Ruth Bars

Predictive Control in Process Engineering: From the Basics to the Applications

Describing the principles and applications of single input, single output and multivariable predictive control in a simple and lively manner, this practical book discusses topics such as the handling of on-off control, nonlinearities and numerical problems. It gives guidelines and methods for reducing the computational demand for real-time applications. With its many examples and several case studies (incl. injection molding machine and waste water treatment) and industrial applications (stripping column, distillation column, furnace) this is invaluable reading for students and engineers who would wish to understand and apply predictive control in a wide variety of process engineering application areas.

Optimal Choice of Horizons for Predictive Control by Using Genetic Algorithms

Abstract For predictive control in industry often very long horizons for control error and manipulated signal are used because of the slow processes which take place in the petrochemical industry. In order to reduce the computational effort some commercial predictive control program packages offer the ability to reduce the number of points in both horizons but do not recommend how to select the points which have to be considered in the horizon of the control error and manipulated variable. In this work the authors introduce an optimal choice not only of the horizon lengths itself but also for the strategy of reducing the number of points in the horizons. A genetic optimization algorithm was used both for the search for the optimal length of the horizons and for the best allocation of the points in the horizons. The results of the optimization process where used to deduct a simple rule.

A New Web/Matlab Based System in Control Education

Abstract At the Department of Automation, Technical University of Budapest in the last years the basic course of control theory has been renewed significantly using CAD devices, especially Matlab in the computer classroom. In this way the theoretical knowledge became more understandable and convincing for the students. Recently the Internet culture has burst into our everyday life, providing new possibilities for control education, too. It was a challenge lo combine the facilities provided both by the Web and Matlab. In the background of the system the Matlab program package supports the teaching process executing the computations. A new control curriculum using these facilities has been being developed.

Nonlinear Predictive Control Algorithms with Different Input Sequence Parameterizations Applied for the Quadratic Hammerstein and Volterra Models

GPC (Generalized Predictive Control) developed by Clarke [1] is extended for different nonlinear input/output models. The nonlinear process models applied are the Hammerstein series model, the Volterra series model, the parametric generalized Hammerstein model and the parametric Volterra model. Extended horizon one-step-ahead and long-range optimal predictive control algorithms are derived. A quadratic cost function is minimized, which considers the quadratic deviations of the reference signal and the output signal predicted in a future point beyond the dead time and also punishes big control signal increments. The future process output is predicted in dependence of the control increment. Using a special assumption on the form of the input signal during the control horizon, the multi-dimensional optimization reduces to a one-dimensional problem. Hard input limitations can be considered, as well.

Nonlinear Long-Range Predictive Control of a Distillation Pilot Plant

Abstract A distillation column separating ethanol-water mixture has been built at our department several years ago. The pilot plant is equipped with analogue instrumentation, manipulating board and alarm devices. It is connected to a microcomputer and a supervisory computer. The aim of the control is to keep the concentration of the products constant. Further on several physical quantities have to be controlled (feed flow rate, feed temperature, reflux flow rate, liquid level in the reboiler, etc.). The plant is highly nonlinear. Long-range predictive control algorithms based on nonparametric system description have been generalized to nonlinear systems characterized by Volterra series by Bars and Haber (1988). These algorithms have been applied successfully to some control loops of the distillation column. Initial experimental results are presented here.

Robust Predictive PI Controller Based on First-Order Dead Time Model

Abstract Predictive control algorithms compute the manipulated variable minimizing a cost function considering expected future errors. PI control algorithms can be equipped with predictive properties. Simple predictive control algorithms are derived using approximation of an aperiodic process by a first-order model with dead time. Applying a noise model the robustness properties of the algorithm are enhanced considering plant-model mismatch. The noise filter is considered as a design parameter. Simulation examples demonstrate the behavior of the predictive PI algorithm and the robustifying effect of the noise filter.

THEORY, ALGORITHMS AND TECHNOLOGY IN THE DESIGN OF CONTROL SYSTEMS: “Status report prepared by the IFAC Coordinating committee on Design Methods”

Abstract Control theory deals with disciplines and methods leading to an automatic decision process in order to improve the performance of a control system. The evolution of control engineering is closely related to the evolution of the technology of sensors and actuators and to the theoretical controller design methods and numerical techniques to be applied in real time computing. New control disciplines, new development in the technologies will fertilize quite new control application fields. The status report gives an overview of the current key problems in control theory and design, evaluates the recent major accomplishments and forecasts some new areas. Challenges for future theoretical work are modelling, analysis and design of systems in quite new applications fields. New effective real-time optimal algorithms are needed for 2D and 3D pattern recognition. Design of very large distributed systems has presented a new challenge to control theory including robust control. Control over the networks becomes an important application area. Virtual reality is developing in impressive rate arising new theoretical problems. Distributed hybrid control systems involving extremely large number of interacting control loops, coordinating large number of autonomous agents, handling very large model uncertainties will be in the center of future research. New achievements in bioinformatics will result in new applications. All these challenges need development of new theories, analysis and design methods.

Long-range optimal model and multi-step-ahead prediction identification for predictive control

Abstract Long-range optimal prediction algorithms use the predicted output for several steps ahead. The prediction based on traditionally estimated model parameters does not result in an optimal prediction if the measurements are noisy or/and model structure differs from real process structure. In this paper two different identification schemes are presented and compared: long-range predictive single-model identification and simultaneous multi-step-ahead prediction identification. It is shown that the first method is easier to realize but the second one leads to more accurate results. Both methods are derived for a first-order model in details. Simulation runs and a level control example illustrate the algorithms presented.

ADAPTIVE PREDICTIVE CONTROL OF NONLINEAR DYNAMIC PROCESSES

Abstract Predictive control algorithms determine a series of the control signal minimizing the deviation between the reference and the output signal in a given future horizon. The output of the plant to be controlled is predicted on the basis of a model, which can be linear or nonlinear, parametric or nonparametric. In adaptive control these process parameters are identified and the control signal is calculated taking the identified values into consideration. In the paper simulation results present some properties of adaptive predictive control. Robustness of the control algorithm is illustrated through control of a simple Wiener model . The relationship between the plant order and the prediction horizon is also mentioned. The promising results indicate that further systematic analysis is worthwhile.

A New Curriculum in Control Education Supported by Matlab

Abstract Recently a great modification has been carried out in the control education at the Department of Automation, Faculty of Electrical Engineering and Informatics, Technical University of Budapest. Beyond the slight renewal and modernization of the control theory education the main goal was to develop the students individual CAD problem solving ability. The paper shortly presents advantages and experiences of this educational illustrates the computer problem solving by an example and outlines the future developments