Languages and Tools for Optimization of Large-Scale Systems

Abstract: Modeling and simulation are established techniques for solving design problems in a wide range of engineering disciplines today. Dedicated computer languages, such as Modelica, and efficient software tools are available. In this thesis, an extension of Modelica, Optimica, targeted at dynamic optimization of Modelica models is proposed. In order to demonstrate the Optimica extension, supporting software has been developed. This includes a modularly extensible Modelica compiler, the JModelica compiler, and an extension that supports also Optimica. A Modelica library for paper machine dryer section modeling, DryLib, has been developed. The classes in the library enable structured and hierarchical modeling of dryer sections at the application user level, while offering extensibility for the expert user. Based on DryLib, a parameter optimization problem, a model reduction problem, and an optimization-based control problem have been formulated and solved. A start-up optimization problem for a plate reactor has been formulated in Optimica, and solved by means of the Optimica compiler. In addition, the robustness properties of the start-up trajectories have been evaluated by means of Monte-Carlo simulation. In many control systems, it is necessary to consider interaction with a user. In this thesis, a manual control scheme for an unstable inverted pendulum system, where the inputs are bounded, is presented. The proposed controller is based on the notion of reachability sets and guarantees semi global stability for all references. An inverted pendulum on a two wheels robot has been developed. A distributed control system, including sensor processing algorithms and a stabilizing control scheme has been implemented on three on-board embedded processors.