CIVINS / Field studies have demonstrated that it is possible to evaluate the state of many shipboard systems by analyzing the power drawn by electromechanical actuators. One device that can perform such an analysis is the non-intrusive load monitor (NILM). This thesis investigates the use of the NILM as a supervisory control system in the engineering plant of gas-turbine-powered vessel. Field tests demonstrate that the NILM can potentially reduce overall sensor count if used in a supervisory control system. To demonstrate the NILM's capabilities in supervisory control systems, experiments are being conducted at the U.S. Navy's Land-Based Engineering Site (LBES) in Philadelphia, Pennsylvania. Following a brief description of the LBES facility and the NILM itself, this thesis presents testing procedures and methodology with results obtained during the extensive field studies. This thesis also describes the on-going efforts to further demonstrate and develop the NILM's capabilities in supervisory control systems. / CIVINS / US Naval Academy (USNA) author.

This thesis proposes a novel graphical model for inference called the Affinity Network,which displays the closeness between pairs of variables and is an alternative to Bayesian Networks and Dependency Networks. The Affinity Network shares some similarities with Bayesian Networks and Dependency Networks but avoids their heuristic and stochastic graph construction algorithms by using a message passing scheme. A comparison with the above two instances of graphical models is given for sparse discrete and continuous medical data and data taken from the UCI machine learning repository. The experimental study reveals that the Affinity Network graphs tend to be more accurate on the basis of an exhaustive search with the small datasets. Moreover, the graph construction algorithm is faster than the other two methods with huge datasets. The Affinity Network is also applied to data produced by a synchronised system. A detailed analysis and numerical investigation into this dynamical system is provided and it is shown that the Affinity Network can be used to characterise its emergent behaviour even in the presence of noise.

The development of AC drives has gained momentum with the advancement of power electronic devices, control theory and technology of microprocessors. Based on high speed digital signal processing theory and power transistor hardware, two axis machine model and adaptive control techniques, a vector controlled fully digital induction motor drive is a high performance, low cost drive which is becoming increasingly popular in many industrial applications in all power ranges and will succeed the dc drive in the near future. The main goal of the thesis is to investigate the various forms of implementations of vector controlled digital signal processor based high performance induction motor drives. For this purpose, a TMS320C30 digital signal processor board with transducer and interfacing circuits has been designed, manufactured and built. Furthermore a conventional bipolar transistor inverter has been modified and interfaced to the DSP board. The entire hardware has been tested and successfully implemented. The control software is versatile and provides a platform for implementing various control configurations. The first part of the thesis reviews the development of various vector controlled drives; different mathematical models are presented and discussed together with digital simulation results of different vector controlled schemes. Intelligent control algorithms based on neural networks, fuzzy control, self-tuning control are also considered, some of them are tested by using real-time simulation or on a real drive. Different PWM strategies are investigated and compared, and a versatile and reliable real-time PWM algorithm is generated and integrated into the drive control software. Various parameter identification and auto-commissioning techniques are included in the control software. Details of experimentally obtained results for various vector drives are presented.

This thesis is concerned with the analysis of rotary induction motors. Chapter two develops a general field analysis based on a multilayer model of the machine. The analysis is shown to be capable of taking the effects of the discrete nature of skewed rotor bars directly into account. The method is further shown to be able to accommodate asymmetrical stator windings and terminal voltages with ease, and to determine the magnetic fields within the model incidentally. In chapter three, the analysis is verified theoretically by imposing certain classical assumptions and comparing with a conventional equivalent-circuit approach. The effects on classical machines text formulae of relaxing the small-airgap assumption are then examined. Chapter four presents a discussion on some of the practical aspects concerning computer implementation of the field analysis. These include selection of field harmonics and the effects of saturation and rotor skin-effeet. Chapter five applies the field analysis to an experimental machine operating in the first instance as a single-phase motor, and in the second, as an unbalanced two-phase motor. The predicted performance characteristics of this machine are compared with those determined experimentally, to indicate the validity of the method.

No description available.

The work, of which this thesis is a record, is concerned with the development of condition monitoring and protection facilities for electrical drives and generators. There are two primary objectives, the first being the development of a consistent theory governing the production of axial - flux harmonics in electrical machines. This development proceeds with the identification of the harmonic spectra expected in machines which are running normally, and is then extended to predict the changes to be expected for a variety of fault conditions. Faults such as: stator winding interturn short circuits, negative phase sequence due to unbalanced supply, eccentric running due to bearing wear, fractured rotorbars in induction machines, and rotor winding short circuits are investigated and the predicted changes verified by test and experiment on machines varying in rating from 2 kW to 100 HW. The second, and overlapping, primary objective is the provision of an integrated, microprocessor-based condition monitoring system that will also afford protection for the machines under surveillance against the fault conditions mentioned above. Such a system based on the FFT analysis of the axial flux signal has been constructed to prototype Level, and has met said objective. The system comprises two 8 bit and one 16 bit processors and is designed to react extremely rapidly to the occurrence of a fault condition. Throughout the design phase, the aims of simplicity of operation and flexibility have been paramount. The provision of test facilities in the industrial environment afforded by the CF.GB proved invaluable in these respects. It is thought that the methods of monitoring and protecting machines, mentioned above, has significant attractions because it allows multi-functional protection from a single transducer to be achieved, and the form of the transducer is such that line voltages and currents need not be sensed.

Brushless motors are increasingly popular because of their high power density, torque to inertia ratio and high efficiency. However an operational characteristic is the occurrence of torque ripple at low speeds. For demanding direct drive applications like machine tools, robot arms or aerospace applications it is necessary to reduce the level of torque ripple. This thesis presents an in depth investigation into the production and nature of torque ripple in brushless machines. Different torque ripple reduction strategies are evaluated and one reduction strategy using Park's transform as a tool is identified as the promising strategy. The unified machine theory is checked to clarify the theory behind Park's transform; in particular assumptions made and general validity of the theory. This torque ripple reduction strategy based on Park's transform is extended to include the effect of armature reaction. A novel adaptive torque ripple reduction algorithm is designed. The ineffectiveness of the conventional approach is demonstrated. Further a novel torque ripple reduction strategy using direct measurements of the torque ripple is suggested, reducing implementation time and allowing higher accuracies for torque ripple reduction. Extensive measurements from the experimental system show the validity of the novel torque ripple reduction strategies. The experimental results allow derivation of a formula for all load 111 situations. This formula makes it possible to further increase the reduction accuracy and enables improved real time implementation of the torque ripple reduction algorithm. The work presented here makes a substantial contribution towards understanding the nature of torque ripple in brushless motors and solving the associated problems. The novel reduction strategies form the basis for the development of intelligent dynamometers for motor test beds. Further the torque ripple reduction method presented here can be used to overcome manufacturing imperfections in brushless machines thus removing the cost for precise manufacturing tools. Future designs of controllers can "build" their own correction formula during set-up runs, providing a motor specific torque ripple correction. IV

This thesis is concerned with the development of a prototype electromagnetic flowmeter for dielectric liquids. A initial prototype flowmeter is tested in laboratory conditions with a liquid of low conductivity, the conclusion of these initial tests being that the stability of the flowmeter is below that required for its design applications. This lack of stability is a problem reported on several occasions in similar projects involving electromagnetic flowmeters for dielectric liquids. Therefore the main body of the work is spent investigating the causes of instability and trying to redress the necessary problems in order that a viable working prototype is developed. Problems of thermal instability, electrostatic and electromagnetic shielding are addressed along with the causes and effects of all the major noise and unwanted signals experienced in such a system. A new prototype is developed incorporating stability driven improvements. This new design is rigorously tested and assessed. The stability of the meter is considered with respect to its flow related output. Several parameters of operational stability are given in the examples considered. A final assessment is then made of the flowmeters baseline performance and a account of any remaining instability is given with respect to the issue developed throughout the thesis. The conclusions indicate that the electromagnetic flowmeter for dielectric fluids does produce a linear output based on Faraday's law of induction. They also indicate that the instability experienced previously can be broken into its constituent parts and each contributor accounted for and reduced, working towards the ideal totally stable flowmeter.

No description available.

The problems of modelling the electromagnetic characteristics of vehicles and the experimental validation of such models are considered. The validity of the measurement methods that are applied in model validation exercises is of particular concern. A philosophy for approaching the validation of automotive electromagnetic models of realistic complexity is presented. Mathematical modelling of the key elements of the measurement processes is proposed as the only reliable mechanism for addressing these issues. Areas considered include: basic elements of numerical models; geometrical fidelity requirements for model elements; calibration and use of experimental transducers; the inclusion of cables in electromagnetic models; essential content for vehicle models. A number of practical measurement processes are also investigated using numerical methods, leading to recommendations for improved practices in: calibration of transducers for current measurement at high frequencies; measurement of radiated emissions from vehicles; identification of range requirements for simple methods of determining antenna gain and related characteristics in EMC test facilities. The impact of such measures on the success of model validation studies for automotive applications is demonstrated. It is concluded that experimental results are no less in need of validation than the numerical results that are, more conventionally, judged against them.