Logo image
Back
A multiple model, state feedback strategy for robust control of non-linear processes
Journal article   Peer reviewed

A multiple model, state feedback strategy for robust control of non-linear processes

F.Y. Wang, P. Bahri, P.L. Lee and I.T. Cameron
European Symposium on Computer-Aided Process Engineering-15, 38th European Symposium of the Working Party on Computer Aided Process Engineering, Vol.20, pp.1111-1116
2005
DOI: https://doi.org/10.1016/S1570-7946(05)80027-6
url
Link to Published Version *Subscription may be requiredView

Abstract

In order to achieve global stability using well-established linear control theory and techniques, a multiple model approach has attracted increased attention in recent years. In our previous work, a mini-max optimisation strategy was developed within the framework of a multiple model approach, in which a global controller can be designed without the requirement of membership/validity functions used in conventional methods, and the regime division was realised using a gap metric method. The major limitation of the reported methods is that robustness against process/controller disturbances cannot be addressed if the process switches from stable to unstable in operation. Furthermore, the number of local models is still large for highly non-linear processes even though the gap-metric method is incorporated. In this paper, a signficantly modified multiple model approach is developed to achieve robust control with global stability. The main new features of the current approach include: (1) stabilization of open-loop unstable plants using a state feedback strategy, (2) incorporation of an adjustable pre-filter to achieve offset-free control, and (3) implementation of a Kalman filter for state estimation where necessary. The improved controller design method is successfully applied to two non-linear processed with different chaotic behaviour, namely a continuous stirred tank reactor and a Zymomonas mobilis reactor. Compared with conventional methods without model modifications, the new approach has achieved significant improvement in control performance and robustness with dramatically reduced number of local models.

Details

Metrics

97 Record Views
Logo image