The thesis investigates a high speed switched reluctance (SR) generator suitable for applications such as aerospace and turbo-charged ground vehicles. The generator is two-phase with 16-8 stator-rotor poles. The stator is 70 mm OD by 25 mm core length made of soft ferrite. The rotor comprises 30 pm laminations of amorphous alloy chosen for its mechanical strength. The rotor of the generator was mounted on the protruding shaft of an air turbine and extensive tests were performed at a variety of speeds and supply voltages. The generator is shown to be capable of delivering around 280 Watts to a resistive load at 60,000 rpm, and 60 volts, with an estimated efficiency of around 77%. The thesis describes in detail the electromagnetic and mechanical construction of the generator and presents a comprehensive survey of operating waveforms and performance measurements. The design utilises a simple magnetic model for the SR generator based on a quasi-linear flux-current characteristics which are shown to provide a very useful tool for performance prediction.

The aim of this research project was to develop a vector controlled induction motor drive operating without a speed or position sensor but having a dynamic performance comparable to a sensored vector drive. The methodology was to detect the motor speed from the machine rotor slot harmonics using digital signal processing and to use this signal to tune a speed estimator and thus reduce or eliminate the estimator’s sensitivity to parameter variations. Derivation of a speed signal from the rotor slot harmonics using a Discrete Fourier Transform-based algorithm has yielded highly accurate and robust speed signals above machine frequencies of about 2 Hz and independent of machine loads. The detection, which has been carried out using an Intel i860 processor in parallel with the main vector controller, has been found to give predictable and consistent results duing speed transient conditions. The speed signal obtained from the rotor slot harmonics has been used to tune a Model Reference Adaptive speed and flux observer, with the resulting sensorless drive operating to steady state speed accuracies down to 0.02 rpm above 2 Hz (i.e. 60 rpm for the 4 pole machine). A significant aspect of the research has been the mathematical derivation of the speed bandwidth limitations for both sensored and sensorless drives, thus allowing for quantitative comparison of their dynamic performance. It has been found that the speed bandwidth limitation for sensorless drives depends on the accuracy to which the machine parameters are known and that for maximum dynamic performance it is necessary to tune the flux and speed estimator against variations in stator resistance in addition to the tuning mechanism deriving from the DFT speed detector. New dynamic stator resistance tuning algorithms have been implemented. The resulting sensorless drive has been found to have a speed bandwidth equivalent to sensored drives fitted with medium resolution encoders (i.e. about 500 ppr), and a zero speed accuracy of ± 8 rpm under speed control. These specifications are superior to any reported in the research literature.

The matrix converter is an attractive topology of power converter for the Aerospace Industry where factors such as the absence of electrolytic capacitors, the potentiality of increasing power density, reducing size and weight and good input power quality are fundamental. The matrix converter potential advantages offers the possibility to achieve the aim of the More Electric Aircraft research which intends to gradually re- place, from the aircraft architecture, the hydraulic power source and its infrastructure with electric power generation and a more flexible power distribution system. The purpose of this work is to investigate the design and implementation of a 40kVA matrix converter for an Electro Mechanical Actuator (EMA) drive system. A SABER simulation analysis of the candidate matrix converter drive systems, for this application, is provided. The design and implementation of the matrix converter is described, with particular attention to the strict requirements of the given aerospace application. Finally, the matrix converter PMSM drive system and the EMA drive system are respectively assembled, tested and commissioned.

The magnetic field within electrical machines causes an interaction between the electrical and mechanical dynamics of the system. In the simplest cases, when the rotor mean position is central in the stator, the interaction manifests itself mainly as a negative stiffness between the rotor and the stator. When the rotor mean position is offset relative to the stator, then components of force arise whose frequency in the stationary frame is twice the electrical frequency of the supply. For induction machines in particular, both the electrical system and the mechanical system may be quite complex dynamically in the sense that over the range of frequencies of interest, it is necessary to consider a number of degrees of freedom in both the electrical part of the model and the mechanical part. This work sets out a structured and formal approach to the preparation of such models. Each different combination of voltage and slip is examined separately. In each case, the first step is to compute the steady-state reference solution for machine currents as a function of time. Then, the electro-magnetic behaviour of the electrical machine is linearised around that reference solution. The result is a linear time-dependent model for the electromagnetic behaviour which is then easily coupled with a linear model for the mechanical dynamics. The mechanical dynamics are usually stationary. Floquet methods can then be applied to determine whether the system is stable and the response of the system to mechanical or electrical perturbations can be computed quickly. The analysis method is applied to a particular three-phase induction machine which has parallel paths integrated into its winding structure in the sense that each of the phases is split into a "Wheatstone-bridge" arrangement following. Currents passing diametrically through a phase in the vertical direction account for the main torque-producing components of stator field. Currents passing diametrically through the phase in the horizontal direction account for transverse forces. The parallel paths can be switched to open-circuit or closed-circuit without affecting the torque-producing function of the machine and all of the stator conductors contribute to torque-production.For a number of combinations of voltage and slip, the machine is stable irrespective of whether the parallel paths are open-circuit or not but the effective damping of the machine for synchronous vibration is shown to be much higher with the parallel paths in closed-circuit.

This thesis concerns the design and construction of a closed-loop controlled matrix converter induction motor drive, using transputer parallel processors. The modulation algorithms used for the matrix converter are described. A 2.5 kW experimental matrix converter using IGBT switching devices has been constructed and tested. An analysis of the losses in the converter has been carried out and this gives good agreement with the measured losses. Two modulation algorithms, the Venturini algorithm and the scalar algorithm have been implemented in real-time on a network of parallel transputer processors. Experimental results are presented to compare the operation of these two algorithms. Open-loop constant V/F control of the matrix converter induction motor drive has been demonstrated. A controller has been designed to achieve closed-loop speed control of the drive system, employing the slip regulation technique. The experimental results under various operating conditions have verified the correct operation of both control systems. The indirect vector control technique has also been implemented. The results demonstrate the steady-state and transient performance as well as the regenerative operation of the drive system. The application of a matrix converter to a high performance induction motor servo drive rated at 2.5 kW with true four quadrant capability and minimum passive components has been demonstrated.

Permanent magnet motors have a series of characteristics that make them attractive for the use in industrial drives: low maintenance, high dynamics, small size and mass to power ratio. In particular its higher efficiency means that permanent magnet synchronous motors may be used instead of electro-magnetically exited motors (such induction machines or commutator DC motors) in applications where the energy savings compensate the higher initial cost. Nevertheless, the need for a shaft mounted position measurement to perform the orientation of the control of the synchronous machine is of concern, because it increases the total drive cost and reduces reliability. In this work the sensorless vector control of a surface mounted permanent magnet machine is presented. The emphasis is in the control at low and zero speed, including position control, by means of saturation saliency tracking. Two different strategies for rotor position detection used in salient synchronous machines and in induction machines are analysed. These are hf voltage injection in the stationary, stator, reference frame of the machine (α-ß injection) and hf voltage injection on the estimated rotor axis (so called d-axis or pulsating injection). These two methods are optimised for its application to the surface mounted PM machine. The small magnitude of the saliency present difficulties and disturbances are significant. A commissioning based method (SMP) is used for enhanced rotor position estimation by the α-ß rotating injection. The two methods are implemented on a 4 kW experimental rig and the sensorless controlled results are compared and discussed. A hybrid structure combining the saliency tracking method with a flux-observer is also presented and provides sensorless control capability over the whole speed range.

This thesis describes the design and construction of a high performance induction motor drive, controlled by a network of parallel (INMOS) Transputer processors. The flexibility and high computational ability of the controller is demonstrated by the implementation of three forms of indirect vector control for the induction motor (here termed "V-Type, "V-Type with Current Feedback" and "I-Type") on two motor drive rigs. Results show that V-Type control with current feedback is superior, and that on-line parameter estimation (namely the rotor time constant) is required. The controller has been expanded to incorporate two parameter identification strategies for assessment. The first, termed "Reactive Power Measurement", has proved successful in matching the controller value of rotor time constant to the actual machine value of rotor time constant. The second, termed "PRBS Injection with Cross- Correlation" has proved inconclusive and is the subject of on-going research. The performance of the transputer parallel processing network for real time control is discussed. This assessment is felt to be significant since parallel architectures are likely to become increasingly exploited as the processors become cheaper, more powerful and flexible, and with enhanced system support.

Similar to a human observer, an automated image vision system is able to recognise most parts of an object if the system could accurately trace and reflect its true shape. This has prompted the development of the many diverse edge detection techniques. Neural networks have been successfully applied to pattern recognition tasks and edge detection. However, there is a great necessity to analyse neural network models so as to achieve close insight into their internal functionality. To this purpose, a new and general training set, consisting of a limited number of prototype edge patterns, is proposed to analyse the problem of neural network edge detection. This thesis also proposes two approaches to the skin lesion image segmentation problem. The first is a mainly thresholding segmentation method where an optimal threshold is determined iteratively by an isodata algorithm. The second method proposed is based on neural network edge detection and a rational Gaussian curve that fits an approximate closed elastic curve between the recognized neural network edge patterns. A quantitative comparison of the techniques is enabled by the use of synthetic lesions to which Gaussian noise is added. The proposed techniques are also compared with an established automatic skin segmentation method. It is demonstrated that for lesions with a range of different border irregularity properties the thresholding segmentation method provides the best performance over a range of signal to noise ratios; the thresholding segmentation method is also demonstrated to have similar performance when tested on real skin lesions.

A human observer is able to recognise and describe most parts of an object by its contour, if this is properly traced and reflects the shape of the object itself. With a machine vision system this recognition task has been approached using a similar technique. This prompted the development of many diverse edge detection algorithms. The work described in this thesis is based on the visual observation that edge maps produced by different algorithms, as the image degrades. Display different properties of the original image. Our proposed objective is to try and improve the edge map through the arbitration between edge maps produced by diverse (in nature, approach and performance) edge detection algorithms. As image processing tools are repetitively applied to similar images we believe the objective can be achieved by a learning process based on sample images. It is shown that such an approach is feasible, using an artificial neural network to perform the arbitration. This is taught from sets extracted from sample images. The arbitration system is implemented upon a parallel processing platform. The performance of the system is presented through examples of diverse types of image. Comparisons with a neural network edge detector (also developed within this thesis) and conventional edge detectors show that the proposed system presents significant advantages.

Three techniques are used to introduce wire models into Transmission Line Matrix (TLM) Electromagnetic field solvers. The first involves the setting up of a one dimensional time-domain transmission-line model of a pair of wires. Voltages and currents coupled into the wires can be obtained from the incident fields. The differentials of these fields are used as source terms of the one dimensional model. Results obtained by this model were compared with frequency domain results and good agreement was found. Non-linear loads can easily be modelled, and the response of a simple non-linear device is presented. Wires of realistic radii can only be modelled in 2- or 3-dimensional TLM by using a fine mesh. Using progressively finer meshes the modelled radius converges to an unexpected result. The problem lies with the two forms of the TLM mesh, each giving a different modelled radius. This can lead to velocity of propagation errors of up to 10%. A short-circuit node was developed but this has poor propagational characteristics and a radius equal to half the mesh spacing. Diakoptics was re-examined and some operating conditions suggested. Frequency Domain Diakoptics was also introduced and tested. Time Domain Diakoptics was used to create blocks of space containing wires which could be pre-solved for future use. Since a very fine mesh was used to describe the wire, but a coarse mesh was used for the remainder of the geometry there is a very crude approximation between the two meshes. This effect meant that long wires could not be constructed out of a multitude of pre-solved short-wires. A longer wire could be pre-solved as a single element and a reasonable modelled radius obtained. The technique became unstable after about 750 iterations. Comparisons between the wire models developed show reasonable agreement. The one-dimensional method was the easiest technique to use.