Abstract—In this paper, a fast-convergent fault detection and
isolation architecture is proposed for linear MIMO continuoustime
systems. By exploiting a system decomposition technique
and making use of kernel-based deadbeat estimators, the
state variables can be estimated in a non-asymptotic way.
Estimation residuals are then defined to detect the occurrence
of a fault and identify the occurring fault function after fault
detection. In the noisy scenario, thresholds are defined for the
residual to distinguish the effect of the noise from that of
the fault. Numerical examples are included to characterize the
effectiveness of the proposed FDI architecture
