Adaptive control with self-tuning for center-driven web winders

Studenteropgave: Speciale (inkl. HD afgangsprojekt)

  • Alexander H. Andersen
  • Thor L. Lemminger
  • Mads P. Z. L. Andersen
2. semester, Elektromekanisk Systemdesign, Kandidat (Kandidatuddannelse)
Many industries require manufacturing with a continuous long piece of material, a so-called web. An example of this is the manufacturing of our daily newspaper. Here a long string of paper is feed to a rotating printing press. The paper is often unrolled from a large reel of paper and distributed through series of rollers and winded to another reel. This leads to the thesis initial problem stated as: What influences the control of the web winder system? A nonlinear model is derived for the system and it is validated partly by experiments with the winder motor by itself and later by test with the entire system. A linear model is derived on basis of the non-linear model, and it is found that the two models are in agreement. A parameter variation analysis is made on basis of the validated linear model. The result of this analysis is that the radius of the winder reel is of great importance to the dynamic behavior of the system. A simplified model with fewer parameters is presented, to ease the task of parameter estimation in the system. From the problem analysis it is concluded that a given controller should be able to adapt to the system due to the changes in radius. The problem is stated as: How is it possible to utilize adaptive control for the given web winder system ? To be able to adapt the controller to the web winders changing behavior its model parameters have to be estimated. It is chosen to use the recursive least square algorithm. The main principle of the least square algorithm is described. To ensure that the parameter estimation have a satisfying performance it is tested by means of simulation in MATLAB Simulink. To mimic the real web winder system noise is added, when the noise is added to the simulation is concluded that filters are necessary to meet the requirements for proper estimation. The estimations algorithm is implemented in the web winder system by means of dSpace, different excitations methods are tested and it is found that a sine signal is best suited. Four different controllers are implemented in the system two non adaptive and two adaptive. The non adaptive controllers are both PID controller, one is based on iterative manual tuning, the other is model based. The two adaptive controllers are an adaptive PID and an adaptive pole placement controller. The four controllers are tested through simulation and on the web winder system. Good relation between the simulations and the actual behavior is found. It is also found that the adaptive pole placement controller is able to control the web winder, the performance is though exceeded by the model based PID. The reason for this is found to be governed by the amount of measurement noise in the given system.
Udgivende institutionInstitut for energiteknik
ID: 17642392