Advanced Control Methods

Professor Urpo Kortela and Professor Raimo Ylinen
Systems Engineering Laboratory, Department of Process and Environmental Engineering, University of Oulu
urpo.kortelaoulu.fi, raimo.ylinenoulu.fi
http://www.infotech.oulu.fi/adcon

Background and Mission

Process control is an efficient means of improving the operation of a process, the productivity of a plant, and the quality of products. In process engineering, even a small improvement in the operation of a process can have great economic and environmental influences.

In order to develop better - fast, accurate, robust and fault tolerant - process control, model-based modern control methods and efficient adaptive and learning techniques are required. During recent years, the developments in these fields have introduced new tools for use in control engineering: neuro-fuzzy systems, guided random search techniques, predictive control, etc. In process engineering, these new tools have found applications in non-linear process modelling and control, plant optimisation, monitoring, scheduling, fault diagnosis, etc. The application area of control engineering methods can be extended also to systems beyond the realm of traditional process engineering.

Although the above mentioned techniques have been the subject of intensive scientific research, there are still many theoretical and practical problems to be investigated. A clear tendency and need for real industrial applications of advanced control methods can be seen. The gap between theory and practice needs to be narrowed. As there is no universal best solution that would apply to all control problems, knowledge of different model structures and alternative learning methods can be used to offer advice on the choice of the most suitable technique for a specific application. The results of the synthesis are readily applicable in industry.

The Infotech research group carrying the name "Advanced Control Methods (ADCON)" has been formed from the Systems Engineering Laboratory in the Department of Process and Environmental Engineering at the University of Oulu.

Scientific Progress

The scientific research of the Infotech Oulu research group in Advanced Control Methods (ADCON) prospered in 2002. There was a number of on-going research and training projects. Several projects found financial support from the Academy of Finland. In basic research work, the group has been concerned with the development of novel structures and algorithms for identification of non-linear systems, as well as robust, predictive, adaptive, constrained and learning control of industrial processes.

Polynomial methods are modern industrial design techniques for complex multivariable systems, signals and processes based on manipulations with polynomials, polynomial matrices, and other similar objects. Our co-operation with the European Network of Excellence for "Industrial Applications of Polynomial Design Methods (EUROPOLY)" has continued.

In a project entitled "Modelling and Control of Non-Linear Processes: Linear Dynamics and Non-Parametric Methods", financed by the Academy of Finland for the period of 2000-2005, algorithms for identification and control of Wiener and Hammerstein systems are being considered. In addition, the project examines stochastic learning automata and their applications in process identification and control. In 2002, new hybrid algorithms were examined, merging learning automata and confidence probabilistics. Applied work on Wiener and Hammerstein systems was conducted in the modelling of a bubbling/circulated fluidized bed power plant.

Development of theory and methods in ADCON includes learning automata in optimization.

"The Generic Hybrid Model of Thermal Power Plants" focuses on the combustion of multi-biofuels (mixtures of peat, wood chips, sawdust and bark, and in some cases coal and municipal waste). The power plant can be approximated as a chain of sub-models for the combustion, the fresh steam generation and the superheating system, completed with models of fuel types (fuel analysis, estimated heat values), fuel preparation (mixing, fuel type ratio) and emission formation. The models serve as an instrument to analyse and optimise thermal efficiency and operation conditions when burning multi-biofuels. Design of advanced controllers is based on these models. The project is funded by the Academy of Finland, 2001-2003. The project started in full effect from the beginning of 2002. In 2002, industrial measurements were carried out together with another combustion-related project. Phenomenological modelling of the superheater system is now almost completed. Fuzzy-neural modelling of the combustion zone has been started.

A typical Code Division Multiple Access (CDMA) receiver in telecommunication contains several non-linear tracking loops. A basic tracking loop consists of a non-linear discriminant function for detecting the delay or phase difference between the received signal and reference signal produced by a controllable signal generator and a so-called loop filter. These are usually tuned by conventional linear methods of control theory. However, due to bad and time varying noise-to-signal ratios, and fast changes in the operation environment, the system often looses its stability. More sophisticated non-linear, time varying, and multivariable control methods have been studied in a project called "Tuning of non-linear feedback loops of CDMA receivers". Using these methods, it seems possible to enlarge significantly the operational area of the receivers.

In the "Fault diagnostics in optimisation of sub-processes in the pulp and paper industry (OVI)" project, fault diagnostics for sub-processes in a TMP plant are being examined. The main target is to minimise the so-called ?grey area?, the area where quality measurements are not within the limits of good quality. This Tekes project is being conducted with UPM-Kymmene, the Valmet Mechanical Pulping and Control CAD. Methods for the OVI project were developed in a project entitled "Advanced Methods in Processes Associated with Problems in Observability and Supervision", funded by the Academy of Finland. Exploitation of fuzzy sensor fusion and clustering has been examined lately, and tests, aiming at improving the observability and control of the refining and screening, were completed successfully. The applied methodology seems to be of a highly generic nature, and some promising results were achieved also in the kraft pulping process. This project terminated in 2002, and continues now in "Diagnosis of the continuous kraft pulping process".

Exploitation of Results

The Advanced Control Methods (ADCON) project unit of Infotech Oulu has several applied projects involving close collaboration with industry. The results from the academic research have been extensively used in applied projects.

The basic equipment in flotation control is an X-ray fluorescence analyser, which measures metal contents of different pulp streams, typically in 5.10-minute sampling intervals. However, measurement of certain variables related to the type and movement of the flotation froth is based on the visual observation of a human operator. In an ESPRIT project called "The Characterization of Flotation Froth Structure and Colour by Machine Vision (ChaCo)" a new vision system for the monitoring and control of flotation processes has been developed. A set-up consisting of a camera, a grabber card, a spectrophotometer and a PC-computer was installed at the Pyhäsalmi flotation plant. The project ended in 2000, but it has been followed by the national VÄSY project financed by Tekes. The project aims to develop and apply multivariable control methods using the information given by a new vision system containing five cameras located on the most important cells.

Machine vision is an integral part in the control of flotation froth structure and color.

The aim of a project called "Modelling and improvement of the continuous pulp cooking process (Jakema)" (2000-2002) was to improve the continuous kraft pulping process by modelling the different sub-processes in normal situations, during grade changes and in disturbance situations. The main sub-processes of interest are impregnation, cooking and washing. The first goal of the study was to ensure optimal and steady cooking conditions, i.e. an even packing degree in the digester scale. The cooking conditions were studied based on the wood and liquor balances and process measurements indicating the horizontal and vertical phenomena in the digester. The second goal was to improve the prediction and control of the Kappa-number. In the Kappa-number calculations, Gustafson?s model was utilised together with accurate temperature and alkali profiles. The research started in Jakema continues now in the Tekes project "Advanced control of continuous cooking (Keha)" (2002-2004) initiated in late 2002. The models developed in Jakema will be utilised in advanced control and in the control of disturbances in the process. The Kappa-number profile will be predicted and different kinds of stability indices will be constructed. The study will be also expanded to Lo-Solids type cooking process. The resulting models will be tested and implemented as a part of VTT?s pulp and paper process simulator APROS. Keha is conducted in co-operation with Stora Enso Fine Papers, Metso Automation and new partners Metsä-Botnia, Andritz and VTT.

Research in "Using an object-oriented approach in modelling fluidized bed combustion boilers (OLMA)", in co-operation with Tekes, Oulun Energia, Air-Ix Tuotantotekniikka Oy and Foster-Wheeler Energia Oy, considers power plant control systems. The main objective is to reduce the flue gas emissions, especially CO and NOx. Models of the process dynamics and variable interactions play an essential role in controller design. The research focuses mainly on oxygen control. In 2002, a combustion-simulator was constructed for control design purposes. The simulator is divided into independent parts, in order to clarify the interfaces and to improve the opportunities to compare the performance and stability of different oxygen controllers. The simulator contains models for the fuel feeding system, the air feeding system, an adaptable semi-phenomenological (grey-box) flue-gas oxygen-content model and a flue-gas oxygen-content control system. The simulator was utilised to improve the performance of control action in a power plant. In this case, the fluctuation of flue-gas oxygen-content around the setpoint decreased by about 70 %. The use of the new control structure allowed a decrease of 1.5% in the flue gas O2 setpoint, thereby improving the boiler efficiency and decreasing CO2 and NOx emissions.

A project on the "Effect of wood fuel on usability and control" (2000-2003) is being lead by VTT Processes and financed by Tekes. The aim of the project is to examine the effects of biofuels (e.g., peat, bark, chips, and waste wood) on the boiler control and the means to improve controls of a biofuel-fired boiler. The main goal is to optimise the behaviour of the fuel feed system and to compensate the fuel quality disturbances. More stable combustion conditions enable a decrease in the excess-air ratio, thus reducing flue gas emissions (NOx, SOx, CO) and improving the efficiency of the boiler. Fuzzy logic is examined to improve the control in a biofuel-fired fluidised bed boiler. In 2002, the fuzzy combustion power control software was developed further, so that users can easily tune and modify the behaviour of the controller. In addition to fuzzy CPC, the program works as a regular fuzzy controller.

Future Goals

The Advanced Control Methods (ADCON) research group of Infotech Oulu has a wide variety of research programmes considering the development and application of modern information technology. In the future, the research will continue to be active, directed towards multivariable and learning systems, model-based modern control design, and fault diagnosis. Collaboration with international partners will be emphasised, along with active publishing in scientifically recognised forums. A major goal is to decrease the gap between control and systems theory and industrial practice.

Personnel

External Funding

Selected Publications

Poznyak AS & Najim K (2002). Learning through reinforcement for n-person repeated constrained games. IEEE Transactions on Systems, Man and Cybernetics Part B, 32 (6): 759-771.

Ikonen E, Kortela U & Najim K (2002). Distributed logic processors in process identification. In: Expert Systems 5: 1457-1495, Ed. Leondes C, Academic Press.

Ikonen E & Najim K (2002) Advanced Process Identification and Control. Marcel Dekker Inc., New York, USA, 310 p.

Rantanen R & Hiltunen J (2002) Improving the usability of measurements in monitoring and control of kraft pulping process. Control Systems 2002, Stockholm, Sweden.

Mannermaa J, Ylinen R & Kiviahde M (2002) Nonlinear control of tracking loops. 10th Mediterranean Conference On Control and Automation (MED2002), Lisbon, Portugal, July 12-15.

Kovács J & Kortela U (2002) Convergence and stability of iterative controller refinement in the presence of constraints. IASTED Control and Applications (CA2002), Cancun, Mexico, May 20-22.

Mononen J, Lerssi I & Leppäkoski K (2002) Disturbance analysis of the fuel feed system in multi-fuel combustion. Proceedings of the IASTED Power and Energy Systems, Marina del Rey, USA, May 13-15, ACTAPress, 69-74.

Ylinen, R & Zenger K (2002) Poles and zeros of multivariable linear time-varying systems. 15th IFAC World Congress 2002, Barcelona, Spain.

Hätönen J & Owens DH (2002) Convex modifications to an iterative learning control law. 15th IFAC World Congress 2002, Barcelona, Spain.

Ahvenlampi T, Hietanen T & Kortela U (2002) Prediction of production rate in pressure screening using fuzzy clustering methods. Proceedings of the IASTED AIA 2002, September 9-12, Benalmadena, Spain.

Najim K, Poznyak AS & Ikonen E (2002) Learning automata-based optimization in a binary coded search space. 15th IFAC World Congress, Barcelona, Spain.

Tervaskanto M, Hietanen T & Kortela U (2002) The process control using SPC and fuzzy modelling techniques. 15th IFAC World Congress, Barcelona, Spain.

Ylinen R (2002) Polynomial systems theory for nonlinear systems. 8th IEEE International Conference on Methods and Models in Automation and Robotics. September 2-5, Szczecin, Poland.