Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Modularisation of State/Event Fault Trees

Design and Implementation of a Structure-Conserving Approach in Support of Hybrid, DSPN- and BDD-Based Evaluation
: Förster, M.
: Polze, A.; Liggesmeyer, P.

Kaiserslautern, 2006, 99 pp. : Ill., Lit.
Potsdam, Univ., Master Thesis, 2006
Master Thesis
Fraunhofer IESE ()

Safety and dependability analyses in industry are customarily performed using fault tree analysis (FTA), a combination of Boolean logic, dependability theory and stochastics. Many engineers today are familiar with FTA, and it has become industry standard because it is easy to grasp and at the same time can be handled efficiently for its strictly mathematical foundations. A drawback of standard FTA is that it operates on a purely combinatorial system model and cannot capture complex failure behaviour, which is becoming a more important aspect of real-world systems as their complexity rises and the more they are software controlled. Addressing this issue, more powerful modelling and evaluation techniques have been proposed. One example are state/event fault trees (SEFTs), a new, state- and component-based formalism for dependability assessment of software-controlled systems. Like other state-based models, SEFTs suffer from the fact that the state space to be handled during evaluation grows exponentially in the number of system components. This thesis proposes an effort-saving modularisation approach to SEFT evaluation that partitions model state space into smaller units, the measures of which can be calculated separately.
Subsequently, measures are integrated by combinatorial, less computationally expensive means. For this purpose, the developed algorithm transforms the SEFT source model, while preserving structural properties that may be exploited to facilitate further state space reduction. For validation of the presented ideas, the algorithm has been implemented in the ESSaRel FTA framework. Its correctness, effectiveness and efficiency are demonstrated, and possible further developments are discussed.