Stiction in control valve is the most common and long standing problem in process industry, resulting in oscillations in process variables which subsequently lower product quality and productivity, accelerates equipment wear and tear, or leads to system instability. In this paper, we apply the Mixed-Integer Quadratic Programming (MIQP)based Model Predictive Controller (MPC) that was originally developed for system with backlash, to the system with the control valve stiction. The objective is to investigate the robustness of the original MIQP-based MPC in the presence of control valve stiction nonlinearity. To have the real stiction model, we apply neural-network modelling from simulated data of real valve stiction and then use it to test the MIQP-based MPC control. Different scenarios of control valve stiction are considered. Simulation studies using an industrial case study of fluid catalytic cracking unit (FCCU) show that, if the approximate dead-band value is known a priori, the MIQP-based MPC can effectively improve the closed-loop performance in the presence of stiction through the reduction of oscillations in the process variables.
[1]
Nina F. Thornhill,et al.
A simple test to confirm control valve stiction
,
2005
.
[2]
S. Joe Qin,et al.
A survey of industrial model predictive control technology
,
2003
.
[3]
Jan M. Maciejowski,et al.
Predictive control : with constraints
,
2002
.
[4]
M.A.A. Shoukat Choudhury,et al.
Detection and Diagnosis of Plant-wide Oscillations: An application study
,
2004
.
[5]
Yudi Samyudia,et al.
A hybrid formulation and design of model predictive control for systems under actuator saturation and backlash
,
2006
.
[6]
Manabu Kano,et al.
Practical Model and Detection Algorithm for Valve Stiction
,
2004
.
[7]
Alberto Bemporad,et al.
Control of systems integrating logic, dynamics, and constraints
,
1999,
Autom..
[8]
Emad Ali,et al.
On-line tuning strategy for model predictive controllers
,
2001
.