Practical applications of timed failure propagation graphs for vehicle diagnosis

This paper describes how a timed failure propagation graph (TFPG) based reasoner can be used to optimally diagnose failures in a complex real time system, such as an aerospace vehicle. The TFPG approach is a specific instantiation of model based diagnostic reasoning which has been well documented (Abdelwahed et al., 2004). The authors have been collaborating over the last several years to identify the practical issues that are associated with the application of the TFPG reasoner for real time fault isolation in a large scale, dynamically changing environment; the resulting modifications to the reasoner have also been previously documented (Karsai et al., 2006) (Abdelwahed & Karsai, 2007). The focus of this paper is to illustrate, using a notional but representative fuel system case study, how the TFPG reasoner can efficiently and effectively diagnose both single and multiple failure scenarios. Using actual results obtained for this example, we will explore the practical implementation challenges that would typically be encountered when building a diagnostic reasoner implementation, and then demonstrate the TFPG features that have been implemented to address these issues.