Thesis Control predictivo de un sistema de levitación magnética
Loading...
Date
2025-10
Authors
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Electrónica
Departament
Campus
Campus Casa Central Valparaíso
Abstract
En este trabajo de titulación se presenta la aplicación de un método avanzado de control de procesos, el Control Predictivo por Modelo (MPC). Esta técnica consiste en implementar un control que utiliza un modelo matemático de la planta para predecir su comportamiento futuro y, en función de ello, optimizar las acciones de control que se aplicarán al sistema. La planta bajo estudio corresponde a un sistema de levitación magnética, compuesto por un electroimán que actúa sobre un rotor suspendido. Este sistema es inherentemente inestable y no lineal, lo que lo convierte en un caso de interés para la implementación de técnicas avanzadas de control. Para facilitar el diseño, se realiza un proceso de linealización y se estudian las propiedades de controlabilidad y observabilidad de la planta. El método de optimización seleccionado para resolver el problema de MPC es el algoritmo Active-Set, implementado en el solver quadprog de MATLAB. Se desarrollan simulaciones en MATLAB/Simulink con el fin de verificar el correcto funcionamiento del bloque de control MPC. Posteriormente, en la implementación física se evalúa el desempeño del controlador variando el horizonte de predicción, el periodo de muestreo y las restricciones de actuación, utilizando hardware de adquisición de datos y bloques de Quanser.
This work presents the application of an advanced process control method, Model Predictive Control (MPC). This technique consists of implementing a control strategy that uses a mathematical model of the plant to predict its future behavior and, based on this prediction, optimize the control actions to be applied to the system. The plant under study corresponds to a magnetic levitation system, composed of an electromagnet that acts on a suspended rotor. This system is inherently unstable and nonlinear, making it an interesting case for the implementation of advanced control techniques. To facilitate the design, a linearization process is carried out and the controllability and observability properties of the plant are analyzed. The optimization method chosen to solve the MPC problem is the Active-Set algorithm, implemented in MATLAB’s quadprog solver. Simulations are carried out in MATLAB/Simulink to verify the correct operation of the MPC control block. Afterwards, in the physical implementation, the controller’s performance is evaluated by varying the prediction horizon, the sampling period, and the actuation constraints, using data acquisition hardware and Quanser blocks.
This work presents the application of an advanced process control method, Model Predictive Control (MPC). This technique consists of implementing a control strategy that uses a mathematical model of the plant to predict its future behavior and, based on this prediction, optimize the control actions to be applied to the system. The plant under study corresponds to a magnetic levitation system, composed of an electromagnet that acts on a suspended rotor. This system is inherently unstable and nonlinear, making it an interesting case for the implementation of advanced control techniques. To facilitate the design, a linearization process is carried out and the controllability and observability properties of the plant are analyzed. The optimization method chosen to solve the MPC problem is the Active-Set algorithm, implemented in MATLAB’s quadprog solver. Simulations are carried out in MATLAB/Simulink to verify the correct operation of the MPC control block. Afterwards, in the physical implementation, the controller’s performance is evaluated by varying the prediction horizon, the sampling period, and the actuation constraints, using data acquisition hardware and Quanser blocks.
Description
Keywords
Control predictivo por modelo (MPC), Levitación magnética, Algoritmo Active-Set, MATLAB / Simulink, ObservadordeEstados, MagneticLevitation, State Observer
