Title: Neuro-fuzzy-based Smith predictor for FOPLDT process control

Authors: Hao Chen; Zoubir Zouaoui; Zheng Chen

Addresses: Institute for Arts, Science and Technology, Glyndwr University, Mold Road, Wrexham, LL11 2AW, Wales, UK. ' Institute for Arts, Science and Technology, Glyndwr University, Mold Road, Wrexham, LL11 2AW, Wales, UK. ' Institute for Arts, Science and Technology, Glyndwr University, Mold Road, Wrexham, LL11 2AW, Wales, UK

Abstract: In this paper, a novel approach for a first-order-plus-long-dead-time (FOPLDT) process control has been developed. The Smith predictor (SP) is a well-known compensation scheme to deal with the problem of long dead time in process control. However, in the presence of process modelling errors in gain, time constant or time delay, the Smith predictor appears to have poor performance, as even instability may be caused as soon as model mismatch becomes significant. An application of neuro-fuzzy-based Smith predictor method is introduced to improve the robustness to modelling errors for FOPLDT process control. From simulation evaluation, this new proposed approach provides superior performance to the original Smith predictor and robustness to modelling errors.

Keywords: dead time; Smith predictor; neural networks; fuzzy logic; modelling errors; process control; process modelling; simulation.

DOI: 10.1504/IJMA.2012.049398

International Journal of Mechatronics and Automation, 2012 Vol.2 No.3, pp.169 - 177

Received: 24 Mar 2012
Accepted: 13 Jul 2012

Published online: 02 Oct 2012 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article