Model-based fault diagnosis for aerospace systems: a survey:

Question

1. What factors should be taken into account during design?

2. What are the contributions in "Model-based fault diagnosis for aerospace systems: a survey" ?

3. How is the extension to non-linear systems addressed?

4. What are the types of observers used for generating residuals?

Accepted Answer

The exis-tence of measurement noise, model uncertainty, andunmodelled exogenous disturbances should be takeninto account during design.

Accepted Answer

This survey of model-based fault diagnosis focuses on those methods that are applicable to aerospace systems.. To highlight the characteristics of aerospace models, generic nonlinear dynamical modelling from flight mechanics is recalled and a unifying representation of sensor and actuator faults is presented.. The approaches that best suit the constraints of the field are emphasized and recommendations for future developments in in-flight fault diagnosis are provided.

Accepted Answer

Extension to non-linear systems is commonlyaddressed through linearization around the currentmean estimate within the extended Kalman filter(EKF) [250].

Accepted Answer

Two types of banks of observershave been designed for generating residuals that aresensitive to desired sets of faults, namely the dedi-cated observer scheme (DOS) and the generalizedobserver scheme (GOS) [73].

Accepted Answer

the design difficulty movesfrom choosing the parameters of the analytical sepa-rator to the selection of an activation function and thechoice of the structure of the network, i.e. the numberof layers and the number of neurons composing eachof them.

Accepted Answer

Norm-based methods [12, 283–285] aimat maximizing the effect of faults on residuals according to the H1 norm (maximum induced gain of the transfer matrix from faults to residuals), while mini-mizing a measure of the influence of disturbances(minimum induced gain of the transfer matrix fromdisturbances to residuals).

Accepted Answer

Non-Linear state estimation is often addressed bylinearizing the model around an operating point oralong a trajectory, in order to apply the previous tech-niques.

Accepted Answer

To allow autono-mous aerial vehicles to continue their missions, thereis an absolute necessity to identify unexpectedchanges (faults) in the system before they lead to acomplete breakdown (failure).

Accepted Answer

To validate thesenew approaches, there is the need for actual flighttests, since most work is done in simulation (admit-tedly with increasing realism).

Accepted Answer

After residuals – presumably noisy and disturbed –have been generated by methods from section 4,there is the need to analyse them on-line to provideBoolean decisions on whether they significantly differfrom those that would be generated during normaloperation.

Accepted Answer

These tasks generally involvethe generation of residuals, which are fault indica-tors based on deviation between measurementsand model-based computations.

Accepted Answer

Combining sensors from Table 1 may even allow theentire state vector to be observed, which makes moreanalytical redundancy available.

Accepted Answer

a new form of non-linear observers hasbeen proposed [233, 234], which is designed via thesolving of a partial differential equation.

Accepted Answer

An interesting idea in this context is active faultdiagnosis, where an auxiliary input may be injec-ted into the system to enhance fault identification[295–297].

Accepted Answer

There is, therefore, the need to call uponanalytical redundancy, i.e. to exploit mathematicalrelations between measured or estimated variablesin order to detect possible dysfunctions.

Model-based fault diagnosis for aerospace systems: a survey:

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What factors should be taken into account during design?

2. What are the contributions in "Model-based fault diagnosis for aerospace systems: a survey" ?

3. How is the extension to non-linear systems addressed?

4. What are the types of observers used for generating residuals?

5. What is the design difficulty of a neural network?

6. What is the way to decouple the effects of faults from disturbances?

7. How is the non-linear state estimation addressed?

8. What is the importance of identifying unexpected faults in the system?

9. What is the need for actual flighttests?

10. What is the need to analyse the residuals?

11. What are the tasks that are usually called fault detection and diagnosis?

12. What is the effect of combining sensors from Table 1?

13. How is the non-linear observer scheme designed?

14. What is the way to improve the detection of faults?

15. What is the need to use to detect possible dysfunctions in the system?

Figures

Citations

Generalized pseudo-Bayesian estimator for networked control systems over UDP-like channels

Fin failure diagnosis for non-linear supersonic air vehicle based on inertial sensors

Control Optimization of a Novel eVTOL/Pusher Transport Aircraft with Known Faults

Performance estimation of a FDI function for Flight Critical Systems: Application to a Sensor Acquisition System

Modern Spacecraft GNC

References

A New Approach to Linear Filtering and Prediction Problems

A New Approach to Linear Filtering and Prediction Problems

Fuzzy identification of systems and its applications to modeling and control

Anomaly detection: A survey

Statistical pattern recognition: a review

Related Papers (5)

A Review of Process Fault Detection and Diagnosis Part I : Quantitative Model-Based Methods

Model-based fault-detection and diagnosis - status and applications §

Bibliographical review on reconfigurable fault-tolerant control systems

Model-based Fault Diagnosis Techniques: Design Schemes, Algorithms, and Tools

A Survey of Fault Diagnosis and Fault-Tolerant Techniques—Part I: Fault Diagnosis With Model-Based and Signal-Based Approaches