At the beginning of the diploma work we tested the hard- and software which was to control our helicopter later on. The software RealLink32 is made to be used with Matlab and Simulink. It is a simple creation tool for controllers and the handling is rather easy. The advantage of the short sampling time was the design in the almost continuous domain, instead of time discrete domain. For the design of the first few controllers we used the transfer function taken from older diploma and project work. During the real time simulation on the real helicopter we recognized that the transfer functions used were very inexact. We created new block diagrams for the main- and the back rotor with the help of the FIT-script. With these simplified block diagrams we succeeded in designing a state feedback with an observer. The main rotor controller from the 1995 diploma work could not be improved with a regular controller. Instead some new features were added to the back rotor. Its reaction for nominal value changes are faster with almost no overshooting.

It was our job to create a discrete control-system for a seesaw. It was based of an old project from 1997. In the first part, we analysed the old project to get an overview what has been done. Afterwards we got into the field of digital signal processing (DSP).

We took the mathematical modell of the seesaw from the control-system lessons. The internal controller could have been taken from the old project. We had to develop the external controller. First we had to measure the ultrasonic sensor and to develop a calibrate circuitry. The external controller gave us some problems, because it wasn't possible to find the right adjustment.

The problem could be solved with the state feedback configuration. Now it was possible to place the poles that we had a usable step response. With this controller configuration the sphere could be placed in 5s at any position. But by a large step change, the seesaw touchs the mecanical impact, although the system wasn't unstable. A solution to prevent this behaviour was to change the step into a ramp. Additional we made some consideration about adaptive controll system to optimize the transient step response. Unfortunately we hadn't enough time to implementate it in the DSP.

The behaviour of different disturbance had been checked. Non constant disturbance were adjustable.

At the end, we realised a print for the calibrate circuitry. The hardware is now completet.

Our job was the design of a discrete controller for a loading crane. The controller should prevent the load from swing in order to bring it in the shortest possible time from A to B, e.g. from a ship to a lorry.

We have chosen a state controller which is providing the possibility of independent pole assignment.The crane model based on a linearisized model (e.g. sin(x) has been replaced by x for little angles). We closed the loop by using the values Xk (position of the sledge), phi (angle of the rope) and phi'(angular speed of the rope) for feedback.

In the beginning, we used a Motorola 68332 evaluation board. Cause there were many malfunctions, we replaced it by a system of the engineering school Burgdorf using the same type of processor. This system seemed to run more stable.

As several modifications on the hardware were required, we have implemented safety mechanisms like an end switch logic for the X and the Y axe and an emergency hook brake with its supply.Furthermore, a converter and driver circuit for the processor board has been realised.

The biggest problems occurred in coherence with the angular speed of the rope. Due to an error of the algorithm, faulty values were generated, especially at the points with values of the angular speed near zero.After the correction of this error, the angular speed feedback caused in best case a slow down of the system; in worst case, the system was getting unstable. For a better behaviour, we cut the angular speed feedback line. The system behave now fast and stable.Unfortunately, we didn't have the time to determine the problem that caused such a behaviour.

For demonstration purposes, the reference input of the sledge speed is fetched by a incremental joystick. Cause we are only able to give a position setting to the controller, we have to integrate the value from the joystick in order to achieve a speed reference.

A truck model was given, which had to run along a track, that was, a HF-line with a 1MHz sine wave signal. By means of the sensor and the electronic controller, the truck model should run along the track as precisely as possible, and be capable to correct the deviation continuously.

The diploma work consists of the following parts: improving of the controller, optimisation of the adjustable sensor structure, development of an algorithm for a discrete regulation of the truck model and of testing the algorithm.

The truck model with the sensor and the electronic controller and most of the controller software were taken over from a former semester work. I improved the sensor positions so that the non-linear signals became symmetric and could be linearized in most of the working arrange. In order to record characteristics of the sensor, e.g. step response, two separate tracks were built up in parallel so that the HF signal could be switched over from one to another.

A mathematic model was developed to simulate the truck model running and controlling along the track. Then two regulating algorithms, i.e. P-regulator and PD-regulator, were designed and compared with each other. The regulator software was written and downloaded into the microcontroller 80C537, and then adjusted to be suitable for the truck model. It was shown that the PD-regulator is faster and more stable.

In this disertation an intelligent battery-charger with Fuzzy Logic controller was developed. The Fuzzy controller guarantees a fast and safe charge. Because of the spezial charging process with short negative current-impulses during the charging process the memory effect does not occur.

Temperature and voltage of the battery are periodically measured, amplified and filtered. This signals are submitted to a personal computer where Visual Designer deals with them. Because the signals are very small a great effort has to be done to make them smooth.

Temperature, change of temperature, voltage and its change are measured and calculated every 26.5 seconds. They are transmittet via DDE (dynamic data exchange) to Matlab, a mathematics programm. Its Fuzzy tool then calculates the output signal of the Fuzzy controller and sends it back to Visual Designer which steers a voltage-controlled current source that charges the battery.

Our battery charger can charge empy or full NiCd-batteries and detect the end of the process reliably.

Noise always disturbs and passive damping measures are quite expensive and ineffective. A alternative way is active noise control, which only uses two microphones, a loud-speaker and some electronics. The principle is simple: Cancel noise with noise.

The principle of active sound control is to cancel an undesirable noise by superposing an inverse noise, which results in the destructive interference of the waves. The noise is first measured with a detection microphone. A digital signal processor calculates the desired inverse noise signal, which is afterwards fed to a loud-speaker, that is placed in the duct 1,3m downstream of the detection microphone. The calculation of the inverse noise is done with a adaptive digital FIR-filter. The adaption of the filter is based on the time-domain filtered-x LMS-algorithm (Least Mean Square), which minimises the square error of the signal from a error microphone that is placed 0,3m downstream of the loud-speaker. To include the rather intense reflections at the downstream end of the duct and the noise source, the FIR-filter should be able to model a impulse response of about 100ms, which results in a filter length of 500 coefficients at a sample rate of 5kHz.

The designed controller for random broadband noise (50-800Hz) allows the noise to be suppressed by more than -16dB (reduction of 85%). The difference of the intensity level of the noise is clearly audible.

With a controller optimised for single frequencies a total cancellation of the sound wave is achieved for most frequencies in the range between 50 and 800Hz.

We are convinced that we have done a first and useful step in active noise control at the Technikum Winterthur.Finally we introduced some possible ways for further research on this topic. This could lead to more detailed knowledge about adaptive algorithms and also a improved damping of noise.

The setting of the task for our diploma was to carry out a first control for the BIGA heat pump with Fuzzycontrol.At first the brine recirculator of the unit was carried out with a conventional controller so that this circuit can be considered as an autonomous, constant heat source like underground water. Additionally a mean was found to embed a fuzzycontroller in the existing process control system, which can be developed in an external fuzzykit.The heat pump is designed for a multiple dwelling, but there is no physical housemodell attached to the circuit. Therefor a virtual and mathematical house has been invoked in the process control system which has the capability of reacting like a realistic house. It converts physically measured quantities from the circuit, to such, that can be measured in a realistic house and calculates for instance the room temperature. Further the effect of the outdoor temperature to the indoor temperature can be simulated.Parallel to these projects a fuzzycontroller has been designed which controls the indoor temperature. It uses the information which is provided by the virtual house and controls in this way the physical circuit.

This diploma requirement contains various control positions of an electro-mechanical system in the discrete world. There was a sample of an AMIRA 300 motor available which was connected with an amplifier and joined with an AD/DA conversion card.

According to exam requirements the motor model needed to be regulated through three different methods.

Our first major objective was the achievement of a cascaded PI-Controller for the stabilization of speed and a P-Controller for the position control loop. The analysis techniques were successful through the use of these control methods.

A second major objective was the creation of a state regulation using speed signals. This requirement could also be achieved even though the simulated disturbance could not be captured satisfactorily.

As a third method of regulation a state control with observation was designed. Due to lack of time these controls could not be implemented nor tested. The readings of the analogue and digital inputs could not measure the bode diagrams satisfactorily. So the correct evaluation of the diagrams were only partially successful.

The goal of our final exam work was, to let a suitable object float in a dynamic variable position, and keep it there. This work was a continuation of an earlier work Machbarkeitsstudie schwebende Kugel (Wl_PA1_97_2), in which we have found a model of the controlled process, a suitable coil, a ball and an idea for a sensor. We still had to define, evaluate and if necessary construct components such as the mechanical parts, the power amplifier and the controller. This as a result of our theoretical considerations, simulations and calculations.

The proposed sensor was not suitable, because of its inexact details from the manufacturer. Leading to the consequence, that we had to evaluate a new sensor for the measurement of the position of the ball. We decided to take a lasersensor and developped an adapter for it. We have built a suitable power amplifier and optimized the ball from our last work. After finishing the mechanical construction, there was not much time left for the controller itself.

We managed the stabilization of the system with two different structures of the controller. One of them is much better and can hold the ball over a wider range of position than the other. We also collected our first experiences with adaptive controllers.

Abstract

In the laboratory of control engineering stands a 5-axial handling robot called SMART-ARMS. The task for a later project is to develop a line control for the robot by neural networks. The aim of our final year project is to prepare the robot for this use.

At first we traced the signal paths and recorded the existing electronics. Then we modified the power amplifiers, recorded the swivel ranges of the axis and analysed the motors. For the five axis we programmed digital controllers. For this we used a program that was designed during an earlier semester project. At last we designed a program in Pascal, that allows to drive a simple line.

Performand StatusThe connections of the plugs and the Boards are listed.A wiring diagram exists for every electronic board.The dates of the swivel ranges, the motors and the conversion factors can be looked up.In consideration of the limits it is possible to drive any line in the y-z-plane.All axis are controlled by a digital control unit.