Abstract
This thesis describes the design, development, and flight-simulation testing of an
integrated Artificial Immune System (AIS) for detection, identification, and evaluation of a
wide variety of sensor, actuator, propulsion, and structural failures/damages including the
prediction of the achievable states and other limitations on performance and handling
qualities. The AIS scheme achieves high detection rate and low number of false alarms for all
the failure categories considered. Data collected using a motion-based flight simulator are
used to define the self for an extended sub-region of the flight envelope. The NASA IFCS F15
research aircraft model is used and represents a supersonic fighter which include model
following adaptive control laws based on non-linear dynamic inversion and artificial neural
network augmentation. The flight simulation tests are designed to analyze and demonstrate
the performance of the immunity-based aircraft failure detection, identification and
evaluation (FDIE) scheme. A general robustness analysis is also presented by determining
the achievable limits for a desired performance in the presence of atmospheric
perturbations.
For the purpose of this work, the integrated AIS scheme is implemented based on
three main components. The first component performs the detection when one of the
considered failures is present in the system. The second component consists in the
identification of the failure category and the classification according to the failed element.
During the third phase a general evaluation of the failure is performed with the estimation of
the magnitude/severity of the failure and the prediction of its effect on reducing the flight
envelope of the aircraft system.
Solutions and alternatives to specific design issues of the AIS scheme, such as data
clustering and empty space optimization, data fusion and duplication removal, definition of
features, dimensionality reduction, and selection of cluster/detector shape are also analyzed
in this thesis. They showed to have an important effect on detection performance and are a
critical aspect when designing the configuration of the AIS.
The results presented in this thesis show that the AIS paradigm addresses directly the
complexity and multi-dimensionality associated with a damaged aircraft dynamic response
and provides the tools necessary for a comprehensive/integrated solution to the FDIE
problem. Excellent detection, identification, and evaluation performance has been recorded
for all types of failures considered. The implementation of the proposed AIS-based scheme
can potentially have a significant impact on the safety of aircraft operation. The output
information obtained from the scheme will be useful to increase pilot situational awareness
and determine automated compensation.
Original language | American English |
---|---|
Qualification | Ph.D. |
Awarding Institution |
|
Supervisors/Advisors |
|
State | Published - Jan 1 2009 |
Keywords
- Artificial Immune Systems
- Artificial Intelligence
- Fault Tolerant Control Systems
- Failure Detection
- Identification and Evaluation
Disciplines
- Aeronautical Vehicles
- Systems Engineering and Multidisciplinary Design Optimization