Adaptiv regulering af web winder
Translated title
Adaptive control with self-tuning for center-driven web winders
Authors
H. Andersen, Alexander ; L. Lemminger, Thor ; P. Z. L. Andersen, Mads
Term
2. term
Education
Publication year
2009
Abstract
Many products are made from a continuous long strip of material called a web, for example paper in newspaper printing. In a web winder, paper is unwound from one reel, guided through rollers, and wound onto another. This thesis asks what most strongly affects control of such a winder, and how adaptive control can handle changes during operation. A nonlinear model of the winder is derived and validated first with the winder motor alone and then with the complete system. A linearized model based on it matches the nonlinear model, and a parameter sensitivity study shows that the reel radius has a dominant impact on the system's dynamic behavior. To make parameter estimation more practical, a simplified model with fewer parameters is proposed. Because the reel radius changes as paper is wound, the controller must adapt. The chosen method estimates model parameters online using the recursive least squares (RLS) algorithm, which updates its estimates as new measurements arrive. Simulations in MATLAB/Simulink with added noise indicate that filtering is required for reliable estimation. The estimation algorithm is then implemented on the winder using the dSpace real-time platform. Several excitation signals are tried; a sine signal gives the best results. Four controllers are implemented and tested in simulation and on the real system: two non-adaptive PIDs (one manually tuned, one model-based) and two adaptive controllers (an adaptive PID and an adaptive pole-placement controller). Simulations agree well with experiments. The adaptive pole-placement controller can regulate the winder, but the model-based PID achieves better performance, mainly due to the amount of measurement noise in the system.
Mange produkter fremstilles af en lang, sammenhængende bane af materiale (en web), for eksempel papir ved avistryk. I en opruller afvikles papir fra en rulle, føres gennem valser og spoles op på en anden. Specialet undersøger, hvad der mest påvirker styringen af en sådan opruller, og hvordan adaptiv styring kan håndtere ændringer under drift. Der udledes en ikke-lineær model af systemet, som valideres først på oprullermotoren alene og derefter på det samlede anlæg. En lineær model baseret herpå stemmer overens med den ikke-lineære, og en parametervariationsanalyse viser, at rullens radius har størst betydning for systemets dynamik. For at gøre parameterestimering mere håndterbar præsenteres en forenklet model med færre parametre. Fordi radius ændrer sig, mens papiret spoles op, skal regulatoren være adaptiv. De nødvendige modelparametre estimeres løbende med recursive least squares (RLS), en metode der opdaterer sine estimater, når nye målinger kommer til. Simulationer i MATLAB/Simulink med tilføjet støj viser, at filtrering er nødvendig for pålidelige estimater. Estimeringsalgoritmen implementeres derefter på anlægget via dSpace realtidsplatformen. Flere eksitationssignaler afprøves; et sinussignal fungerer bedst. Fire regulatorer implementeres og testes i simulation og på anlægget: to ikke-adaptive PID'er (en manuelt trimmet og en modelbaseret) og to adaptive (en adaptiv PID og en adaptiv polplaceringsregulator). Simulationer og forsøg stemmer godt overens. Den adaptive polplacering kan styre oprulleren, men den modelbaserede PID giver bedre ydeevne, primært på grund af målerstøj i systemet.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords