No description available.

No description available.

No description available.

The recent Generalised Predictive Control algorithm (Clarke et al, 1984,87) is a self-tuning/ adaptive control algorithm that is based upon long-range prediction, and is thus claimed to be particularly suitable for process control application. The complicated nature of GPC prevents the application of standard analytical techniques. Therefore an alternative technique is developed where an equivalent closed loop expression is repeatedly calculated for various control scenarios. The properties of GPC are investigated and, in particular, it is shown that 'default' values for GPC's design parameters give a mean-level type of control law that can reasonably be expected to provide robust control for a wide variety of processes. Two successful industrial applications of GPC are then reported. The first series of trials involve the SISO control of soap moisture for a full-scale drying process. After a brief period of PRBS assisted self-tuning default GPC control performance is shown to be significantly better than the existing manual control, despite the presence of a large time-delay, poor measurements and severe production restrictions. The second application concerns the MIMO inner loop control of a spray drying tower using two types of GPC controller: full multivariable MGPC, and multi-loop DGPC. Again after only a brief period of PRBS assisted self-tuning both provide dramatically superior control compared to the existing multi-loop gain-scheduled PID control scheme. In particular the use of MGPC successfully avoids any requirement for a priori knowledge of the process time-delay structure or input-output pairing. The decoupling performance of MGPC is improved by scaling and that of DGPC by the use of feed-forward. The practical effectiveness of GPC's design parameters (e.g. P, T and λ) is also demonstrated. On the estimation side of adaptive control the current state-of-the-art algorithms are reviewed and shown to suffer from problems such as 'blowup', parameter drift and sensitivity to unmeasurable load disturbances. To overcome these problems two novel estimation algorithms (CLS and DLS) are developed that extend the RLS cost-function to include weighting of estimated parameters. The exploitation of the 'fault detection' properties of CLS is proposed as a more realistic estimation philosophy for adaptive control than the 'continuous retention of adaptivity'.

This thesis reports the successful application of the recently introduced Generalised Predictive Control self-tuner to the high-performance positioning of a real flexible single-link robot arm. The large amount of experimental time available on this high bandwidth system allowed exhaustive testing of the 'tuning-knobs' and 'design-filters' available to the user for tailoring the closed-loop. Based upon these experiments a coherent philosophy for configuring GPC in practice is generated. Configuration details found to be necessary for satisfactory GPC control of this high-order neutrally stable and non-minimum-phase plant, with its lightly damped resonant modes, are isolated. In particular it is found that band-pass filtering of data is essential for stable offset-free control using finite-order models of the plant. These aspects are considered in detail both theoretically and experimentally. In this application, as is often the case in practice, some information about the plant dynamics is available beforehand. Novel methods for the inclusion of this prior knowledge are introduced and their beneficial effects on the convergence of the recursive least squares estimation scheme, upon which most self-tuners are based, are demonstrated in simulation and experiment. A novel high-speed 68010/20 multi-processor computer system is described which allows the implementation of GPC at the required high sample rate (60Hz). The software division of the self-tuning algorithm into concurrently and sequentially executing tasks is discussed in detail.

This thesis is concerned with suboptimal adaptive control of discrete linear stochastic processes whose parameters are unknown. The suboptimal adaptive controllers considered are (i) Open Loop Feedback Optimal (OLFO) controller, (ii) self-tuning controller, and (iii) optimal k step ahead controller. Two more controllers, certainty about parameter (CAP) controller and no learning (NOL) controller, that provide bounds on the performance of these adaptive controllers are also considered. Performance of these controllers have been evaluated for a first order process through monte-carlo simulations. Simulation of OLFO controller together with the bounding controllers for the first order process when there is only one unknown parameter revealed that OLFO controller is unsuitable to control unstable processes and would be an unwise choice even for controlling stable processes. Selftuning and OK controllers have been simulated for the first order process with all the parameters unknown. Three cases for the unknown parameters have been considered. They are: (i) constant unknown parameters (ii) slowly time-varying unknown parameters and (iii) rapidly time-varying unknown parameters. Simulation results showed that in certain regions of the unknown parameter space the cost produced by self tuning controller and OK controller are very similar, in certain regions the OK controller produces lesser cost than the self-tuning controller and in certain other regions both controllers perform very badly. But self-tuning controller always took only half as much computing time as OK controller. A necessary condition for convergence of OK controller to a linear constant parameter controller having the same functional form as CAP controller is found out using the ideas of uniform complete observability. For a first order process under OK controller the only occasion the condition would be violated is when there is 'turn-off'. Finally, it is shown that using the combined state/parameter estimator in the place of extended Kalman filter the computational requirement of OK controller can be reduced. For the first order process, OK controller with the combined estimator took only sixty percent as much computing time as the OK controller with extended Kalman filter without any appreciable deterioration in the performance.

In classical model reference adaptive control, the goal is to design a controller to make the closed-loop system act like a prespecified stable reference model. A recent approach yields a linear periodic controller which simultaneously performs probing, estimation, and control. This linear controller is not only able to handle time-varying systems, but also provides exponential stability. In addition, from simulations, it is found that the controller has excellent noise rejection in certain cases. In this thesis, we used the induced noise gain as the measurement of noise rejection. For plants that are minimum phase with relative degree one, we started with the case where the plant is first order and linear time-invariant. Then we moved to the case where the plant is first order and linear time-varying. Finally, we extended to the general case where the plant is linear time-varying with relative degree one. For the above cases, we quantitatively investigated how certain control parameters affect the induced noise gain.

Thesis (MEng)--University of South Australia, 1997

Thesis (Ph.D.) -- McMaster University, 2002. / Includes bibliographical references (leaves 246-253). Also available via World Wide Web.

Thesis (Ph. D.)--Virginia Polytechnic Institute and State University, 1992. / Vita. Abstract. Includes bibliographical references (leaves 144-148). Also available via the Internet.