SIMULTANEOUS PROCESS DESIGN AND CONTROL FOR OPTIMAL GRADE TRANSITION IN A STYRENE POLYMERIZATION REACTOR

M. ASTEASUAIN, A. BRANDOLIN, C. SARMORIA, A. BANDONI

Valencia, España

Congreso; SLAP 2004; 2004

SIMULTANEOUS PROCESS DESIGN AND CONTROL FOR OPTIMAL GRADE TRANSITION IN A STYRENE POLYMERIZATION REACTOR Mariano Asteasuain, Adriana Brandolin, Claudia Sarmoria, Alberto Bandoni PLAPIQUI, UNS-CONICET, CC 717, (8000) Bahía Blanca, Argentina. Grade transition optimization is very important for a profitable operation of continuous polymerization processes. Polymers have many application areas, and each area demands different grade specifications. In order to achieve market requirements, continuous polymer plants usually alternate between the production of several polymer grades in the same equipment by switching the operating points. Since this operation can be performed even every few days, it is essential to determine optimal transition policies that minimize the production of off-specification material and the transition time. In addition to this, it is necessary to count with a suitable control system that guarantees that the optimal transition policies are actually followed, and also ensures safe process operation. This has motivated intense research on grade transition optimization and control of polymerization reactors (1). In most cases, however, a sequential approach has been used to deal with control system design and optimization of transition policies. By following this sequential methodology, the strong interaction between process design and control is not accounted for. Although several works in other fields of chemical engineering have shown the great benefits of incorporating control aspects in the process design stages (2), few efforts have been done in this direction in polymer engineering (3). In this work we focus on grade transition operation in a CSTR styrene polymerization reactor. A Mixed – Integer Dynamic Optimization (MIDO) approach is used to simultaneously design the process and its control system for A ↔ B grade transition sequences. The process design involves reactor size and initiator type selection (discrete decisions), and the two steady state operating points in which each polymer grade is produced (continuous decisions). Simultaneously, feedforward – feedback controllers are optimally designed to drive the process between steady states. The proposed control superstructure for PI feedback controllers involves the monomer, initiator and coolant flow rates as possible manipulated variables, and the reactor temperature, jacket temperature, number average molecular weight and polymerization rate as possible controlled variables. Optimal pairings between them must be determined (discrete decisions) as well as the controllers tuning parameters and the set points of the controlled variables (continuous decisions). The feedforward controllers’ signals consist of time-varying profiles for each manipulated variable, which are calculated by the optimizer. The MIDO was solved with the gPROMS/gOPT v2.2.6 package (Process Systems Enterprise Ltd.). A few examples of the optimal design features are shown in Figures 1 and 2. Figure 1 presents the optimal feedback control structure for the A → B transition. All allowed loop pairings are used in this structure. The combination of feedback loops is probably intended to compensate for the strong interaction between process variables in a polymerization system. Figure 2 shows the Mn trajectory for the same transition. It can be seen that the optimal transition policy implemented by the control system takes only 20 min to reach grade B specification (in a reactor with a residence time of 2.5 h), and closely follows this value thereon. The optimal time profile of the Mn set point resulted in a constant value coincident with Mn↔ B grade transition sequences. The process design involves reactor size and initiator type selection (discrete decisions), and the two steady state operating points in which each polymer grade is produced (continuous decisions). Simultaneously, feedforward – feedback controllers are optimally designed to drive the process between steady states. The proposed control superstructure for PI feedback controllers involves the monomer, initiator and coolant flow rates as possible manipulated variables, and the reactor temperature, jacket temperature, number average molecular weight and polymerization rate as possible controlled variables. Optimal pairings between them must be determined (discrete decisions) as well as the controllers tuning parameters and the set points of the controlled variables (continuous decisions). The feedforward controllers’ signals consist of time-varying profiles for each manipulated variable, which are calculated by the optimizer. The MIDO was solved with the gPROMS/gOPT v2.2.6 package (Process Systems Enterprise Ltd.). A few examples of the optimal design features are shown in Figures 1 and 2. Figure 1 presents the optimal feedback control structure for the A → B transition. All allowed loop pairings are used in this structure. The combination of feedback loops is probably intended to compensate for the strong interaction between process variables in a polymerization system. Figure 2 shows the Mn trajectory for the same transition. It can be seen that the optimal transition policy implemented by the control system takes only 20 min to reach grade B specification (in a reactor with a residence time of 2.5 h), and closely follows this value thereon. The optimal time profile of the Mn set point resulted in a constant value coincident with Mn→ B transition. All allowed loop pairings are used in this structure. The combination of feedback loops is probably intended to compensate for the strong interaction between process variables in a polymerization system. Figure 2 shows the Mn trajectory for the same transition. It can be seen that the optimal transition policy implemented by the control system takes only 20 min to reach grade B specification (in a reactor with a residence time of 2.5 h), and closely follows this value thereon. The optimal time profile of the Mn set point resulted in a constant value coincident with MnMn trajectory for the same transition. It can be seen that the optimal transition policy implemented by the control system takes only 20 min to reach grade B specification (in a reactor with a residence time of 2.5 h), and closely follows this value thereon. The optimal time profile of the Mn set point resulted in a constant value coincident with MnMn set point resulted in a constant value coincident with MnMn of grade B polymer. Optimal values for the other process and control system deign features mentioned before were also calculated. Equivalent results have been obtained for the B → A transition sequence.→ A transition sequence. ACKNOWLEDGEMENTS The authors thank CONICET and UNS for financial support. REFERENCES (1) Embiricu, M., E.L. Lima and J.C. Pinto (1996). Polym. Eng. Sci., 36(4), 433-447. (2) Bansal, V., J.D. Perkins and E.N. Pistikopoulos (2002). Ind. Eng. Chem. Res., 41, 760-778.Polym. Eng. Sci., 36(4), 433-447. (2) Bansal, V., J.D. Perkins and E.N. Pistikopoulos (2002). Ind. Eng. Chem. Res., 41, 760-778.Ind. Eng. Chem. Res., 41, 760-778., 41, 760-778. Fig. 1: Optimal pairings between manipulated and controlled variables in A → B grade transition.→ B grade transition. Coolant flow rate Initiator flow rate Monomer flow rate Reactor temperature Jacket temperature Number average molecular weight Polymerization rate Fig. 2: Mn and Mn set point profiles in A → B grade transition.Mn and Mn set point profiles in A → B grade transition. 0 40 80 120 160 200 Time (min) 35000 40000 45000 50000 55000 Mn (g/mol) Polymer Mn Mn set point Mn of Grade A Mn of Grade B (3) Chatzidoukas, C., J.D. Perkins, E.N. Pistikopoulos, and C. Kiparissides (2003). Chem. Eng. Sci., 58, 3643-3658.Chem. Eng. Sci., 58, 3643-3658.