In important safety–critical systems, such as aircraft or spacecraft applications, malfunctions affecting actuators, sensors or other system components may lead to unsatisfactory performance, or even instability. To overcome these drawbacks, Fault Tolerant Control (FTC) systems have been developed in order to tolerate component malfunctions, while maintaining desirable stability, and satisfactory performances. In general, FTC methods are classified into two types, i.e. Passive Fault Tolerant Control (PFTC), and Active Fault Tolerant Control (AFTC) [9, 3]. In PFTC systems, controllers are fixed, and designed to be robust against a class of presumed faults, thus resulting in limited fault-tolerant capabilities In contrast to (PFTC), AFTC systems react to the faults actively by reconfiguring the control actions, maintaining acceptable performances Moreover, AFTC schemes rely heavily on real–time Fault Detection and Diagnosis (FDD), i.e. fault detection, isolation and estimation too. Usually, this information can be used from a logic–based switching controller or a feedback of the fault estimate. The approach proposed in this paper relies on the latter strategy. Over the last three decades many FDD techniques have been developed, see the survey works [12, 8, 2, 13]. Unfortunately, disturbance affecting the system can cause false alarms or, even worse, missed faults. Disturbance decoupling in FDD is therefore very important [7, 3]. This paper presents two novel aerospace AFTC systems based on differential geometry: in particular the FDD mudule is based on NLGA ([4, 6]. The FDD’s filters structure is derived using the coordinate change of the NLGA theory [11]: the application of the NLGA is investigated in order to obtain actuator fault estimates decoupled from aerodynamic disturbance. The actuator fault estimation is accomplished by Adaptive Filters (AF) structurally designed thanks to NLGA, hence denoted with NLGA–AFs, and exploiting Recursive Least Squares or Radial Basis Function Neural Network (RBF-NN).
Issues of Fault Diagnosis and Fault Tolerant Control for Aerospace System
SIMANI, Silvio
2015
Abstract
In important safety–critical systems, such as aircraft or spacecraft applications, malfunctions affecting actuators, sensors or other system components may lead to unsatisfactory performance, or even instability. To overcome these drawbacks, Fault Tolerant Control (FTC) systems have been developed in order to tolerate component malfunctions, while maintaining desirable stability, and satisfactory performances. In general, FTC methods are classified into two types, i.e. Passive Fault Tolerant Control (PFTC), and Active Fault Tolerant Control (AFTC) [9, 3]. In PFTC systems, controllers are fixed, and designed to be robust against a class of presumed faults, thus resulting in limited fault-tolerant capabilities In contrast to (PFTC), AFTC systems react to the faults actively by reconfiguring the control actions, maintaining acceptable performances Moreover, AFTC schemes rely heavily on real–time Fault Detection and Diagnosis (FDD), i.e. fault detection, isolation and estimation too. Usually, this information can be used from a logic–based switching controller or a feedback of the fault estimate. The approach proposed in this paper relies on the latter strategy. Over the last three decades many FDD techniques have been developed, see the survey works [12, 8, 2, 13]. Unfortunately, disturbance affecting the system can cause false alarms or, even worse, missed faults. Disturbance decoupling in FDD is therefore very important [7, 3]. This paper presents two novel aerospace AFTC systems based on differential geometry: in particular the FDD mudule is based on NLGA ([4, 6]. The FDD’s filters structure is derived using the coordinate change of the NLGA theory [11]: the application of the NLGA is investigated in order to obtain actuator fault estimates decoupled from aerodynamic disturbance. The actuator fault estimation is accomplished by Adaptive Filters (AF) structurally designed thanks to NLGA, hence denoted with NLGA–AFs, and exploiting Recursive Least Squares or Radial Basis Function Neural Network (RBF-NN).I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
