Probabilistic Performance Analysis of Fault Diagnosis Schemes
Probabilistic Performance Analysis of Fault Diagnosis Schemes
Probabilistic Performance Analysis of Fault Diagnosis Schemes
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
the performance metrics, a comprehensive performance analysis would consider all<br />
possible values <strong>of</strong> u, which is clearly not feasible. One reasonable compromise is to<br />
compute the worst-case performance over a specified family <strong>of</strong> inputs. To this end,<br />
we consider families <strong>of</strong> inputs that have the following form:<br />
B n u<br />
p (u ◦ ,γ) = { u + u ◦ ∈ S n u<br />
: ‖u‖ p < γ } ,<br />
where u ◦ ∈ l n u<br />
p is a fixed nominal input, p ∈ [1,∞] specifies the l p -norm, and γ > 0 is<br />
the desired bound.<br />
2. Bounded Disturbances: Thus far, we have assumed that the system G θ is affected by<br />
a noise signal v. It is also useful to consider the case where a deterministic signal w,<br />
called a disturbance, affects the system in such a way that the fault diagnosis scheme<br />
cannot use w to generate a residual. We consider disturbances in the bounded set<br />
B n w<br />
p (0,γ) = { w ∈ S n w<br />
: ‖w‖ p < γ } ,<br />
where p ∈ [1,∞] specifies the l p -norm, and γ > 0 is the desired bound.<br />
3. Uncertain <strong>Fault</strong> Signals: In Chapters 3 and 4, it is assumed that the fault signal f k<br />
at time k is a known, fixed function <strong>of</strong> the fault parameter sequence θ 0:k . While this<br />
approach may work for certain types <strong>of</strong> faults, it <strong>of</strong>ten useful to consider the worst-case<br />
performance <strong>of</strong> a fault diagnosis scheme over a set <strong>of</strong> possible fault signals. Hence, for<br />
a given parameter sequence ϑ, we assume the fault signal lies in a bounded set <strong>of</strong> the<br />
form<br />
B n f<br />
p<br />
(<br />
f (ϑ) ◦ ,γ ) = { f + f ◦ (ϑ) ∈ S n f<br />
: ‖f ‖ p < γ } ,<br />
where f ◦ (ϑ) ∈ l n f<br />
p is the nominal value <strong>of</strong> the fault signal, p ∈ [1,∞] specifies the<br />
l p -norm, and γ > 0 is the desired bound.<br />
4. Model Uncertainty: In model-based fault diagnosis schemes, the residual generator<br />
is usually designed according to the nominal system model G 0 . However, it useful to<br />
consider cases where G 0 does not perfectly model the system or the designer <strong>of</strong> the<br />
residual generator does not have accurate knowledge <strong>of</strong> the true model. Both <strong>of</strong> these<br />
cases are addressed by assuming that the parameterized system G θ is uncertain. In<br />
particular, we assume that the system consists <strong>of</strong> an interconnection <strong>of</strong> the system G θ<br />
and an uncertain operator ∆. We consider two classes <strong>of</strong> uncertain operators. First,<br />
we consider the class norm-bounded linear time-invariant uncertainties<br />
∆ 2,lti (γ) := { ∆ ∈ ∆ 2 (γ) : ∆ is lti, causal, stable } ,<br />
where γ > 0 is the desired bound. Second, we consider the class <strong>of</strong> norm-bounded<br />
79