Time delay integrating systems: a challenge for process control industries. A practical solution
Introduction
Control of processes involving the heating and cooling of a closed batch reactor are a real problem for conventional proportional-integral-derivative (PID) based loop controllers, due to the reduced stability margins typical for these applications. These processes exhibit long dead times and time constants and have an integrating response due to the circulation of the heating or cooling medium through coils within the reactor or jackets on the outside of it.
The advanced controller described in this paper has the ability to model and control marginally stable processes with long and time varying delays. This controller exhibits the ability to incorporate and model the effect of known and unknown disturbances. The field application results presented are demonstrating that reactors which could previously only be operated manually can now automated using model predictive control technology. The barrier to automation of the reactor batch controls is removed resulting in improvements in batch consistency, batch cycle times and productivity.
A number of industrial applications of advanced control methods are reported for batch processes. The limitations of these schemes are fundamentally connected with the application. These schemes lack the generality required to solve batch reactor industrial control in an unified manner. Schemes previously proposed for this application include conventional feedback control with feed forward compensation (Berber, 1996), gain scheduling or multiple models (Krishnan & Kosanovich, 1997; Jutan & Uppal, 1984), generic internal model control (Morari & Zafiriou, 1989) and adaptive regulators (Cabassud, Le Lann, Chamayou, & Casamatta, 1989; Chen & Peng, 1998). The typical batch process variables evolve over a wide range therefore time linear invariant models tend to fail in describing completely the process dynamics. Few authors are looking into these challenges from the perspective of predictive control (Garcia, 1984; Koncar, Koubaa, Legrand, Bruniaux, & Vasseur, 1997; Lee, Chin, Lee, & Lee, 2000; Nagy & Agachi, 1997). Some authors (Jarupintusophon, Lelann, Cabassud, & Casamatta, 1994; Fileti & Pereira, 1997) are reporting applied adaptive control techniques. However, a potential problem of their approach can arise from the identification scheme adopted (e.g. the use of ARMAX models limits the generality of the approach). Also, in the case of grey box models used, as in (Jarupintusophon et al., 1994), an intimate knowledge of the plant is required.
The paper content is split in six sections. After the introduction containing achievements to date in the second section the theory behind dynamic modelling and control is addressed. Section 3 describes the process to be controlled. In Section 4 the results concerning the controlled process are provided. Other successful applications are encompassed in Section 5 followed by conclusions in Section 6.
Section snippets
The adaptive predictive control strategy
Based on an original theoretical development by Dumont and Zervos (1986); Zervos (1988) the controller was first developed for self regulating systems. This controller was credited by various users with features like reduced effort required to obtain accurate process models, inclusion of adaptive feed forward compensation and ability to cope with severe changes in the process etc.
These features together with a recognized need in process control industry made the authors of this paper consider a
A batch reactor processes
The chemical batch reactor in this application is used to produce various polyester compounds. The process involves combining the reagents and then applying heat to the mixture in order to control the reactions and resulting products. A specific temperature profile sequence for the batch reaction must be followed to ensure that the exothermic reactions occur in a controlled fashion and that the resulting products will have consistent properties. An additional requirement is that the reaction
Application results
The advanced model based controller was implemented on the reactor temperature control loop. The controller parameters were estimated from the observed system response from a previous batch and an approximate model of the system was developed in the controller using 15 Laguerre filters. There was some concern that a single model of the system may not be valid for the entire batch sequence because the composition and viscosity of the polyester in the reactor changes substantially during the
Other successful applications
The model based predictive controller has been implemented on many other industrial processes involving significant time delays. Control of pulp bleaching to achieve a desired brightness following a 40 min reaction time in a retention tower is among them. In this example, the mill had designed a Smith Predictor compensated PID controller and the model predictive controller was installed to compare performance. The results of this application demonstrated a 48% reduction in pulp brightness
Conclusions
An advanced model based predictive controller MBPC developed for use on processes with an integrating response exhibiting long dead time and time constants has been successfully applied to the temperature control of a batch reactor. The controller was easy to apply and configure. It has achieved very good control performance on a reactor that could not be controller in a satisfactory manner using PID controls implemented in the plant DCS.
The automatic control of the reactor temperature now
References (23)
- et al.
Properties of generalized predictive control
Automation
(1989) - et al.
Adaptive control
(1995) Control of batch reactorsA review
Transactions Institute of Chemical Engineering
(1996)- Cabassud, M., Le Lann, M. V., Chamayou, A., & Casamatta, G. (1989). Modelling and Adaptive Control of a Batch Reactor,...
- et al.
A simple adaptive control strategy for temperature trajectory tracking in batch processes
The Canadian Journal of Chemical Engineering
(1998) - Clarke, D. (1993). Advances in model-based predictive control (pp. 3–21). Oxford: Oxford University...
- de Dona, J. A., Goodwin, G. C., & Seron, M. M. (1999). Connections between model predictive control and anti-windup...
- Dumont, G. A., & Zervos, C. C. (1986). Adaptive controllers based on orthonormal series representation. In Proceedings...
- et al.
Adaptive and predictive control strategies for batch distillation
Computers Chemical Engineering
(1997) - Garcia, C. E. (1984). Quadratic dynamic matrix control of nonlinear processes—an application to a batch reaction...
Adaptive filtering, prediction and control
Cited by (50)
Discrete-time domain two-degree-of-freedom control design for integrating and unstable processes with time delay
2016, ISA TransactionsCitation Excerpt :Note that most of existing control methods as aforementioned were designed in continuous-time or frequency domain, which in fact, need to be discretized for implementation in digital control systems. This may provoke undesirable loss in control performance or even closed-loop instability for integrating or unstable processes [1–3]. Besides, discrete-time models are extensively established based on sampled data for control design in engineering practice.
Improved PI controller based on predictive functional control for liquid level regulation in a coke fractionation tower
2014, Journal of Process ControlIdentification of integrating processes with time delay
2013, IFAC Proceedings Volumes (IFAC-PapersOnline)A novel dead time compensator for stable processes with long dead times
2012, Journal of Process ControlHybrid intelligent control scheme of a polymerization kettle for ACR production
2011, Knowledge-Based SystemsCitation Excerpt :When the process time delay becomes significant relative to the time constant, the performance of the closed-loop system can be improved by using a predictor structure. These predictor based controllers are known as dead-time compensators (DTC) and have been applied to many process industries [18,32]. The first DTC structure, Smith predictor which is still being used today was introduced at the end of 1950s to improve the performance of classical controllers (PI or PID controllers) for plants with dead time [39].