 
Summary: The Center for Control, Dynamical Systems, and Computation
Spring Seminars
Presents
Theory and Practice of Fault Detection
and Identification
by
Professor Jason Speyer
Mechanical and Aerospace Department
University of California, Los Angeles
Friday, April 28th, 2006 3:00  4:00 PM Engineering II Pavilion
Abstract:
The restricted diagonal fault detection filter was first developed from a geometric and spectral approach. The
idea of the filter is to place each fault into invariant subspaces which do not overlap each other. Then, when a
nonzero residual is detected, the fault can be announced and identified by projecting the residual onto each of
the invariant subspaces. In this way, multiple faults can be monitored in one filter. Strict geometric structures can
be sensitive to system uncertainties. Therefore, derivations based on minimizing a cost criterion are explored
which can be shown to acquire the properties of the geometric fault detection filter in an appropriate limit. One
approach, the unknown input observer, simplifies the restricted diagonal fault detection filter problem by dividing
the faults into two groups: a single target fault and possibly several nuisance faults. The nuisance faults are
placed in an approximate invariant subspace that is unobservable to the residual. Therefore, the residual
