MODULAR TECHNIQUES FOR DYNAMIC FAULT-TREE ANALYSIS

Current approaches used to assess the dependability of complex systems such as Space Station Freedom and the Air Traffic Control System are incapable of handling the size and complexity of these highly integrated designs. A new technique for modeling such systems is described which is built upon current techniques in Markov theory and combinatorial analysis. It enables the development of a hierarchical representation of system behavior which is more flexible than either technique alone. A solution strategy is also described which is based on an object-oriented approach to model representation and evaluation. The technique is virtually transparent to the user since the fault tree models can be built graphically and the objects defined automatically. The tree modularization procedure allows the two model types, Markov and combinatoric, to coexist and does not require that the entire fault tree be translated to a Markov chain for evaluation. This effectively reduces the size of the Markov chain required and enables solutions with less truncation, making analysis of longer mission times possible. Using the FTPP (Fault Tolerant Parallel Processor) as an example, a model is built and solved for a specific mission scenario and the solution approach is illustrated in detail.