DATA DRIVEN INTELLIGENT AGENT NETWORKS FOR ADAPTIVE MONITORING AND CONTROL

To analyze the characteristics and predict the dynamic behaviors of complex systems over time, comprehensive research to enable the development of systems that can intelligently adapt to the evolving conditions and infer new knowledge with algorithms that are not predesigned is crucially needed. This dissertation research studies the integration of the techniques and methodologies resulted from the fields of pattern recognition, intelligent agents, artificial immune systems, and distributed computing platforms, to create technologies that can more accurately describe and control the dynamics of real-world complex systems. The need for such technologies is emerging in manufacturing, transportation, hazard mitigation, weather and climate prediction, homeland security, and emergency response. Motivated by the ability of mobile agents to dynamically incorporate additional computational and control algorithms into executing applications, mobile agent technology is employed in this research for the adaptive sensing and monitoring in a wireless sensor network. Mobile agents are software components that can travel from one computing platform to another in a network and carry programs and data states that are needed for performing the assigned tasks. To support the generation, migration, communication, and management of mobile monitoring agents, an embeddable mobile agent system (Mobile-C) is integrated with sensor nodes. Mobile monitoring agents visit distributed sensor nodes, read real-time sensor data, and perform anomaly detection using the equipped pattern recognition algorithms. The optimal control of agents is achieved by mimicking the adaptive immune response and the application of multi-objective optimization algorithms. The mobile agent approach provides potential to reduce the communication load and energy consumption in monitoring networks. The major research work of this dissertation project includes: (1) studying effective feature extraction methods for time series measurement data; (2) investigating the impact of the feature extraction methods and dissimilarity measures on the performance of pattern recognition; (3) researching the effects of environmental factors on the performance of pattern recognition; (4) integrating an embeddable mobile agent system with wireless sensor nodes; (5) optimizing agent generation and distribution using artificial immune system concept and multi-objective algorithms; (6) applying mobile agent technology and pattern recognition algorithms for adaptive structural health monitoring and driving cycle pattern recognition; (7) developing a web-based monitoring network to enable the visualization and analysis of real-time sensor data remotely. Techniques and algorithms developed in this dissertation project will contribute to research advances in networked distributed systems operating under changing environments.

Screening, optimization and extraction of polyhydroxyalkanoates and peptidoglycan from Bacillus megaterium

Bioplastics are polymers (such as polyesters) produced from bacterial fermentations that are biodegradable and nonhazardous. They are produced by a wide variety of bacteria and are made only when stress conditions allow, such as when nutrient levels are low, more specifically levels of nitrogen and oxygen. These stress conditions cause certain bacteria to build up excess carbon deposits as energy reserves in the form of polyhydroxyalkanoates (PHAs). PHAs can be extracted and formed into actual plastic with the same strength of conventional, synthetic-based plastics without the need to rely on foreign petroleum. The overall goal of this project was to select for a bacteria that could grow on sugars found in the lignocellulosic biomass, and get the bacteria to produce PHAs and peptidoglycan. Once this was accomplished the goal was to extract PHAs and peptidoglycan in order to make a stronger more rigid plastic, by combing them into a co-polymer. The individual goals of this project were to: (1) Select and screen bacteria that are capable of producing PHAs by utilizing the carbon/energy sources found in lignocellulosic biomass; (2) Maximize the utilization of those sugars present in woody biomass in order to produce optimal levels of PHAs. (3) Use room temperature ionic liquids (RTILs) in order to separate the cell membrane and peptidoglycan, allowing for better extraction of PHAs and more intact peptidoglycan. B. megaterium a Gram-positive PHA-producing bacterium was selected for study in this project. It was grown on a variety of different substrates in order to maximize both its growth and production of PHAs. The optimal conditions were found to be 30°C, pH 6.0 and sugar concentration of either 30g/L glucose or xylose. After optimal growth was obtained, both RTILs and enzymatic treatments were used to break the cell wall, in order to extract the PHAs, and peptidoglycan. PHAs and peptidoglycan were successfully extracted from the cell, and will be used in the future to create a new stronger co-polymer. Peptidoglycan recovery yield was 16% of the cells’ dry weight.

Compressed sensing based cyclic feature spectrum sensing for cognitive radios

Spectrum sensing is currently one of the most challenging design problems in cognitive radio. A robust spectrum sensing technique is important in allowing implementation of a practical dynamic spectrum access in noisy and interference uncertain environments. In addition, it is desired to minimize the sensing time, while meeting the stringent cognitive radio application requirements. To cope with this challenge, cyclic spectrum sensing techniques have been proposed. However, such techniques require very high sampling rates in the wideband regime and thus are costly in hardware implementation and power consumption. In this thesis the concept of compressed sensing is applied to circumvent this problem by utilizing the sparsity of the two-dimensional cyclic spectrum. Compressive sampling is used to reduce the sampling rate and a recovery method is developed for re- constructing the sparse cyclic spectrum from the compressed samples. The reconstruction solution used, exploits the sparsity structure in the two-dimensional cyclic spectrum do-main which is different from conventional compressed sensing techniques for vector-form sparse signals. The entire wideband cyclic spectrum is reconstructed from sub-Nyquist-rate samples for simultaneous detection of multiple signal sources. After the cyclic spectrum recovery two methods are proposed to make spectral occupancy decisions from the recovered cyclic spectrum: a band-by-band multi-cycle detector which works for all modulation schemes, and a fast and simple thresholding method that works for Binary Phase Shift Keying (BPSK) signals only. In addition a method for recovering the power spectrum of stationary signals is developed as a special case. Simulation results demonstrate that the proposed spectrum sensing algorithms can significantly reduce sampling rate without sacrifcing performance. The robustness of the algorithms to the noise uncertainty of the wireless channel is also shown.

Graphite exfoliation by supercritical carbon dioxide extraction

Supercritical carbon dioxide is used to exfoliate graphite, producing a small, several-layer graphitic flake. The supercritical conditions of 2000, 2500, and 3000 psi and temperatures of 40°, 50°, and 60°C, have been used to study the effect of critical density on the sizes and zeta potentials of the treated flakes. Photon Correlation Spectroscopy (PCS), Brunauer-Emmett-Teller (BET) surface area measurement, field emission scanning electron microscopy (FE-SEM), and atomic force microscopy (AFM) are used to observe the features of the flakes. N-methyl-2-pyrrolidinone (NMP), dimethylformamide (DMF), and isopropanol are used as co-solvents to enhance the supercritical carbon dioxide treatment. As a result, the PCS results show that the flakes obtained from high critical density treatment (low temperature and high pressure) are more stable due to more negative charges of zeta potential, but have smaller sizes than those from low critical density (high temperature and low pressure). However, when an additional 1-hour sonication is applied, the size of the flakes from low critical density treatment becomes smaller than those from high critical density treatment. This is probably due to more CO2 molecules stacked between the layers of the graphitic flakes. The zeta potentials of the sonicated samples were slightly more negative than nonsonicated samples. NMP and DMF co-solvents maintain stability and prevented reaggregation of the flakes better than isopropanol. The flakes tend to be larger and more stable as the treatment time increases since larger flat area of graphite is exfoliated. In these experiments, the temperature has more impact on the flakes than pressure. The BET surface area resultsshow that CO2 penetrates the graphite layers more than N2. Moreover, the negative surface area of the treated graphite indicates that the CO2 molecules may be adsorbed between the graphite layers during supercritical treatment. The FE-SEM and AFM images show that the flakes have various shapes and sizes. The effects of surfactants can be observed on the FE-SEM images of the samples in one percent by weight solution of SDBS in water since the sodium dodecylbenzene sulfonate (SDBS) residue covers all of the remaining flakes. The AFM images show that the vertical thickness of the graphitic flakes can ranges from several nanometers (less than ten layers thick), to more than a hundred nanometers. In conclusion, supercritical carbon dioxide treatment is a promising step compared to mechanical and chemical exfoliation techniques in the large scale production of thin graphitic flake, breaking down the graphite flakes into flakes only a fewer graphene layers thick.

Skin Lesion Extraction And Its Application

In this thesis, I study skin lesion detection and its applications to skin cancer diagnosis. A skin lesion detection algorithm is proposed. The proposed algorithm is based color information and threshold. For the proposed algorithm, several color spaces are studied and the detection results are compared. Experimental results show that YUV color space can achieve the best performance. Besides, I develop a distance histogram based threshold selection method and the method is proven to be better than other adaptive threshold selection methods for color detection. Besides the detection algorithms, I also investigate GPU speed-up techniques for skin lesion extraction and the results show that GPU has potential applications in speeding-up skin lesion extraction. Based on the skin lesion detection algorithms proposed, I developed a mobile-based skin cancer diagnosis application. In this application, the user with an iPhone installed with the proposed application can use the iPhone as a diagnosis tool to find the potential skin lesions in a persons' skin and compare the skin lesions detected by the iPhone with the skin lesions stored in a database in a remote server.

EXTRACTION OF PICTORIAL ENERGY INFORMATION FROM CAMPUS UNMETERED BUILDINGS USING IMAGE PROCESSING TECHNIQUES

In recent years, advanced metering infrastructure (AMI) has been the main research focus due to the traditional power grid has been restricted to meet development requirements. There has been an ongoing effort to increase the number of AMI devices that provide real-time data readings to improve system observability. Deployed AMI across distribution secondary networks provides load and consumption information for individual households which can improve grid management. Significant upgrade costs associated with retrofitting existing meters with network-capable sensing can be made more economical by using image processing methods to extract usage information from images of the existing meters. This thesis presents a new solution that uses online data exchange of power consumption information to a cloud server without modifying the existing electromechanical analog meters. In this framework, application of a systematic approach to extract energy data from images replaces the manual reading process. One case study illustrates the digital imaging approach is compared to the averages determined by visual readings over a one-month period.

IONIC LIQUID EXTRACTION UNVEILS PREVIOUSLY OCCLUDED HUMICBOUND IRON IN PEAT POREWATER

Extracellular iron reduction has been suggested as a candidate metabolic pathway that may explain a large proportion of carbon respiration in temperate peatlands. However, the o-phenanthroline colorimetric method commonly employed to quantitate iron and partition between redox species is known to be unreliable in the presence of humic and fulvic acids, both of which represent a considerable proportion of peatland dissolved organic matter. We propose ionic liquid extraction as a more accurate iron quantitation and redox speciation method in humic-rich peat porewater. We evaluated both o-phenanthroline and ionic liquid extraction in four distinct peatland systems spanning a gradient of physico-chemical conditions to compare total iron recovery and Fe2+:Fe3+ ratios determined by each method. Ionic liquid extraction was found to provide more accurate iron quantitation and speciation in the presence of dissolved organic matter. A multivariate approach utilizing fluorescence- and UV-Vis spectroscopy was used to identify dissolved organic matter characteristics in peat porewater that lead to poor performance of the o-phenanthroline method. Where these interferences are present, we offer an empirical correction factor for total iron quantitation by o-phenanthroline, as verified by ionic liquid extraction. The written work presented in this thesis is in preparation for submission to Soil Biology and Biochemisrty by T.J. Veverica, E.S. Kane, A.M. Marcarelli, and S.A. Green.