Fast Online Constrained Optimizing Controllers

  • Awo King-Hans

Student thesis: Master of Philosophy


Research in recent years has focused on the application of more advanced control technologies in industrial controllers, however the low computing power of standard industrial controllers has limited its implementation to higher end hardware. In multivariable input-constrained plants, actuator saturation causes two major problems for control engineers namely windup and directionality. This research focuses on a Two-stage Multivariable IMC Antiwindup (TMIA) structure for open-loop stable plants which requires minimal computing power and tackles the aforementioned control problems in an intuitive and easy to tune way. The highlight of this IMC-based structure is the solution of two low-order quadratic programs to control both steady-state and transient behaviours of the plant. The TMIA structure is further developed to handle constraints on the input rate in a simple but effective way. The controller is tested by application to a Quadruple Tank process in both minimum and non-minimum configurations controlled by a PLC. Although the focus of this paper is on computation and not performance, the TMIA structure is found to outperform its IMC counterparts in handling windup and directionality. Comparison of the TMIA controller and Model Predictive Controller is also carried out and shows competing results, hence is a suitable alternative for this class of systems. Results on computation obtained demonstrate the realizability of the advanced control technique on an off-the-shelf low-end industrial PLC using three different quadratic programming methods. Worst case computation time is in the region of 5ms using the projected fast gradient method, this shows that the controller embedded on the PLC can be applied to much faster processes. Thus the TMIA structure is presented as a competitive alternative for input-constrained multivariable plants in terms of tuning transparency and reduced computations compared to other advanced control techniques such as MPC which are limited by the low computational power offered by standard PLCs.
Date of Award1 Aug 2018
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorWilliam Heath (Supervisor) & Joaquin Carrasco Gomez (Supervisor)


  • Multivariable control
  • Input Constraints
  • Anti Windup
  • Model Predictive Control
  • Quadratic Programming
  • Programmable Logic Controllers
  • Internal Model Control

Cite this