Date of Award
Master of Engineering (Research)
Dr. Tom O'Mahony
One of the most popular approaches for dealing with process that undergo large variations with time, of late, is the multiple model approach. This method has considerable intuitive appeal and has received widespread dissemination in academia. Numerous successful simulations have been reported although few real-time results have been published. The principal contribution of this thesis is the application of the multiple model control scheme in real-time.
A number of practical concepts that enable the use of multiple model control in real-time are discussed. Many of these issues arose when the multiple model control scheme was applied to a laboratory scale system, namely a flexible link, and do not appear to be well documented in the current literature. Firstly, the issue of computational burden is addressed and a recursive cost function is proposed that eliminates unnecessary computation. Secondly, the impact of an actuator non-linearity, namely a dead-zone, is investigated. Finally, the issue of disturbances and their effect on the multiple model control scheme is considered. Appropriate filtering can be incorporated to negate the effect of disturbances and the dead-zone non-linearity. Both simulations and real-time experiments are conducted to examine the effect of these solutions.
Also, the closed-loop performance of the multiple model control scheme is analysed. The evaluation is performed on a flexible link which has inherent difficulties, namely, the system i) is integrating ii) has widely varying dynamics due to load changes iii) includes an actuator non-linearity (dead-zone) iv) is subjected to large deterministic disturbances. Identification, using the impulse response technique overcomes the first difficulty, while point ii) is surmounted through the identification of a number of local models and the design of a suitable supervisor. A pair of filters is introduced to ameliorate the effect of both dead-zones and deterministic disturbances. Finally, the performance of the multiple model control scheme is compared to the case where a single, robust controller is applied to the process and it is shown that for more difficult control problems the multiple model control approach outperforms the robust approach.
Cunningham, Peter, "Multiple Model Control with Intelligent Switching to Ensure Robustness to Process Variations." (2003). Theses [online].
Available at: https://sword.cit.ie/allthe/169
Submitted to the Higher Education & Training Awards Council.