Any technical system is liable to the occurrence of faults. A major fault in the plant like the complete loss of an actuator or sensor breaks the control loop and renders the system in-operational, unless the control structure is changed in response to the fault. This paper shows how the LQG (linear quadratic Gaussian) optimal control problem can be posed and solved automatically for the faulty plant, resulting in a new controller that allows to stabilise the plant despite of an actuator fault. The same approach can be used to design a bank of observers (one per fault case) that can be used for sensor fault diagnosis and reconfiguration at the same time. The approach is experimentally verified at a two-degrees-of-freedom helicopter model.
[1]
Thomas Steffen,et al.
Hybrid Reconfigurable Control
,
2002
.
[2]
Siep Weiland,et al.
H2 Optimal Control
,
2000
.
[3]
Uwe Mackenroth,et al.
H 2 Optimal Control
,
2004
.
[4]
P. Antsaklis,et al.
Stability of the pseudo-inverse method for reconfigurable control systems
,
1991
.
[5]
Jan M. Maciejowski,et al.
Predictive control : with constraints
,
2002
.
[6]
Mogens Blanke.
Fault-tolerant Control Systems
,
1999
.
[7]
Anuradha M. Annaswamy,et al.
Robust Adaptive Control
,
1984,
1984 American Control Conference.