Test signal generation for service diagnosis based on local structural properties

Abstract

The paper presents a new approach to the generation of test signals used in service diagnosis. The tests make it possible to isolate faults, which are isolable only if the system is brought into specific operating points. The basis for the test signal selection is a structure graph that represents the couplings among the external and internal signals of the system and the fault signals. Graph-theoretic methods are used to identify edges that disappear under certain operating conditions and prevent a fault from changing the system behavior at this operating point. These operating conditions are identified by validuals, which are indicators obtained during the graph-theoretic analysis. The test generation method is illustrated by a process engineering example.

Certifying Robustness of Separating Inputs and Outputs in Active Fault Diagnosis for Uncertain Nonlinear Systems

“…FDI approaches have to cope with uncertain measurements (Blanke et al, 2006), which impede making conclusive statements. An important class of FDI methods are so-called active approaches in which auxiliary input signals are used to enhance the FDI performance (Niemann and Poulsen, 2005;Ungermann et al, 2012;Campbell and Nikoukhah, 2004;Zhang, 1989). Active diagnosis becomes important if faults are not detectable or isolable during nominal plant operation, in particular, if the reachable output sets for different faults and the nominal behavior overlap, or if the control actions mask the effects of faults.…”