Dynamic modelling and control of a flexible manoeuvring system.

In this research a twin rotor multi-input multi-output system (TRMS), which is a
laboratory platform with 2 degrees of freedom (DOF) is considered. Although, the
TRMS does not fly, it has a striking similarity with a helicopter, such as system
nonlinearities and cross-coupled modes. Therefore, the TRMS can be perceived as
an unconventional and complex "air vehicle" that poses formidable challenges in
modelling, control design and analysis, and implementation. These issues constitute
the scope of this research.
Linear and nonlinear models for the vertical movement of the TRMS are
obtained via system identification techniques using black-box modelling. The
approach yields input-output models without a priori defined model structure or
specific parameter settings reflecting any physical attributes of the system. Firstly,
linear parametric models, characterising the TRMS in its hovering operation mode,
are obtained using the potential of recursive least squares (RLS) estimation and
genetic algorithms (GAs). Further, a nonlinear model using multi-layer perceptron
(MLP) neural networks (NNs) is obtained. Such a high fidelity nonlinear model is
often required for nonlinear system simulation studies and is commonly employed in
the aerospace industry. Both time and frequency domain analyses are utilised to
investigate and develop confidence in the models obtained. The frequency domain
verification method is a useful tool in the validation of extracted parametric models.
It allows high-fidelity verification of dynamic characteristics over a frequency range
of interest. The resulting models are utilized in designing controllers for low
frequency vibration suppression, development of suitable feedback control laws for
set-point tracking, and design of augmented feedforward and feedback control
schemes for both vibration suppression and set-point tracking performance. The
modelling approaches presented here are shown to be suitable for modelling
complex new generation air vehicles, whose flight mechanics are not well
understood.
Modelling of the TRMS revealed the presence of resonance modes, which are
responsible for inducing unwanted vibrations in the system. Command shaping
11
control strategies are developed to reduce motion and uneven mass induced
vibrations, produced by the main rotor during the vertical movement around the
lateral axis of the TRMS rig. 2-impulse, 3-impulse and 4-impulse sequence input
shapers and Iow-pass and band-stop digital filters are developed to shape the
command signals such that the resonance modes are not overly excited. The
effectiveness of this concept is then demonstrated in both simulation and real-time
experimental environments in terms of level of vibration reduction using power
spectral density profiles of the system response.
Combinations of intelligent and conventional techniques are commonly used
the control of complex dynamic systems. Such hybrid schemes have proved to be
efficient and can overcome the deficiencies of conventional and intelligent
controllers alone. The current study is confined to the development of two forms of
hybrid control schemes that combine fuzzy control and conventional PID
compensator for input tracking performance. The two hybrid control strategies
comprising conventional PO control plus PlO compensator and PO-type fuzzy
control plus PlO compensator are developed and implemented for set-point tracking
control of the vertical movement of the TRMS rig. It is observed that the hybrid
control schemes are superior to other feedback control strategies namely, PlO
compensator, pure PO-type and PI-type fuzzy controllers in terms of time domain
system behaviour.
This research also witnesses investigations into the development of an
augmented feedforward and feedback control scheme (AFFCS) for the control of
rigid body motion and vibration suppression of the TRMS. The main goal of this
framework is to satisfy performance objectives in terms of robust command tracking,
fast system response and minimum residual vibration. The developed control
strategies have been designed and implemented within both simulation and real-time
environments of the TRMS rig. The employed control strategies are shown to
demonstrate acceptable performances. The obtained results show that much
improved tracking is achieved on positive and negative cycles of the reference
signal, as compared to that without any control action. The system performance with
the feedback controller is significantly improved when the feedforward control
component is added. This leads to the conclusion that augmenting feedback control
with feedforward method can lead to more practical and accurate control of flexible
systems such as the TRMS.