159 Nonlinear Multivarlabie Control of Product Properties In an Industrial Gas Phase Polyethylene Reactor K.B. McAuley, J.F. MacGregor, pp 173-178 A nonlinear model-based scheme is developed for product property control in industrial gas phase polyethylene reactors. The controller is designed for instantaneous melt index and density regulation over a range of products, and for control during grade changeovers. The nonlinear model-based feedback controller design is based upon global input/output linearization methods. Mismatch between the model and the process is removed using an extended Kalman filter. Through simulations on a complex mechanistic model of the process, it is shown that the controller performs well for both regulatory and servo-eontrol. The simplicity of the control algorithm makes it an excellent candidate for industrial application.
160 A Comparison of Strategies for the Control of a Polypropylene Reactor B. Lie, J.G. Balchen, pp 179-184 The paper discusses various structures for the control of the reactor in an industrial polypropene plant. The present structure uses single-loop PI controllers. In a suggested new structure, refluxed liquid is used to control the pressure while removed energy in the heat exchanger controls the accumulator liquid level. A multivariable PID controller is investigated, in addition to a decoupling structure. The different structures are compared with regard to set point tracking, disturbance rejection and suppression of varying process parameters. Overall, the multivariable PID controller yields the best result. The present structure may also give a satisfactory result if properly tuned.
161 An Analytical Approach to Modelling in Distillation Control T. Liider, G. Wozny, pp 185-192 This paper presents a second-order linear process model for a distillation column. It gives a steady-state exact representation of the column for any control configuration in the time and frequency domains, and preserves the physical significance of all parameters. Because of the low computational effort and the consistency of the modelling information it is an effective tool for control structure analysis and controller design. A systematic comparison with related work reveals the analytical strength of the approach. A comparison to simulation results shows very satisfactory reproduction of the column behaviour in the low- to midfrequency range as well as the high-frequency deficiencies of the model.
162 The Unique Challenges of Cryogenic Distillation Column Control for Integrated Coal Gasification Combined Cycle Applications D.M. Esple, J.A. Mandler, D. Miller, D. O'Connor, pp 193-199 The world's largest coal gasification plant for electricity generation is being built in Buggenum, The Netherlands and will be on-stream in 1993. A program of dynamic modelling, simulation, and control system design was undertaken at a very early stage in the plant design, prior to detailed equipment specification. This paper describes the approach taken to understand the dynamics of the proposed air separation unit under the expected load-
567 following conditions in order to verify the process design and to develop and optimise the control strategy. The work highlights the importance of incorporating control and operability considerations early during plant design.
163 Dynamics and Control of Unstable Distillation Columns E.W. Jacobsen, S. Skogestad, pp 201-206 The paper addresses the dynamics and control of distillation columns operated at open-loop unstable operating points. Ideal two-product distillation columns may have multiple steady states and right half plane poles. With reflux and boilup as independent variables the operating points become unstable if the internal flows are sufficiently large. An open-loop unstable operating point may be stabilized by use of one-point control, i.e., feedback control. If the control is insufficiently fight, the column may go into a stable limit cycle. With distillate flow and boilup as independent variables the operating points may also become unstable. The column may then also go into a stable limit cycle.
164 Relating the DRD Structure to Conventional Model Based Controllers J.B. Wailer, K.V. Waller, pp 207-212 A systematic way of constructing control structures, introduced in an earlier paper, has been investigated. A parametrization of a suggested static disturbance rejecting and decoupling structure (DRD) has enabled analysis of the structure in terms of other model-based controllers (such as IMC and Inferential Control). The analysis shows that in practice there will be a need for rel'ming the sla'ucture design by introducing suitable dynamics. Illustrative means of how to do this are presented.
165 Control Structures for a Sidestream Distillation Column Separating a Ternary Mixture A KoggersbOI, S.B. Jergensen, pp 213-218 The distillation under study is that of an impurity which accumulates on intermediate trays due to the thermodynamics of the ternary system. The resulting control problem is to minimize the sidestream flow (i.e. maximize main product recovery) while maintaining the two high-quality end products. This paper first investigates sensor selection with the goal of identifying the column states using relatively few sensors. Next, actuators are selected with the goal of obtaining a control problem of low sensitivity. A simple strategy for controlling the location of the maximum concenlration of the accumulating impurity is described and lowest control sensitivity is obtained for an (LV) type of control structure.
166 Selection of the Best Control Configuration for an Industrial Distillation Column D. Ariburnu, C. Ozgen, T. Gurkan, pp 219-224 The best regulatory control structure is determined for an industrial, multicomponent, high-purity ethylbenzene distillation column. Steady-state plant data and design information are utilized with a ste-dy-state model of the column. The column-model match is obtained by adjusting the column efficiency. Dynamic column behaviour is experimentally obtained by pulse and step testing. Steady-state rating programs are used to obtain the steady-state process and disturbance gain matrices of the control configurations. Steady-state control configuration selection methods are used. Single-end and dual composition control of the column is examined
by IMC-hased PI controllers. The results reveal that for feed flow rate and composition disturbances, the column can be controlled with a single-end structure.
167 Non-Linear Analysis of Distillation Control Structures J. Alvarez, J. Aivarez, C. Martinez, pp 225-230 With geometric control tools, the two-point control configuration problem for distillation columns is analyzed. The notion of left-invertability for nonlinear systems is used to establish solvability for the closedloop stabilization with noninteractive output dynamics. Results are analytic and use of first-principle modelling enables physical interpretation of derivations and results. The approach provides information about the necessity or not of integral action and about coupling of the underlying control strategy. Findings agree with and support earlier results obtained from finear techniques, with simulations and experimental testing.
application of the predictive controller to a nonlinear distillation system. The enhanced performance using the artificial neural network based control methodology is demonstrated.
171 Application of a Bilinear Long-Range Predictive Control Method in a Distillation Process Kun Lo, En Sup Yoon, Yeong.Koo Yeo, Hyung-Keun Song, pp 249-254 Lung-range predictive controllers for discrete-time MIMO bilinear processes are derived based on a new bilinear model with integral action. Using this model a multi-step-ahead optimal predictor is derived. Two alternative solution methods - rigorous and shortcut - of the minimization problem of a long-range objective function are established and used to calculate control inputs. Several simulation results show that the proposed control methods have robusmess to the limited a priori knowledge of the process. Dynamic simulation on dual composition control of a binary distillation process showed satisfactory servo and regulatory performances of the proposed algorithms.
168 A Comparative Study of Linear and Nonlinear Multivarlable Binary Distillation Column
Control R. Pullinen, P. Pletili, T. Jussila, P. Lautala, pp 231-236 Distillation columns present several problems for a control engineer because of their nonlinear nature and strong interactions. In this paper model predictive control and multivariable PI control are compared. Also, a method for constructing a nonlinear process model and linearizing it using Simulab software is presented. The controller simulation results indicate that model predictive control behaves better than PI control. Anyhow, the ease of tuning of the PI controller makes it a very attractive choice.
172 Combining Adaptive and Neural Control for Distillation Control M. Roele, K. Warwick, pp 255-260 Although developments in the computer industry have moved towards highly integrated parallel processing, the control industry generally only makes use o f the computer as a digital numerical manipulation tool for controlling the plant with a monitoring ability for failure detections. However, different types of control algorithms with a certain degree of intelligence can work in parallel, giving the possibility of better control performance under increased uncertainty, learning abilities and the possibility of intelligent fault tolerant design. In this paper, a simple parallel control scheme is presented for the control of the reflux, reboiler and condenser in a batch distillation process.
169 Feedforward/Feedback Control of a Binary High Purity Distillation Column J. Broil, H. Gelbe, pp 237-242 The objective of feedforward control is to measure the disturbances in process loads and to correct the manipulated variables so that the controlled variables remain unaffected. In this paper a nonlinear timediscrete rigorous model of a distillation column was used to calculate the manipulated variables relating to the measured disturbances, the controlled variables remaining constant. It is shown in simulations that the control behaviour can be improved, if the feedforward model is used in combination with a conventional feedback controller. The feedforward control is also advantageous when compared with a robust controller designed and tuned with the same model.
170 Predictive Control of Distillation Columns using Dynamic Neural Networks G.A. Montague, M.T. Tham, M.J. Willis, A.J. Morris, pp 243-248 In this paper a nonlinear multivariable predictive controller is proposed where the model used for control law synthesis is an artificial neural network. The controller makes use of an on-line optimisation routine which determines the future inputs that will minimise the deviations between the desired and predicted process outputs. Control is implemented in a receding horizon fashion. The paper highlights the importance of selection of the network training philosophy by
173 Discrete-Event Controlled Systems in the Chemical Processing Industry H.A. Preisig, pp 261-266 What are discrete-event dynamic systems and where do they show up in the chemical industry? The paper tries to fabricate a pair of glasses for the reader which will enable him to recognize the discrete-event dynamic nature of chemical processes. What are the modelling methods used and what are their strength and weaknesses? The paper tries to point them out, unbiased but brief. References are made to major cona'ibutions and information sources. What chemical processes have been analyzed and what techniques have been used? The paper lists the processes and briefly describes their nature. Where are things going? The author's view is presented.
174 The Dynamic Modeling and Optimization of an Industrial Batch Reactor S.E. Keeler, J.W. Hull, Jr., G.L. Agln, pp 267-272 The development of a kinetic mechanism for an industrially important process is presented. This mechanism is then implemented into a process model describing the operation of a batch reactor. The resulting model is used to examine the effects of alternative operating strategies on the production process. The Maximum Principle is used to determine optimal temperature profiles for the operation of the reactor,