Loading...

Application of Nonlinear System Identification Techniques in the Modeling of Agile Fighter Aircraft Dynamics

Roudbari, Alireza | 2015

612 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 47352 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Saghafi, Fariborz
  7. Abstract:
  8. Modeling and simulation are widely used as essential tools to predict and analyze complex systems in various scientific and engineering fields. For an aerospace system such as aircraft, mathematical models are useful used to carry out such tasks as dynamic analysis, autopilot development and combination of flight control laws, hazardous behaviors prediction, the controller design and validation, the study of the handling qualities, or the implementing simulators used for training pilots and many other tasks. In general, the aircraft flight dynamics is a nonlinear and coupled system whose dynamic modeling, in addition to pilot control inputs, depends on flight conditions (i.e., Mach and altitude) and limited to aerodynamic coefficients and derivatives. These kinds of data are not often accurate enough and their computations are often costly and even in some cases, unavailable. These models are usually linearized or are only valid in a limited domain around a specific (operational) point. Furthermore, when the degree of nonlinearity increases, the modeling process becomes even more difficult. The identification methods which work based on the measurement of the whole system input/output can serve as better and faster approaches for complicated systems such as aircrafts in order to obtain accurate models. Thus far, numerous system identification algorithms have been applied by researchers. Generally, the structures of the identification algorithms are classified as linear or nonlinear. Because linear models are only valid in a small range around the operational point, and most systems in the real world are nonlinear, when a wider range of the system’s dynamic behavior is of interest, the linear models may not be valid; thus, it is necessary to develop nonlinear mathematical models to capture the system’s nonlinear behavior. In the this thesis, a new approach has been proposed to model the coupled nonlinear six-degree-of-freedom dynamics of a highly maneuverable aircraft. For this purpose, three different types of nonlinear identification algorithms including nonlinear autoregressive with exogenous (NARX), block-oriented nonlinear systems (i.e., Hammerstein model, Wiener model, and the combination of Hammerstein and Wiener model) Intelligent techniques (i.e., artificial neural networks and fuzzy algorithms) have been used and compared. In order to develop a better training method for the ANNs, both normal and modified genetic algorithms(GAs) have been used and compared to train the network in this study. A modified genetic algorithm (MGA) has simultaneously been applied to ANNs at the three different levels of connection weights, structures, and learning rules. In order to validate the identified model, two types of data including experimental data from fourth-generation fighter aircraft and simulated data from the aircraft are used
  9. Keywords:
  10. System Identification ; Neural Networks ; Flight Test ; Aircraft Nonlinear Dynamic ; Block-Oriented Nonlinear Model ; Regression-Based Nonlinear Model

 Digital Object List

 Bookmark

No TOC