faults. Finally, our framework includes visualization algorithms that depict fault trees and architectures at various levels of abstraction. We provide a case study of a model inspired by the Boeing 777 IMA architecture.
The framework is compositional because safety engineers only need to
define reliability and fault propagation aspects at the component
level. This is in contrast with current methods used in the field of
avionics, where safety engineers directly construct system-level fault
trees. Defining such fault trees requires significant expertise, time
and care. Small changes to architectures can result in significant
changes to fault trees. All of this makes analyzing a collection of
architectures error- prone and prohibitive both in terms of time and
money. We developed an open source tool that implements our framework,
and provide an experimental evaluation consisting of the modeling and
analysis of a collection of architectures. Our model-based framework
provides a new paradigm, allowing significant automation in the area
of safety analysis of architectures for complex avionics systems.
PDF (294K) © IEEE