Two self-test methods applied to an inertial system problem

A survey of design methods for failure detection in dynamic systems

“…Recently a technique involving a bank of observers has been devised, and a successful application to a hydrofoil sensor failure problem is reported by Clark et al in [34]. Also, Wiiisky, Deyst, and Crawford [ 15], [ 16] have applied the methodology devised by Buxhaum and Haddad in [14] to study failure detection for an inertial navigation problem. We will briefly describe this technique to illustrate some of the concepts underlying the bank of filters approach.…”

“…We will briefly describe this technique to illustrate some of the concepts underlying the bank of filters approach. We also refer the reader to Wernersson [42] for a technique that is similar to that discussed in [16].…”

“…We note that the techniques devised in [10]-- [12] do not involve growing banks of filters, since the number of hypothesized models do not grow in time. However, it is possible for all of the filters in the bank to become oblivious, and thus shifts between the hypotheses may go undetected (see [16], [36] for examples). The technique of periodic fusing of the densities and initiation of new bank effectively avoids this problem, as would designing the original bank using age-weighted filtering techniques.…”

“…(27) and Note that the implementation of (26) requires an exponentially growing bank of filters: there are (r + i)* terms in (26). To avoid this problem a number of approximation techniques have been proposed [14]- [16]. The one used in [16] involves hypothesizing shifts only once every N steps.…”

“…To avoid this problem a number of approximation techniques have been proposed [14]- [16]. The one used in [16] involves hypothesizing shifts only once every N steps. At the end of each N step period we "fuse" the (r + !)…”