Comparing two methods for diagnosis of imprecise dynamic systems

We compare two methods for diagnosing faults using imprecise dynamic models and imprecise measurements. By imprecise we mean that a parameter or measurement is only known to an interval rather than a point on the real number line. The first uses 'Temporal Band Sequences' (TBS - due to [loi96]) to encode the measurements and parameters - it checks the consistency of these sequences against the model using symbolic integration. The second uses a numerically-based interval simulator. The interval behaviour predicted by the simulator is simply compared against the interval description of measured output. Although the simulation-based method suffers from numerical interval integration inaccuracies, it still outperforms the consistency checking method because it exploits the interrelations among the state variables and parameters. The simulation method is, however, computationally expensive and is not complete due to the sampling.