No description available.

The speed control of a Brushless DC Motor Drive in the presence of load disturbance is investigated. Firstly some practical results are presented where a simple proportional-integral speed controller is used in the presence of a large step input speed demand as well as load disturbance. The wind-up problem caused by the saturation of the controller is discussed. In order to improve the performance of the proportional-integral speed controller in the presence of load variation, a load estimator is used with torque feedforward control. The results presented show the speed holding capability in the presence of load variation is significantly improved. A genetic algorithm is used on line to optimise the controller for different conditions such as large and small step input speed demand and load disturbance. The results presented show that a genetic algorithm is capable of finding the tuning of the controller for optimal performance. Single-input single-output and two-input two-output fuzzy speed controllers are also used and the results compared to a proportional-integral controller. Results are presented showing that a single-input single-output fuzzy controller works as a proportional controller with variable gain whereas the two-input two-output fuzzy controller is capable of driving the motor at variable speed and load torque with excellent performance. The robustness of the fuzzy controllers is compared to the proportional-integral controller and the results presented show that the fuzzy one is more robust then the proportional-integral. A genetic algorithm is also used on line for the optimisation of the two-input twooutput fuzzy speed controller and the results show that despite the large number of parameters to be optimised, the tuning for optimal performance is also possible.

Small Unmanned Aerial Vehicles (UAVs) are currently used in several surveillance applications to monitor a set of targets and collect relevant data. One of the main constraints that characterize a small UAV is the maximum amount of fuel the vehicle can carry. In the thesis, we consider a single UAV routing problem where there are multiple depots and the vehicle is allowed to refuel at any depot. The objective of the problem is to find a path for the UAV such that each target is visited at least once by the vehicle, the fuel constraint is never violated along the path for the UAV, and the total length of the path is a minimum. Mixed integer, linear programming formulations are proposed to solve the problem optimally. As solving these formulations to optimality may take a large amount of time, fast and efficient construction and improvement heuristics are developed to find good sub-optimal solutions to the problem. Simulation results are also presented to corroborate the performance of all the algorithms. In addition to the above contributions, this thesis develops an approximation algorithm for a multiple UAV routing problem with fuel constraints.

It is important for engineers to understand the capabilities and limitations of the technologies they consider for use in their systems. Several researchers have investigated approaches for modeling the capabilities of a technology with the aim of supporting the design process. In these works, the information about the physical form is typically abstracted away. However, the efficient generation of an accurate model of technical capabilities remains a challenge. Pareto frontier based methods are often used but yield results that are of limited use for subsequent decision making and analysis. Models based on parameterized Pareto frontiers—termed Technology Characterization Models (TCMs)—are much more reusable and composable. However, there exists no efficient technique for modeling the parameterized Pareto frontier. The contribution of this thesis is a new algorithm for modeling the parameterized Pareto frontier to be used as a model of the characteristics of a technology. The novelty of the algorithm lies in a new concept termed predicted dominance. The proposed algorithm uses fundamental concepts from multi-objective optimization and machine learning to generate a model of the technology frontier.

We address the problem of multi-view multi-target tracking using multiple stationary cameras in the application of hockey tracking and test the approach with data from two cameras. The system is based on the previous work by Okuma et al. [50]. We replace AdaBoost detection with blob detection in both image coordinate systems after background subtraction. The sets of blob-detection results are then mapped to the rink coordinate system using a homography transformation. These observations are further merged into the final detection result which will be incorporated into the particle filter. In addition, we extend the particle filter to use multiple observation models, each corresponding to a view. An observation likelihood and a reference color model are also maintained for each player in each view and are updated only when the player is not occluded in that view. As a result of the expanded coverage range and multiple perspectives in the multi-view tracking, even when the target is occluded in one view, it still can be tracked as long as it is visible from another view. The multi-view tracking data are further processed by trajectory smoothing using the Maximum a posteriori smoother. Finally, automatic camera selection is performed using the Hidden Markov Model to create personalized video programs.

Recent developments in materials and material joining [specifically Aluminum and Pulse Gas Metal Arc Welding (GMAW-P) technology] have increased the scope and extent of their areas of application. However, stern market demand for the improved weld quality necessitates the need for automation of the welding processes. As a result, improvements in the process parameter feedback, sensing and control, are necessary to successfully develop the automated control technology for the welding processes. Hence, several aspects of the GMAW-P process have been investigated in this study in order to improve its control techniques. Welding was conducted on 6XXX aluminium, using 1.2 mm diameter 4047 aluminum electrode and argon shielding gas. An extensive collection of high speed camera pictures were taken over a wide range of pulse parameters and wire feed rates using a xenon shadowgraph setup to improve understanding of the physics of GMAW-P process. Current and voltage signals were recorded concurrently too. This investigation explores the effects of different process parameters namely pulsing parameters (Peak current (IP), Base Current (IB), Peak time (TP), Base Time (TB)) and wire feed rate on metal transfer phenomena in GMAW-P. Number of drops per pulse, arc length and droplet diameter were measured for aluminium electrodes by high speed videography. The pulsing parameters and wire feed rate were varied to investigate their effect on the metal transfer behaviour. Analysis showed that transition between the different metal transfer modes is strongly influenced by the electrode extension. Lower electrode extension reduced the number of droplets detached per pulse, while at higher electrode extension, spray mode is observed due to increased influence of the resistance heating. Analysis of the current and voltage signals were correlated with the high speed films. A simple derivative filter was used to detect the sudden changes in voltage difference associated with metal transfer during GMAW-P. The chosen feature for detection is the mean value of the weld current and voltage. A new algorithm for the real time monitoring and classification of different metal transfer modes in GMAW-P has been developed using voltage and current signals. The performance of the algorithm is assessed using experimental data. The results obtained from the algorithm show that it is possible to detect changes in metal transfer modes automatically and on-line. Arc stability in the GMAW-P has a close relationship with the regularity of metal transfer, which depends on several physical quantities (like voltage, current, materials, etc.) related to the growth and transfer of the metal droplet. Arc state in GMAW-P can be assessed quantitatively in terms of number of drops per pulse, droplet diameter and arc length. In order to assess the arc state in GMAW-P quantitatively, statistical and neural network models for number of drops/pulse, droplet diameter and arc length were developed using different waveform factors extracted from the current waveform of GMAW-P. To validate the models, estimated results were compared to the actual values of the number of drops per pulse, droplet diameter and arc length, observed during several welding conditions. Determination of stable one drop per pulse (ODPP) parametric zone containing all the combinations of peak current (IP), base current (IB), peak time (TP), and base time (TB) that results in stable operation of GMAW-P, is one of the biggest challenges in GMAW-P. A new parametric model to identify the stable ODPP condition in aluminium which also considers the influence of the background conditions and wire feed has been proposed. Finally, a synergic control algorithm for GMAW-P process has been proposed. Synergic algorithm proposed in this work uses the sensing and prediction techniques to analyse state of the arc and correct the pulsing parameters for achieving the stable ODPP. First arc state is estimated using the signal processing techniques and statistical methods to detect the occurrence of short circuit, unstable ODPP or multiple drops per pulse (MDPP) in GMAW-P system. If the arc state is not stable ODPP, then parametric model and genetic algorithm (GA) is used to assess the deviation of the existing pulsing parameters from the stable operation of GMAW-P process and automatically adjust pulsing parameters to achieve stable ODPP.

Microarray data clustering represents a basic exploratory tool to find groups of genes exhibiting similar expression patterns or to detect relevant classes of molecular subtypes. Among a wide range of clustering approaches proposed and applied in the gene expression community to analyze microarray data, mixture model-based clustering has received much attention to its sound statistical framework and its flexibility in data modeling. However, clustering algorithms following the model-based framework suffer from two serious drawbacks. The first drawback is that the performance of these algorithms critically depends on the starting values for their iterative clustering procedures. Additionally, they are not capable of working directly with very high dimensional data sets in the sample clustering problem where the dimension of the data is up to hundreds or thousands. The thesis focuses on the two challenges and includes the following contributions: First, the thesis introduces the statistical model of our proposed normalized Expectation Maximization (EM) algorithm followed by its clustering performance analysis on a number of real microarray data sets. The normalized EM is stable even with random initializations for its EM iterative procedure. The stability of the normalized EM is demonstrated through its performance comparison with other related clustering algorithms. Furthermore, the normalized EM is the first mixture model-based clustering approach to be capable of working directly with very high dimensional microarray data sets in the sample clustering problem, where the number of genes is much larger than the number of samples. This advantage of the normalized EM is illustrated through the comparison with the unnormalized EM (The conventional EM algorithm for Gaussian mixture model-based clustering). Besides, for experimental microarray data sets with the availability of class labels of data points, an interesting property of the convergence speed of the normalized EM with respect to the radius of the hypersphere in its corresponding statistical model is uncovered. Second, to support the performance comparison of different clusterings a new internal index is derived using fundamental concepts from information theory. This index allows the comparison of clustering approaches in which the closeness between data points is evaluated by their cosine similarity. The method for deriving this internal index can be utilized to design other new indexes for comparing clustering approaches which employ a common similarity measure.

A novel decomposition method that is referred to as Contribution-based Decomposition is presented in this thesis. The influence of variables on the values of objective functions and/ or constraints is interpreted as their contributions. Based on contributions of variables, a design problem is decomposed into a number of sub-problems so that variables have similar relative contributions within each sub-problem. The similarity in contributions among variables will lead to an even pressure on the variables when they are driven to better solutions during an optimisation process and, as a result, better solutions can be obtained. Due to nonlinearity of objectives and/ or constrains, variables??? contributions may vary significantly during the solution process. To cope with such variations, a Dynamic Contribution-based Decomposition (DCD) is proposed. By employing DCD, decomposition of system problems is carried out not only at the beginning, but also during the optimisation process, and as a result, the decomposition will always be consistent with the contributions of the current solutions. Further more, a random decomposition is also developed and presented to work in conjunction with the Dynamic Contribution-based Decomposition to introduce re-decompositions when it is required, aiming to increase the global exploring ability. To solve multidisciplinary engineering optimisation problems more efficiently, new solvers are also developed. These include a mixed discrete variable Pattern Search (MDVPS) algorithm and a mixed discrete variable Genetic Algorithm (MDVGA). Inside the MDVGA, new techniques including a flexible floating-point encoding method, a non-dominance ranking strategy and heuristic crossover and mutation operators are also developed to avoid premature convergence and enhance the GA???s search ability. Both MDVPS and MDVGA are able to handle optimisation problems having mixed discrete variables. The former algorithm is more capable of local searching and the latter has better global search ability. A hybrid solver is proposed, which incorporates the MDVPS and the MDVGA and takes advantage of both their strengths. Lastly, a Dynamic Sub-space Optimisation (DSO) method is developed by employing the proposed Dynamic Contribution-based Decomposition methods and the hybrid solver. By employing DSO, decomposed sub-problems can be solved without explicit coordination. To demonstrate the capability of the proposed methods and algorithms, a range of test problems have been exercised and the results are documented. Collectively the results show significant improvements over other published popular approaches.

Thesis (Ph.D)--Aerospace Engineering, Georgia Institute of Technology, 2010. / Committee Chair: Mavris, Dimitri; Committee Member: Beeson, Don; Committee Member: Duncan, Scott; Committee Member: German, Brian; Committee Member: Kumar, Viren. Part of the SMARTech Electronic Thesis and Dissertation Collection.

This article investigates the detection and identification problems for changing of regimes about non-linear time series process. We apply the concept of genetic algorithm and AIC criterion to test the changing of regimes. This way is different from traditional detection methods. According to our statistical decision procedure, the mean of moving average and the genetic detection for the underlying time series will be considered to decide change points. Finally, an empirical application about the detection and identification of change points for the Taiwan Business Cycle is illustrated.