Professional Documents
Culture Documents
1 Research
Laboratory of Automatic Signal and Image Processing, (LARATSI)
National School of Engineers of Monastir, Tunisia
2 Conservatoire National des Arts et Métiers (CNAM), Cedric - Laetitia
3 The Department of Electrical Engineering, Indian Institute of Technology (BHU)
Outline
1 Introduction
Switched systems
Interval Observers
Fault detection: An overview
2 Problem Formulation
3 Framer design
Augmented system
4 L∞ fault detection interval observer
5 Numerical Example
6 Conclusion
Switched systems
Introduction
Switched systems
Introduction
Switched systems
Switched systems
Interval Observers
Interval Observers
Interval Observers
Interval Observers
Definition
The dynamical system
(
ż(t) = α(z, y , u)
T T T
x , x
Interval Observers
An interval observer:
compute the set of admissible values.
provide the lower and upper bounds of state vector.
Fault diagnosis:
Fault detection: Detect malfunctions in real time, as soon and
as surely as possible.
Fault isolation: Find the root cause, by isolating the system
component(s) whose operation mode is not nominal
Fault identification: to estimate the size and type or nature of
the fault
Fault detectability: to define residuals that are affected by the
faults, i.e., residuals that permit to detect faults:
Residual generation
Residual evaluation: several approaches:
Fault diagnosis:
Fault detection: Detect malfunctions in real time, as soon and
as surely as possible.
Fault isolation: Find the root cause, by isolating the system
component(s) whose operation mode is not nominal
Fault identification: to estimate the size and type or nature of
the fault
Fault detectability: to define residuals that are affected by the
faults, i.e., residuals that permit to detect faults:
Residual generation
Residual evaluation: several approaches:
Fault diagnosis:
Fault detection: Detect malfunctions in real time, as soon and
as surely as possible.
Fault isolation: Find the root cause, by isolating the system
component(s) whose operation mode is not nominal
Fault identification: to estimate the size and type or nature of
the fault
Fault detectability: to define residuals that are affected by the
faults, i.e., residuals that permit to detect faults:
Residual generation
Residual evaluation: several approaches:
x 0 ≤ x(0) ≤ x 0 , ∀k ≥ 0,
ω ≤ ω(k) ≤ ω, ∀k ≥ 0,
v ≤ v (k) ≤ v , ∀k ≥ 0,
where x 0 , x 0 , ω ∈ Rn and v ∈ Rp are known vectors.
x(k): the upper estimate of x(k). y (k) the upper estimate of y (k).
x(k): the lower estimate of x(k). y (k) the lower estimate of y (k).
+ − − −
(5)
∆ = (Tσk Dσk ) ω − (Tσk Dσk ) ω + L+ +
σk + Nσk v − Lσk + Nσk v
+ − − −
∆ = (Tσk Dσk ) ω − (Tσk Dσk ) ω + L+ +
σk + Nσk v − Lσk + Nσk v
Theorem 1:
Let given assumptions hold, the lower bound x(k) and upper
bound x(k) for the state x(k) given by (5) satisfy in the fault free
case (f (k) = 0, ∀k ∈ Z+ )
Proof.
By calculating the upper and the lower estimate error(e and e)
dynamics, we get:
From the fact that e(0) = x(0) − x(0) ≥ 0 and e(0) = x(0) − x(0) ≥ 0,
then x(k) ≤ x(k) ≤ x(k) holds ∀k ∈ Z+ , if (Tσk Aσk − Lσk Cσk ) ≥ 0. This
ends the proof.
Augmented system
Θσk = I Lσk N σk
Augmented system
e(k)
ẽ(k) = Tσk Aσk − Lσk Cσk 0
e(k) Ãσk =
0 Tσk Aσk − Lσk Cσk
r (k)
r˜(k) = Θσ k 0
r (k) Θ̃σk =
0 Θ σk
Λ(k)
Λ̃(k) = −Tσk Fσk
Λ(k) F̃σk =
Tσk Fσk
v (k)
Cσ−k Cσ+k
I I −I
ṽ (k) = v (k) C̃σk = ,ν =
−Cσ+k −Cσ−k I −I I
v (k)
Theorem 2:
Assume that all assumptions of Theorem 1 hold. For given scalars
γ1 > 0, γ2 > 0 and 0 < α < 1, if there exist positive scalars
α2 > α1 > 0, ρ > 0, a positive definite matrix Pσk ∈ Rn×n and a
G1σk 0
diagonal matrix Gσk such that Gσk = ≻0,
0 G1σk
Wσk ∈ Rn×n , Rσk ∈ Rn×p and Hσk ∈ Rn×(n+p) such that
Υ1σk 0
Υσk = ≥0 (9)
0 Υ1σk
Theorem 2:
(α − 1) Pσk 0 ⋆
⋆ −ρI ⋆ ⪯0 (10)
Υσk Ωσk Pσk T
− Gσk − Gσk
Wσl Pσk
⪰0 (12)
Pσk Pσk
Theorem 2:
with
G1σk Rσ k κσk 0 0 0
Ωσk =
0 0 0 G1σk Rσ k κσk
where
I I 0
ξσk = , λ1 = , λ2 =
C̃σk+1 0 I
Theorem 2:
and
0.05 0.05 0.05 1
A1 = 0.05 0.1 0.05 , B1 = 0.5 ,
0.05 0.05 0.15 0.7
1 2
1 1 0
C1 = , D 1 = 2 , F1 = 1
1 0 1
1 0
0.238 −0.119 0.119 0.4
A2 = 0 0.476 0.238 , B2 = 1.6 ,
0 0 0.119
0
1 0
1.01 0 1
C2 = , D 2 = 0 , F2 = 1
1 1 1
1.73 2
0.238 −0.119 0.119 0.1
A3 = 0 .2 −0.5 , B3 = 0.0 ,
0 0 0.2
0.1
1 1
1 0 1
C3 = , D3 = 0.5 , F3 = 0
0 1 1
1 1
The disturbances w (k) and the measurement noises v (k) are uniformly
such that w ≤
bounded w (k) ≤ w and v ≤ v (k) ≤ v with
w = 0.1 0.1 0.1 , w = −0.1 −0.1 −0.1 ,
v= 0.02 0.02 and v = −0.02 −0.02 .
0.8437 −0.0667 −0.0896
T1 = 0.0704 1.0305 0.0399 ,
−0.0659 −0.0274 0.9615
0.0667 0.0896
N1 = −0.0305 −0.0399 ,
0.9901 0.0207 0.0108
0.0274 0.0385 T3 = −0.0051 0.9192 −0.0858 ,
−0.0209 0.0375 1.0166
0.0099 −0.0207
1.2925 0.3769 0.2933
T2 = −0.6751 0.1974 −0.6763 , N3 = 0.0051 0.0808 .
−0.0667 −0.0495 0.9334 0.0051 0.0808
0.0836 −0.3769
N2 = −0.1263 0.8026 ,
0.0171 0.0495
15
10
-5
-10
-15
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
15
10
-5
-10
-15
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
1.5
-0.5
-1
-1.5
-2
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
2
The residual signal
-1
-2
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
Figure 3: The residiual signals and estimated bounds in the fault-free case
Faulty case
1.2, 32 < k ≤ 75,
We set f (k) =
0, otherwise.
We get:
6
-1
-2
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
-2
10 20 30 40 50 60 70 80 90 100 110 120
Time(s)
The End