A fully remote all optical technique for the generation and detection of ultrasonics in epoxy materials has been developed at Victoria University. This requires a high power pulsed laser to induce ultrasound in the sample and an optical fibre interferometer for detection. The output of the interferometer can be recorded using a digital storage oscilloscope and subsequently processed by a personal computer. This Thesis presents the use of this arrangement for further studies of cure monitoring of composite materials. Samples of 3.0 cm diameter and 1, 2, 4, 6, 8, 10 and 12 cm thickness were made and the interferometer waveforms were recorded in 10 min time intervals after the curing agent was mixed with the epoxy. Attenuation coefficients and travel times were calculated for each waveform. During the analysis phase of this research an algorithm was developed to find the exact onset of ultrasonic signals. Different thickness samples were made and the interferometric waveforms were obtained and analysed. The relative travel time and wave attenuations were calculated and close correlations were found between these two factors and the state of cure. In this work, optically-generated ultrasonic waves, when reflecting from the rear surface of an object under test, cause a small surface motion (typically in the range of one tenth of a nanometre to few nanometres). A second diagnostic laser is used to illuminate the rear surface of the test object. The ultrasonic surface motion produces a small phase shift or frequency shift in the reflected light, which is detected by an interferometric system. Signals from the interferometer were recorded using a digital storage oscilloscope (Tektronix DSA 602) and this stored signal is subsequently read and processed using a personal computer. In these experiments, epoxy was poured into moulds and measurements were taken every few minutes thereafter.

This thesis presents an investigation and a comparative study of four different approaches namely ANSI/IEEE standard models, Adaptive Neuro-Fuzzy Inference System (ANFIS), Multilayer Feedforward Neural Network (MFNN) and Elman Recurrent Neural Network (ERNN) to modeling and prediction of the top and bottom-oil temperatures for the 8 MVA Oil Air (OA)-cooled and 27 MVA Forced Air (FA)-cooled class of power transformers. The models were derived from real data of temperature measurements obtained from two industrial power installations. A comparison of the proposed techniques is presented for predicting top and bottom-oil temperatures based on the historical data measured over a 35 day period for the first transformer and 4.5 days for the second transformer with either a half or a quarter hour sampling time. Comparisons of the results obtained indicate that the hybrid neuro-fuzzy network is the best candidate for the analysis and prediction of the power transformer top and bottom-oil temperatures. The ANFIS demonstrated the best comparative performance in temperature prediction in terms of Root Mean Square Error (RMSE) and peak error.

This thesis describes a study of intragrating position (localised heat source) and strain measurement sensor systems. The design, development and performance of intensity (power) reflection spectrum based intragrating sensing systems employing chirped fibre Bragg gratings (CFBGs) are investigated. Intragrating sensing is the process of obtaining a continuous profile of a measurand over the grating length from either the amplitude or phase, or both the amplitude and phase of its measured reflection response. Techniques for intragrating sensing employing conventional fibre Bragg gratings have been reviewed and analysed in order to design an intragrating sensor which overcomes the problems associated with the existing sensor systems. The inability to determine the position of the disturbance, direction of the strain gradient and the broadening of the spectra together with the reduction in reflectance complicates the grating inverse reconstruction technique to recover the disturbance profile and is also computationally expensive. Thus the selection of CFBGs is desirable because each Bragg wavelength in the broadband reflection spectrum from a CFBG corresponds to particular local position along the grating. Thus a change in the reflection intensity spectrum, due to a disturbance, at a particular wavelength can be quantified to a local position in the grating and hence the spatial distribution of the measurand can be evaluated. The thesis consists of four major sections: fabrication of CFBGs, sensor calibration, intragrating position and strain measurements. Techniques for the production and controlling the spectral response of CFBGs suitable for intragrating sensing application were investigated using two fabrication techniques; the prism interferometer and the scanning phase mask system. The modelling and experimental results showed that fabrication constraints exist for designing a suitable sensor for intragrating measurements with the prism technique, and thus most of the gratings used for this research were fabricated with the scanning phase mask system. The analysis of results for strain/temperature characterisation, showed that CFBGs fabricated in hydrogenated standard telecommunications fibre have a temperature coefficient approximately 20% higher than standard FBGs in the 1550 nm region with identical strain coefficients. Thus a simple technique for strain-independent temperature measurement is proposed and demonstrated using a sensor head with a combination of a standard FBG and a CFBG. This is an additional finding of the research. The intragrating position measurements within a CFBG sensor were investigated as a function of grating strength. Five linear CFBGs of chirp rate 20 nm/cm, approximate length of 15 mm with varying levels of reflectance (6-53%), fabricated with the scanning phase mask technique were employed for the determination of the centre position of a localised heat source within a CFBG. An iterative approach in conjunction with a fast Fourier transform (FFT) was implemented for solving the inverse problem of obtaining the nonuniform temperature distribution from the measured intensity reflection spectrum. The extracted temperature distribution was characterised by the three parameters of amplitude, width and the centre position. The precision of the inferred centre position of the localised heat source was investigated as a function of grating strength. It was demonstrated that the centre position root mean square (rms) error generally improves with grating strength up to approximately 30% peak reflectance and then decreases with further increase in grating strength for both the high and moderate temperature amplitudes which were investigated. A position rms error below 0.03 mm and a repeatability rms error of 0.005 mm were obtained for the centre position measurements with the CFBG intragrating sensor for grating strengths in the range 20-30%. The position rms error reported is the best to date for CFBG based intragrating sensing systems. The intragrating strain measurements were investigated using three embedded CFBGs sensors to determine nonuniform strain distribution near a stress concentration within a notched aluminium specimen subjected to an axial tensile force. The strain distribution was determined from the analysis of the intensity reflection spectrum through the use of an integration method which did not require an initial strain distribution hypothesis. The best case performance was obtained with the sensor having a reflectance of around 23%. The shape of the recovered strain distributions were in reasonable agreement with that predicted by the finite element method (FEM) modelling however some discrepancies were noted. These are most likely due to incomplete transfer of surface strain within the specimen through 3-layers (host/adhesive/fibre) to the fibre core. A repeatability rms error of less than 4 me [mu epsilon] in the extracted strain profiles was obtained.

Device power dissipation has grown exponentially due to the rapid transistor technology scaling and increased circuit complexity. Motivated by the ultra low power requirements of emerging implantable and wearable biomedical devices, novel power management schemes are presented in this thesis to increase device run-time. The schemes involve several techniques suitable for ultra low power biomedical integrated circuit design. This thesis presents a combination of two novel power reduction schemes to reduce the total device power comprising of dynamic and static power dissipation. One of the schemes used is the supply voltage (Vdd) scaling, also known as Dynamic Voltage Scaling (DVS). DVS is an effective scheme to reduce dynamic power (Pdynamic) dissipation. The DVS architecture primarily consists of a DC-DC power regulator which is customised to handle scaling variability of the Vdd. The implemented DVS can dynamically vary the Vdd from 300 mV to 1.2 V. The second scheme presented in this thesis to reduce static power (Pstatic) dissipation is threshold voltage scaling. The variable threshold keeper technique is used to perform threshold voltage scaling, which comprises of a keeper transistor whose threshold voltage is scaled by a body bias generator. The use of the keeper transistor increases the device noise immunity. This combination of supply and threshold voltage scaling techniques offers a further reduction in the overall device power dissipation and enhances reliability without degrading circuit speed. A power reduction of 23% to 31% is achievable with up to 90% efficiency. The thesis discusses the primary design challenges of ultra low power biomedical devices. System and circuit levels design techniques are described which help meeting the stringent requirements imposed by the biomedical environment. This thesis presents a new DVS architecture and investigates the effect of lowering the supply voltage combined with threshold voltage scaling on dynamic power dissipation using 0.13 μm ST-Microelectronic® 6-metal layer CMOS dual- process technology.

Soft computing is a concept that has come into prominence in recent times and its application to power system analysis is still more recent. This thesis explores the application of soft computing techniques in the area of voltage stability of power systems. Soft computing, as opposed to conventional “hard” computing, is a technique that is tolerant of imprecision, uncertainty, partial truth and approximation. Its methods are based on the working of the human brain and it is commonly known as artificial intelligence. The human brain is capable of arriving at valid conclusions based on incomplete and partial data obtained from prior experience. It is an approximation of this process on a very small scale that is used in soft computing. Some of the important branches of soft computing (SC) are artificial neural networks (ANNs), fuzzy logic (FL), genetic computing (GC) and probabilistic reasoning (PR). The soft computing methods are robust and low cost. It is to be noted that soft computing methods are used in such diverse fields as missile guidance, robotics, industrial plants, pattern recognition, market prediction, patient diagnosis, logistics and of course power system analysis and prediction. However in all these fields its application is comparatively new and research is being carried out continuously in many universities and research institutions worldwide. The research presented in this thesis uses the soft computing method of Artificial Neural Networks (ANN’s) for the prediction of voltage instability in power systems. The research is very timely and current and would be a substantial contribution to the present body of knowledge in soft computing and voltage stability, which by itself is a new field. The methods developed in this research would be faster and more economical than presently available methods enabling their use online.

Wireless communication devices are currently enjoying increasing popularity and widespread use. The constantly growing number of users, however, results in the shortage of the available spectrum. Various techniques have been proposed to increase the spectrum efficiency of wireless systems to solve the problem. Multiple Input Multiple Output (MIMO) is one solution that employs multiple antennas at the transmitter and receiver. The MIMO algorithms are usually highly complex and computationally intensive. This results in increased power consumption and reduced battery lifespan. This thesis investigates the complexity – performance trade-off of two MIMO algorithms. Space Time Block Coding (STBC) is a MIMO-based algorithm, which efficiently exploits spatial and temporal diversity. Recently, it has been specified in a number of 3G standards. However, not much attention has been paid to the implementation issues of this algorithm. One such issue, clipping of the Analog to Digital Converter (ADC) at the receiver, is described in the first part of the thesis (chapter 3). A small amount of clipping in an ADC can improve dynamic range and reduce the power consumption. However, the increased clipping distortion of the signal, can adversely affect the overall performance of the system. It will be shown in this dissertation that STBC are more sensitive to clipping, compared to the uncoded single antenna systems. Two receiver structures are considered: Direct Conversion (DC) structure, where the ADCs impose a square clipping function, and a Log-Polar structure, where ADC induces a circular clipping function. Log-Polar receivers were found to be clipping insensitive for the given target Symbol Error Rate (SER) of 1*10-3. This makes Log-Polar receivers an obvious choice for the system designers. The second part of the thesis (chapter 4) addresses the complexity problem associated with the QR decomposition algorithm, which is frequently used as a faster alternative to channel inversion in a MIMO scheme. Channel tracking can be employed with QR equalization in order to reduce the pilot overhead of a MIMO system in a non-stationary environment. QR decomposition is part of the QR equalization method and has to be performed in every instance that the channel estimate is obtained. The high rate of the QR decomposition, a computationally intensive technique, results in a high computational complexity per symbol. Some novel modifications are proposed to address this problem. Reducing the repetition rate of QR decompositions and tracking R (the upper triangular matrix) directly, while holding unitary matrix Q fixed, can significantly reduce complexity per symbol at the expense of some introduced error. Additional modification of the CORDIC algorithm (a square root- and division-free algorithm used to perform QR decomposition) results in more than 80% of computational complexity savings. Further, Minimum Mean Squared Error (MMSE) detection is applied to Least Mean Squared (LMS) based R tracking and channel tracking algorithms and then compared in complexity and performance to the Recursive Least Squares Decision Feedback Equalizer (RLS-DFE) tracking system in [1]. The R tracking scheme is shown to achieve more accurate channel estimates compared to the channel tracking scenario, but this advantage does not translate into better Bit Error Rate (BER) results due to errors on the first layer of the detector. Both LMS strategies have an inferior BER performance compared to the DFE RLS-based system of [1], and surprisingly the LMS schemes show no significant complexity improvement.

The increasing demand for bandwidth in optical fibre communication systems has prompted a significant research effort into developing efficient fibre based optical amplifiers at operating wavelengths neighbouring the erbium-doped fibre amplifier. Of the possible candidates, thulium-doped fibre amplifiers appear best suited to serve this need with amplification available from 1460 to 1530 nm. However, the current limitation with these devices is that the amplifying transition can only provide sufficient gain when doped into host materials with relatively low phonon energies. The ultimate goal for thulium-doped amplifiers is to incorporate the ion into a host material which can be easily integrated into standard communication systems without compromising gain; this thereby involves, to some degree, a silica based host material. To date, optical amplification in the telecommunication S-band using thulium-doped silica fibres has been inefficient due to the high phonon energy associated with Si-O bonds in the glass. This work undertakes a systematic study on the effects of network modifiers on the S-band amplifying transition in an effort to improve the radiative quantum efficiency in silica based glasses. To this end, the techniques employed in this investigation included modifying the glass network with elements which may act to reduce the local phonon energy surrounding the thulium (Tm3+) ion and co-doping the Tm3+ ions with sensitising ytterbium (Yb3+) ions.