The goal of our final project was to enable a metal ball to float freely in an electromagnetic field. The position of the ball in the three-dimensional space is determined by a video-camera and an image-processing-system. The field is produced by a electro-magnet above the ball. Our first task was to put the camera system into operation and ensure that the values from the image-processing system could be read over the serial interface of the PC into the Real-Time-Tool Real-Link. To dimension our controller, a model for the plant had to be found. Since we obtain sampled values of the camera system, our plant has to be discrete. With the help of the Root-Locus-Tool of Matlab we dimensioned a discrete compensator. Upon realising that a compensator is not suitable for a 6th order complex system, we were obliged to design a state-controller with observer, which has the advantage of including several control states. In order to regulate our non-linear system with an adaptive state-controller, we consider all signals during a cycle as constant and calculate the feedback signal coefficients again for each cycle. The simulation with the adaptive state-controller ran satisfactorily; however, it was not applicable for the Real-Time-Tool (the automatic controller-algorithm demanded too much computing time). To solve this problem, we calculated feedback signal coefficients for numerous working points in advance and entered these into look-up-tables. The controller with the look-up-tables had to be implemented in C so as to integrate it in the Real-Time-Tool. Although a correct controller action was evident, the ball did not float as required. Disappointingly close to reaching our goal, the necessary time for further examinations was simply missing. Although crowning success evaded us, the project nevertheless set a highly interesting challenge which acquainted us with different automatic controller types and simulation tools - with all their artful tricks!

Our aim in this diploma was to design a control system for a model of an 18-wheeler truck. It should be able to maneuver the truck independently backwards with a coupled semitrailer. The routing is given by a guidance conductor which is laid on the floor. We are satisfied with the achieved result. The control system is capable of follow straight lines and curves with a radius of up to 3 meters represent no problem for the control system.

The work showed, how much a control system technician has to deal with the object itself which is to be controlled. Before the actual design of the control system can be begun, there are different problems to be solved which do not seem to belong to the control systems technology. Therefore, in this work we improved our knowledge in different disciplines.

The idea of this practical diploma is to develop a platform driven by four propellers. The so-called FlyCat is installed cardanically in a protected zone, where it is able to rotate around its three axes freely. The task is to control pitch, roll and yaw angles as well as the main thrust of the platform. The FlyCat is equipped with four electric motors, transmissions and propellers. Car batteries are used to supply the engines. A commercial remote control serves as command input. Now a discrete state-space controller is to be implemented with a digital signal processor, which is able to control the given system. The sensors as well as their data analysis are to develop for this purpose. A further goal of this work is to implement a data link between the DSP and a PC, which makes it possible to show the actual platform data in real time on the screen.

The tasks of this thesis could successfully be performed. All angles mentioned and the main thrust can be controlled very satisfyingly. The data link to the PC as well as a three dimensional representation of the platform position on the screen were carried out. The development flow ran usually straight-lined, that means our ideas always turned out to be realistic, nevertheless some unexpected problems occurred. Due to mechanical vibrations we were forced to filter the sensor signals. This caused signal delays, which disturbed the controller in an unacceptable way. We found the solution in the use of an estimator which brought an enormous improvement of the controlled system behavior. The platform is operated in a closed space, this partially causes strong turbulence. These disturbances only can be corrected moderately with the designed controller/sensor system. Nevertheless we got a deep view of a very broad spectrum, which reaches from control engineering, digital signal processing over filter theory up to microelectronics.

Our task was to develop a controller for an industrial robot.

To find a good mathematical model of the system, we made an identification with realistic measuring signals. It turned out that the transfer functions found from the identifications indicate large deviations with higher frequencies. The reasons for these deviations are the different measuring signals, the limitations of the under-stored control loops, the different weights of the grip arms and measuring at different Top-Loaders. In order to develop a controller, we selected a transfer function, whose behaviour was average compared with the other ones. We studied different controller types, because we had to wait very long to start with our first measurements and because we had not known then, how critical the position regulation would be. Mainly we examined the controlling with a proportional controller, a shortening controller and a state controller with observer. For the implementation of the controller we decided for a PC-based solution, since one is very flexible with the organisation of the controller and one has the possibility to do trajectory calculations. After the first measurements with a proportional controller it turned out that the position regulation is good. A trajectory must be given as reference signal that the joints move in the desired way. The course of this curve is given by points in space (teach-in). To fit a curve through these points, we used different curve fitting procedures. In order to follow this curve, the angles of the two joints must be calculated from the coordinates. Both, the position regulation and the movement of the joints along the trajectory, had been tested with self-made simulation programs.

During the second measurement, different test runs with the system were made. The position controller supplies good results and the movement along the trajectory corresponds with our theoretical calculations.

The task of this project was to control two mechanically connected slides, operated by DC-motor, to a constant velocity. The slides carry a mirror system which assists the digital exposure of photographic prints. The targets were, that within 100ms, the fast slide should accelerate to its running velocity of 6 inch/sec. This should be controlled within a tolerance of 0.5%. Apart from these targets, it was also in the center of our work to determine realizability of this new system at all (the previous solution has a stepper motor for its drive).

For the implementation of a controller we chose a DSP. For velocity measurement, we used a scale system with an incremental sensor, which generated firstly a position signal of the fast slide. With a derivative algorithm, which had been programed into the DSP, we transformed the position signal into a speed signal. This should serve as the actual value for the further controller. The system identification was carried out by the bodedia-gramm and the step response, as well as the speed calculation was included into plant. The identification showed that it may relate to a system of higher order, but nevertheless marks a very dominant pole. Therefore, the plant had provisionally been considered as a first order system and the other roots (poles and zeros) disregarded. On account of the identification, we decided to assemble a control loop with a PI-controller. Measurements with this control loop showed that because of two different disturbances (motor axle, mechanical resonance), the given tolerance had not been fulfilled.

Finally, the targets had not completely been reached. To improve the values, it would be necessary to investigate the problems with the motor axle, to possibly run the motor with a higher speed and to subtilize the measurement shrewdly. Basically it can be said that operating this system by dc-motor and speed controlling is possible.

The goal of our project was to adapt a micro controlled driveystem from a sliding door to a swing door. We did this job on behalf of the company Record. Record produces drivesytems for automatic doors. As this project serves for the transfer of know-how, we were careful to apply a clearly structured conception.

We did not use Record's existing classical controller, since it has a totally different structure.

One task was to develop another classical controller. This was carried out by M.Berweger, who developed a current-controller, speed-controller and a position-controller. He managed to lay out the speed-controller in a way that only one controller was required for all applications. This controller was able to follow every speedprofile.

Another task was to control the system with a Fuzzy-Controller. The developement of this controller was undertaken by L.Bert. The Fuzzy-Controller was designed with the Tool Matlab 5.2/Fuzzytoolbox and tested with Matlab/Simulink. The tests were done on our model. Afterwards a matrix was build using the data from the Fuzzy-Controller, wich when implemented in the microcontroller simulates the Fuzzy-Controller. In this we also reached the goal of controlling all applications with only one controller.

If we compare the classical controller with the Fuzzy-Controller, it becomes evident that the Fuzzy-Controller is usable in every working point without a speed profile. On the other hand, the classical controller is able to control the follow of speed profile, whereas the drive behaviour of Fuzzy-Controller is very difficult to handle.

The Technikum Winterthur has an rpm-adjustable brine/water heat pump plant for testing purposes. The energy is drawn form an electrically heated brine cycle and transferred to cool water by a heat exchanger in a user cycle. The performance output is regulated through a mathematically reconstructed model house, simulating the user. A timed sequence of the outside temperature and sunlight can be programmed for the model house.

The installation was equipped with the new control and visualization program, LabVIEW, which offers more comfort and has more options than the existing Lab Tech. The features developed in Lab Tech have now been realized in LabVIEW. The regulating strategy and model home were tested and improved where necessary. The Sole regulator and some regulator settings were overhauled and successfully adjusted. The converted concept for indoor temperature regulation allowed the use of meteorological data and weather forecast. Subsequent measurements either met or exceeded past efforts.

The new LabVIEW software provides a reasonable base from which to take full advantage of what the heat pump installation has to offer. All measure- and output data can be saved to a file or monitored online by utilizing graphs.

The Institute of Hygiene and Applied Physiology of the Swiss Federal Insitute of Technology Zurich investigates the accommodation of the human eye. A so called optometer has been developed for this purpose, by means of which the accommodation of the human eye is recorded continuously. An infrared diode shines light into the eye. The light is reflected on the retina and projected out of the eye by means of the dipotric system so to form an image in front of the eye. A sensor is used to detect the focal point of that image and to quantify the amount of accommodation. An exactly measurement is not possible actually because the electronic controller used, is unstable. In a previous exposition, a big delay time of the system has been identified. Therefore, a controller with a predictor has been proposed to eliminate the delay time. The aim of this study was to reproduce the identification of the controlled system, as well as to implement the new controller with predictor. With a new concept it was possible to minimise the delay-time and to realize a faster controller. In a next step, further experiments will be performed to optimize the new controller and to compare with the old one.