The function of this diploma consisted of implementing a speed control for a mirror system. The mirror system consists of two mechanically coupled carriages, which are propelled with an DC engine. The system is used in the area of Photofinishing (production of pictures from an exposed and developed film). The target was to accelerate the two carriages within 100 ms on their desired velocity of 3 Zoll/s and 6 Zoll/s. The deviation from the desired speed may amount to max. 1%. Besides a new developed card for the measurement of the velocity should be tested and put into operation.

In an earlier work the objective could not be achieved. With a new measuring concept for the measurement of the velocity the specification should be better achieved now. In a first step the card was tested, improved and put into operation. Subsequently, we set up a physical block diagram and executed the parameter identification by different measurements on the system. The found parameters could be acknowledged by means of check measurements. The block diagram enabled us to simulate different control algorhitm. We decided for two switchable PI controllers, which were implemented in a DSP. The first PI controller accelerates the carriages on the desired speed, while the second controller is responsible for the constant velocity. The implementation was made more difficult by mechanical oscillations. The oscillations could be absorbed by improvements on the mechanical system as well as a notch filter on the controller output. The given tolerance could not be kept due to the remai-ning mechanical disturbances. We achieved a tolerance of 2% and an acceleration phase of 100 ms.

The goal of our degree dissertation was to judge and optimise the control loops of the preceding project works Nr. PAILk0005/06 and Nr. PAILk0015/16. In addition, the disturbance reaction had to be regarded separately from the reference reaction and be improved with a overlaid integrating control loop and a disturbance estimator. One in the institute for automatic of the ETH Zurich developed software for computer supporting identification, design and implementation for control systems should be implemented in the RT-Laboratory and be tested with the three-mass oscillator. In a further step it applied to examine new regulation concepts like the Riccatti design and the robust controller design and as far as possible to implement them.

The model of the plant of the project works was complemented, which supplied better results by the identification of each disk. Thereby the simulation and that real model corresponded better, whereby the reference reaction of the controllers could be improved.

The disturbance reaction of the controllers could be improved by overlaying a integrating control loop. Without integrating control loop, the controllers weren't able to compensate the influence of a disturbance. With an overlaid integrating control loop the disturbance could be compensated after a short time. The disturbance estimator could not be implemented with the chosen point of attack of the disturbance, because the disturbance estimator wasn't controllable.

The Riccati design supplied another state feedback matrix k for the state regulator. The Riccati design can be used for many control problems and supplies a good basic adjustment. We received a good state regulator, which we optimised. In comparison with the other controllers, this design indicates a somewhat worse reference reaction.

The software of the ETH is not compatible with our operating system in the lab. Therefore we couldn't transfer them.

The Top Loader is a robot designed to pack chocolates or things with similar sizes in cardboard boxes. To control the angle positions three cascaded loops are used for each of the two axis. The closed loop controllers for current and speed are integrated in the power amplifier. For the overlaid position control loop they use a device from Adept Technology, inc.

First I verified the parameters of the speed loop. Next step was to copy the existing controller algorithm to the xPC Target system. This is important for the following work: Now it's possible to carry out measurements depending on special command signals or even whole trajectories.

But the main task of this work was to develop a program which calculates an optimal trajectory. Optimal means in this relation the Top Loader has to move between the given points as fast as possible without cause damage or loosing the transported cargo (the Top Loader holds them pneumatically). To solve this problem it is important to know the limits of acceleration and speed. I got these values by measurements on both axis. A trajectory, calculated with the developed program can be carried out 43 times a minute without exceed the limits described above.

The existing control algorithm is not a very good solution for controlling non-linear plants like robots. The last task thus was to improve or replace it. Studying the possibilities, I discovered there are plenty different schemes. So I decided two write a short overview with the main features of each scheme instead of realising one (maybe not suitable) concept.

The company Sulzer Hexis works on an extensive development project in the field of fuel cells at present. These shall be used for the heating of detached family houses on the one hand an for the rationing of hot-water tanks on the other. Our dissertation now deals with some selected partial problems in connection with the tank heating. The work is restricted to a simulation with a model of the tank because the total system of the tank heating is not completely developed yet.An additional task is to infer the present consumption of hot water from only one temperature sensor in the tank, so that the strategy to control the tank heating as well as the manager for the fuel cell can rely on a prediction of this consumption. The problem can be solved either with a neuronal network or with a system with observer. However, the verification of the prediction is impossible because there is no other self-contained control variable. In addition we are confronted with the problem of a prediction on consumptions which are changing extremely abruptly and which are temporally not correlated. Therefore we dismissed a control strategy with a prediction on consumption. For the manager of the fuel cell a profile of the pump in the heating circuit is much more reasonable than a prediction of the consumption, which is therefore not needed for this. Much better results were reached with a strategy which incorporates the temperature in the forerun and the return of the heating coil. With this information and the temperature of a precise position in the tank we can infer the capacity of energy in the tank. This is more meaningful than the consumption derived from the temperature of the tank. The controller itself is a simple fuzzy controller which is operating without any prediction. However, this new strategy is dependent on a precise position of the temperature sensor and can only be adapted to other tanks with readjustments of the fuzzy sets in the controller.

The goal of the degree dissertation 'The Labyrinth' was to control a ball on a platform which can be toppled in two axes. Beside the realisation of the hardware, the main tasks were to find a suitable controller structure, to develop a simulation and to issue a graphical user interface.

The design of the controller was the biggest challenge. Due to mechanical friction the present system contains some non-linear correlations. Such frictions make great demands on the quality of the controller while linear, unstable systems are relatively easy to control. This might surprise at first sight because in this matter quite the opposite is true as far as human abilities are concerned. A classic controller structure on four levels was realised. Four controllers had to be developed to stabilize the system: two for the angle of the platform and two for the position of the ball.

Generating the desired values by a computer gave us the possibility to increase the performance of the system by changing the software. Measurements showed that the type of transferring the desired values to the controller is decisive for the quality of the process. In addition it was possible to design a suitable graphical user interface and a simulation in Matlab.

Further inquiries were done with the xPC-tool, which permits fast prototyping with a computer. This enabled us to work more efficiently.

The present dissertation allows the operation of the diverse modules as a whole system, which fulfills most of the requirements of the specifications. Furthermore the system may serve as a basis for further investigations in different areas of interest.

In this dissertation a navigation problem was treated. An automatic controller was to be found which operates as a kind of "electronic anchor" and keeps a ship in position in a suddenly occurring current. This problem had to be solved with the help of control engineering.

The idea to this dissertation grew out of a real problem: A ship has to wait on a river in front of a lock, keeping its position by withstanding the attacking currents.

The feasibility study to this topic, made before in a work on the project, served as the basis for the implementation of this project. Thereby the automatic controller should fulfil its function only with the help of position data. A simulation model built in Matlab Simulink and a model ship in a basin with flow mechanism were used for the experiments.

The investigations and experiments showed that it is very difficult to find an automatic controller which yields satisfying results for all flow situations from standing water up to strong current. Even if in the computer simulation such an automatic controller could be found, some problems showed up with the implementation in the real model ship. The consequence of this was an automatic controller, which supplies satisfying results at least for a specific situation of the current.

The laboratory model "magnetic suspension" produced by Amira essentially consists of an electromagnet and an arrow-like object floating beneath. The plant is unstable and nonlinear. In an earlier project work discrepancies between the present hardware and the specification of the manufacturer had occurred. For this reason a careful identification of the different components (plant, actuator, measuring element and built-in PID-controller) were done. The force action of the electromagnet on the arrow in function of coil current and distance was measured and approximated by a function. From these measurements a differential equation of the plant could be derived by linearization. Thanks to controller found empirically in an earlier project work, the transfer function of the plant could be measured directly and compared with this derivation. It was shown that an identification of the plant is possible without a controller with sufficient accuracy.

The mentioned discrepancies could be recovered by the adjustment of the conversion formulas in the DT-part of the PID controller. Additional views of stability by Maple acknowledged the simulations. An external controller and a digital controller acknowledged the modifications in the reality. The remaining components corresponded to theoretical expectations in the reality without adjustments.

The measured frequency response of the closed control loop corresponds with the simulation except a resonance pronounced very strongly in the reality.

This diploma thesis represents a solid basis for further projects and practical courses.