Robust Fault Tolerant Control Framework Using Uncertain Takagi-Sugeno Fuzzy Models - Fuzzy Modeling and Control: Theory and Applications - page 117

The robust TS framework presented in Sect. 6.2 can be used to deal with faults.
More precisely, depending on how much information is available about the fault and
the type of FTC technique to be used, the proposed control framework gives rise
to different FTC strategies. An advantage of the proposed robust TS framework is
that it allows to represent a wide spectrum of fault types, for example actuators and
process faults, affecting the matrices B and A , respectively. Moreover, different fault
dynamics (abrupt or incipient) can be represented through the time variance of the
parameters. Another advantage of the proposed framework is that, in contrast with
other FTC design methodologies that only allow to take into account finite sets of
faulty behaviors (for instance, a finite set of constant pairs ), the proposed

framework allows to specify intervals of fault magnitudes, that represent an infinite
set of faulty behaviors the FTC system has to deal with.

( A f , B f )

6.3.2 Passive FTC

In the passive FTC approach, it is assumed that no information about the faults
is available. Hence, tolerance against faults can only be achieved by considering
faults as if they were uncertainties. A single controller is designed in such a way
that it exhibits some robustness properties. More specifically, the faults f are

considered to be unknown parameters and a single set of subsystem controllers K i is
designed so as to be inferred by the premise variables

∈ F

ϑ(τ) and to be robust against

B i and faults. This strategy has the advantage of not
needing a fault detection, isolation and estimation (FDIE) algorithm but, on the other
hand, the controller has the highest possible conservativeness.

A i

model uncertainties and

6.3.3 Active FTC Without Controller Reconfiguration

The conservativeness of the passive approach can be overcome by considering that
some information available about the fault can be used to infer accordingly the
controller. More specifically, the faults f

are considered to be varying parameters
whose values are known or can be estimated through the information coming from
a fault estimation (FE) module. This information can be used to infer between a
single set of subsystem controllers K i that are designed to be robust against model
uncertainties. In this case, the controller is not reconfigured, as it is the same as the
one designed in the nominal case.

∈ F

6.3.4 Active FTC with Controller Reconfiguration
and Fault Detection

In this case, the faults f are considered to be uncertain parameters, but a fault

detection (FD) algorithm that can detect the fault occurrence at time instant T D is

∈ F