QDMC OF DISTILLATION COLUMNS

Abstract

Many industrial control problems are nonlinear and multivariable in nature. It is common for dynamic models of industrial processes to have strong interactions between the loops. Distillation columns also exhibit elegantly complex dynamics which include strong interactions and large dead times. Though product quality control of simple binary distillation columns somewhat easy, it is challenging job for control engineer to control product quality of complex distillation columns like crude distillation columns. This is due to strong interactions among sidestream product quality control loops and large input-output delays. Significant economic benefits can result from improved control of crude distillation towers because of their large throughput. Present work is about control of crude tower product quality and it is a 4 x 4 control problem which include strong interactions and large dead times. It is common for crude towers to change products quality on market specifications. So controller should maintain product quality on specification. Leo Hsie and McAvoy (1991) worked on product quality control of crude distillation tower. Both conventional PI control and model based QDMC control were applied to crude tower product quality variables. They had used old guide lines in tuning QDMC controller. This tuning involves trial and error procedure in selecting move suppression coefficients. Present work includes redesign of conventional PI controller, redesign of QDMC with old tuning guidelines and design of QDMC with Novel tuning strategy for product quality control of crude distillation column. Present study considers the transfer function models and operating conditions (i.e. constraints on manipulated and controlled variables) given by Leo Hsie and McAvoy (1991). We have followed multiloop BLT tuning method proposed by Luyben (1986) in redesigning conventional PI controller and old tuning guidelines given in Cutler and Ramaker (1980) in redesigning QDMC for product quality control. Present work also followed tuning strategy proposed by Shridhar and Cooper (1998) in designing QDMC with novel tuning strategy. MATLAB simulink and MPC Toolbox are used for simulating results for PI and QDMC respectively. Finally performance of QDMC with Novel tuning strategy is compared with results of conventional PI and QDMC with old tuning guidelines. It was found that the QDMC with Novel tuning strategy performs better than other two control structures.