The identification of a structural system deterministically or probabilistically is a topic of considerable interest and importance for its condition assessment and prediction. Many identification approaches, however, require the input which is not always available. Specifically, it may be impossible to know the input or, alternately, the measurement of the input is much more unreliable than the dynamic state measurement. Along these lines, engineers try to extract as much information as possible from the available output data to reduce the need for knowing the input. Three new methodologies are developed here to address this challenge. Initially, the input-parameter-state estimation capabilities of a novel unscented Kalman filter, for real time monitoring applications, is examined on both linear and nonlinear systems. The unknown input is estimated in two stages within each time step. Firstly, the predicted dynamic states and the system's parameters provide an estimation of the input. Secondly, the corrected with measurements (updated) dynamic states and parameters provide a final input estimation for the current time step. Subsequently, the estimation of the dynamic states, the parameters, and the input of systems subjected to wind loading is examined using a sequential Kalman filter. The procedure considers two Kalman filters in order to estimate initially the dynamic states using kinematic constraints, and subsequently the system parameters along with the input, in an online fashion. Finally, the input-parameter-state estimation capabilities of a new residual-based Kalman filter are examined for both complete and limited output information conditions. The filter is based on the residual of the predicted and measured dynamic state output, as well as on the residual of the system model estimation. The considered sensitivity analysis is developed using a real time sensitivity matrix formulated by the filtered dynamic states.

付記する学位プログラム名: 霊長類学・ワイルドライフサイエンス・リーディング大学院 / 京都大学 / 新制・課程博士 / 博士(理学) / 甲第24467号 / 理博第4966号 / 新制||理||1709(附属図書館) / 京都大学大学院理学研究科生物科学専攻 / (主査)准教授 Huffman Michael Alan, 教授 古市 剛史, 教授 今井 啓雄 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

Propulsion systems are exposed to a variety of foreign objects that can significantly damage or impact their performance. These threats can range from severe dangers such as sandstorms and volcanic eruptions, which can induce engine failure in minutes, to condensation and moisture during ground tests that can negatively impact the engine's fuel efficiency. While numerous computational and experimental studies have investigated the effects of particle ingestion on the component level, an accurate in-situ measurement technique is needed for a systematic understanding of the effects and real-time engine health monitoring. Optical measurement techniques are attractive for this application due to their non-intrusive nature. However, conventional optical particle measurement methods assume the particle to be spherical, which introduces large errors for measuring particles with complex and irregular shapes, such as sand, volcanic ash, and ice crystals. The light-particle interaction contains information on the desired parameters, such as particle shape and size. The research presented in this dissertation uses this idea for a novel particle sensor concept. Scattering and extinction of light by particles are chosen as crucial features that can identify the particle as its unique signature. Numerical tools are used to simulate the scattering and extinction for particles the sensor is expected to encounter. Machine learning models are trained using the data to use scattering and extinction as inputs and estimate the particle parameters. Different types and applications of supervised machine learning models were investigated, including a layered approach with numerous models and a generalized approach with a single neural network. The particle sensor is first demonstrated using data found in the literature. This study confirmed the importance of non-spherical particles in the library to guide the machine learning models. Further demonstrations are made at a full engine and wind tunnel scale to measure injected condensation and sand sprays, respectively. The mass flow rates of the ingested material were calculated using the model outputs and validated. / Doctor of Philosophy / Foreign objects ingested into gas turbines can cause serious damage and degrade their performance. Threats can range from sand, dust, and volcanic ash to condensation on ground and high altitude ice crystals. On the component level, experiments and simulations have been performed to establish the performance decrease and risks to continued operations. However, there is a need for a real-time and non-intrusive measurement technique for the ingested mass. While there are established optical methods applicable for this use, these existing methods assume the particle shape to be spherical. The light-particle interaction contains information on the desired parameters, such as particle shape and size. Optical measurements of these interactions, such as scattering and extinction, can serve as "fingerprints" that can be used to estimate particle parameters. A novel particle measurement technique utilizing supervised machine learning models is presented. The models are trained using a library containing numerically calculated scattering and extinction data. Laser scattering and extinction measurements are used as inputs for the models. This new technique is first demonstrated by sizing particles found in a particle scattering database in the literature. The method's versatility and ruggedness are then demonstrated by accurately estimating the volume flow rate of a spray nozzle spraying water into a research engine. Additionally, the mass flow of sand particles is measured using this technique in a high-speed wind tunnel, in a similar environment to an engine inlet.

Wind turbine blades experience tremendous stresses while in operation. Failure of a blade can damage other components or other wind turbines. This research focuses on developing an electromagnetic energy harvester for powering structural health monitoring (SHM) equipment inside a turbine blade. The harvester consists of a magnet inside a tube with coils outside the tube. The changing orientation of the blade causes the magnet to slide along the tube, inducing a voltage in the coils which in turn powers the SHM system. This thesis begins with a brief history of electromagnetic energy harvesting and energy harvesters in rotating environments. Next a model of the harvester is developed encompassing the motion of the magnet, the current in the electrical circuit, and the coupling between the mechanical and electrical domains. The nonlinear coupling factor is derived from Faraday's law of induction and from modeling the magnet as a magnetic dipole moment. Three experiments are performed to validate the model: a free fall test to verify the coupling factor expression, a rotating test to study the model with a load resistor circuit, and a capacitor charging test to examine the model with an energy storage circuit. The validated model is then examined under varying tube lengths and positions, varying coil sizes and positions, and variations in other parameters. Finally a sample harvester is presented that can power an SHM system inside a large scale wind turbine blade spinning up to 20 RPM and can produce up to 14.1 mW at 19 RPM. / Master of Science

The data rich nature of instrumented civil structures has brought attention to alternative applications outside of the traditional realm of structural health monitoring. An interest has been raised in using these vibration measurements for other applications such as human occupancy. An example of this is to use the vibrations measured from footsteps to locate occupants within a building. The localization of indoor footsteps can yield several benefits in areas such as security and threat detection, emergency response and evacuation, and building resource management, to name a few. The work described herein seeks to provide supplementary information to better define the problem of indoor footstep localization, and to investigate the use of several localization techniques in a real-world, operational building environment. The complexities of locating footsteps via indoor vibration measurements are discussed from a mechanics perspective using prior literature, and several techniques developed for localization in plate structures are considered for their applicability to indoor localization. A dispersion compensation tool is experimentally investigated for localization in an instrumented building. A machine learning approach is also explored using a nearest neighbor search. Additionally, a novel instrumentation method is designed based on a multi-point coupling approach that provides directional inference from a single point of measurement. This work contributes to solving the indoor footstep localization problem by consolidating the relevant mechanical knowledge and experimentally investigating several potential solutions. / Ph. D.

This dissertation presents the exploration of two fiber optics techniques involving high power pulse laser delivery. The first research topic is "Embedded Active Fiber Optic Sensing Network for Structural Health Monitoring in Harsh Environments", which uses the fiber delivered pulse laser for acoustic generation. The second research topic is "Fiber Optics Guided Laser Micromachining", which uses the fiber delivered pulse laser for material ablation. The objective of the first research topic is to develop a first-of-a-kind technology for remote fiber optic generation and detection of acoustic waves for structural health monitoring in harsh environments. Three different acoustic generation mechanisms were studied in detail, including laser induced plasma breakdown (LIB), Erbium-doped fiber laser absorption, and metal laser absorption. By comparing the performance of the acoustic generation units built based on these three mechanisms, the metal laser absorption method was selected to build a complete fiber optic structure health monitoring (FO-SHM) system. Based on the simulation results of elastic wave propagation and fiber Bragg grating acoustic pulse detection, an FO-SHM sensing system was designed and built. This system was first tested on an aluminum piece in the room temperature range and successfully demonstrated its capability of multi-parameter monitoring and multi-point sensing. With additional studies, the upgraded FO-SHM element was successfully demonstrated at high temperatures up to 600oC on P-91 high temperature steels. During the studies of high power pulse laser delivery, it was discovered that with proper laser-to-fiber coupling, the output laser from a multimode fiber can directly ablate materials around the fiber tip. Therefore, it is possible to use a fiber-guided laser beam instead of free space laser beams for micromachining, and this solves the aspect ratio limitation rooted in a traditional laser beam micromachining method. In this dissertation, this Guided Laser MicroMachining (GLMM) concept was developed and experimentally demonstrated by applying it to high aspect ratio micro-drilling. It was achieved that an aspect ratio of 40 on aluminum and an aspect ratio of 100 on PET, with a hole diameter less than 200 um. / PHD / This dissertation presents two research topics both related to high power laser and fiber optic. The first topic studies the application of using optical fiber and high power laser for ultrasonic non-destructive evaluation. The general idea is to use fiber optic to remotely generate and monitor ultrasonic waves on a workpiece. Due to the fact that there are no electronic components involved in the sensing part of the system, this system can work at high temperature and is unsusceptible to EMI. The second topic studies the usage of optical fiber in high aspect ratio micromachining. The key concept is to use a fiber tip and the output high power laser as a "drilling tip", which eliminate the aspect ratio limitation rooted in the traditional free-space laser micromachining method. With this concept and a demonstrative micromachining system, we achieved record-breaking aspect ratio on both aluminum and plastic.

The ability of a Doppler LIDAR (Light Detection and Ranging) system to measure the speed of a moving rail vehicle in a non-contacting manner is extended to capture the lateral and vertical irregularities of the track itself and to evaluate the rail track quality. Using two pairs of lenses to capture speed signals from both rails individually, the track speed, curvature, and lateral and vertical geometry variations on each side are determined. LIDAR lenses are installed with a slight forward angle to generate velocity signals that contain two components: 1) the left and right track speeds, and 2) any lateral and/or vertical speed caused by track motion and/or spatial irregularities. The LIDAR system collects and outputs the track information in time domain. Separating each speed component (forward, vertical, and lateral) is possible due to the inherent separation of each phenomenon with respect to its spatial/temporal frequencies and related bandwidths. For the measurements to be beneficial in practice, the LIDAR data must be spatially located along the track. A data-mapping algorithm is then simultaneously developed to spatially match the LIDAR track geometry measurements with reference spatial data, accurately locating the measurements along the track and eliminating the need for a Global Positioning System (GPS). A laboratory-grade LIDAR system with four Doppler channels, developed at the Railway Technologies Laboratory (RTL) of Virginia Tech, is body-mounted and tested onboard a geometry measurement railcar. The test results indicate a close match between the LIDAR measurements and those made with existing sensors onboard the railcar. The field-testing conducted during this study indicates that LIDAR sensors could provide a reliable, non-contact track-monitoring instrument for field use, in various weather and track conditions, potentially in a semi-autonomous or autonomous manner. A length-based track quality index (TQI) is established to quantify the track geometry condition based on the geometry data collected by the LIDAR sensors. A phenomenological rail deterioration model is developed to predict the future degradation of geometry quality over the short track segments. The introduced LIDAR's TQI is considered as the condition-parameter, and an internal variable is assumed to govern the rail geometry degradation through a deterioration rule. The method includes the historical data, current track conditions collected by the LIDAR system, and traffic data to calculate the track deterioration condition and identify the geometry defects. In addition to rail geometry inspection, a LIDAR system can potentially be used to monitor the rail surface structure and integrity. This is possible due to the fact that the Doppler shift imposed on the laser radiation reflected from a moving surface has the Doppler bandwidth broadened in proportion to the height and width of the surface features. Two LIDAR-based rail surface measures are introduced based on LIDAR measurements to identify different rail surface conditions and materials. / Ph. D.

Railroad tracks require consistent and periodic monitoring to ensure safety and reliability. Unmanned Aerial Vehicles (UAVs) have great potential because they are not constrained to the track, allowing trains to continue running while the UAV is inspecting. Also, they can be quickly deployed without human intervention. For these reasons, the first steps towards creating a track-monitoring UAV system have been completed. This thesis focuses on the design of algorithms to be deployed on a UAV for the purpose of monitoring the health of railroad tracks. Before designing the algorithms, the first steps were to design a rough physical structure of the final product. A small multirotor or fixed-wing UAV will be used with a gimbaled camera mounted on the belly. The camera will take images of the tracks while the onboard computer processes the images. The computer will locate the tracks in the image as well as perform defect detection on those tracks. Algorithms were implemented once a rough physical structure of the product was completed. These algorithms detect and follow rails through a video feed and detect defects in the rails. The rail following algorithm is based on a custom-designed masking technique that locates rails in images. A defect detection algorithm was also created. This algorithm detect defects by analyzing gradient data on the rail surface. / Master of Science

Civil infrastructure failure is hard to predict, both in terms of occurrence and impact. This is due to combination of many factors, including highly variable environmental and operational conditions, complex construction and materials, and the sheer size of these structures. Often, the mitigation strategy is visual inspection and regular maintenance, which can be time-consuming and may not address root causes of failure. One potential solution to anticipating infrastructure failure and mitigating its consequences is the use of distributed sensors to monitor the physical state of a structure, an area of research known commonly as structural health monitoring, or SHM. This approach can be applied in a variety of contexts: safety during and after natural disasters, evaluation of building construction quality and life-cycle assessment for performance based design frameworks. In one way or another, SHM methods always require a ``baseline,'' a set of physical features which describes the behavior of a healthy structure. Often, the baseline is defined in terms of modal parameters: natural frequencies, damping ratios, and mode shapes. Although changes in modal parameters are indicative of structural damage, they are also indicative of a slew of non-damage factors, such as signal-to-noise ratio, environmental conditions, and the characteristics of forces exciting the structure. In many cases, the degree of observed modal parameter changes due to non-damage factors can be much greater than that due to damage itself. This is especially true of low-frequency modal parameters. For example, the fundamental frequency of a building is more sensitive to global influences like temperature than local structural changes like a cracked column. It has been proposed that extracting modal parameters at higher frequencies may be the key to improving the damage-sensitivity of SHM methods. However, for now, modal analysis of civil structures has been limited to low frequency ambient excitation and sparse sensor networks, due to practical limitations. Two key components for high-frequency modal analysis have yet to be studied: 1) Sufficient excitation at high frequencies and 2) high-resolution (high sensor density) measurements. The unifying goal of this work is to expand modal analysis in these two areas by applying novel instrumentation and experimental methods to two full-scale buildings, Goodwin Hall and Ernest Cockrell Jr. Hall. This enables realistic, practical insights into the limitations and benefits of the high-frequency SHM approach. Throughout, analyses are supported through the novel integration of uncertainty quantification techniques which so far has been under-utilized in the field. This work is divided into three experimental areas, with approaches centering on the identification of modal parameters. The first area is the application of high spacial resolution sensor networks in combination to ambient vibration testing. The second is the implementation of a robust automation and monitoring strategy for complex dynamic structures. The third is the testing of a novel method for performing experimental modal analysis on buildings emph{in situ}. The combination of results from these experiments emphasizes key challenges in establishing reliable high-frequency, high-resolution modal parameter ``baselines'' for structural health monitoring (SHM) of civil infrastructure. The first study presented in this work involved the identification of modal parameters from a five-story building, Goodwin Hall, using operational modal analysis (OMA) on ambient vibration data. The analysis began with a high spacial density network of 98 accelerometers, later expanding this number to 117. A second extensional study then used this data as reference to implement a novel automation method, enabling the identification of long-term patterns in the building's response behavior. Three dominant sources of ambient excitation were identified for Goodwin Hall: wind, human-induced loading, and consistent low-level vibrations from machinery, etc. It was observed that the amplitude of excitation, regardless of source, had significant effects on the estimated natural frequencies and damping ratios. Namely, increased excitation translated to lower natural frequencies and higher damping. In addition, the sources had different characteristics in terms of excitation direction and bandwidth, which contributed to significantly different results depending on the ambient excitation employed. This has significant implications for ambient-based methods that assume that all ambient vibrations are broadband random noise. The third and final study demonstrated the viability of emph{in situ} seismic testing for controlled excitation of full-scale civil structures, also known as experimental modal analysis (EMA). The study was performed by exciting Ernest Cockrell Jr. Hall in Austin, Texas with both vertical and lateral ground waves from seismic shaker truck, T-Rex. The EMA results were compared to a standard operational modal analysis (OMA) procedure which relies on passive ambient vibrations. The study focused on a frequency bandwidth from 0 to 11 Hz, which was deemed high frequency for such a massive structure. In cases were coherence was good, the confidence comparable or better than OMA, with the added advantage that the EMA tests took only a fraction of the time. The ability to control excitation direction in EMA enabled the identification of new structural information that was not observed OMA. It is proposed that the combination of high spacial resolution instrumentation and emph{in situ} excitation have the potential to achieve reliable high-frequency characterization, which are not only more sensitive to local damage but also, in some cases, less sensitive to variations in the excitation conditions. / Ph. D. / Civil structures, like buildings and bridges, become weaker as they age, increasing their risk of collapse due to sever weather, earthquakes, and heavy traffic. Engineers regularly inspect civil structures to ensure they are in good shape, but it is difficult do a full assessment by eye since many defects can be hidden. Structural Health Monitoring, or SHM for short, is an approach that uses permanent vibration sensors to continuously inspect civil structures. Any activity, like blowing wind or moving traffic on bridge, produces small vibrations which can be analyzed to assess the “health” of the structure. This approach can detect some invisible defects, but there is still debate about whether it can detect them when they are small and early on in the life of a structure. If SHM can’t issue early warnings, then there is little incentive to spend large amounts of money on a sensor system. To capture small defects, a sensor system needs a large number of sensors, hence the term high-resolution in the title. In addition, the structure being tested needs to vibrate rapidly (that is at high-frequencies) in order for the high-resolution information to be useful. So far, there have been no tests of this kind on civil structures, especially buildings. Instead, most sensor systems have contained a relatively low number of sensors tested with low-frequency vibrations. This works fills in this gap by testing two different buildings with SHM sensor systems. The first experiment uses a very high number of sensors to analyze the vibrations of Goodwin Hall on the Virginia Tech campus. The vibrations in this building are produced by wind and people walking inside. The second experiment uses a standard number of sensors, but explores a new method of vibrating buildings. This method uses a truck with a large hydraulic piston to shake the ground near the E. Cockrell Jr. building (University of Austin-Texas), essentially creating a tiny earthquake. The experiments show that both testing techniques provide more useful information than standard ones alone. For the first experiment, using more sensors meant the analysis could better distinguish the structural characteristics of the building. For the second, the artificial “earthquake” enabled the measurement of high-frequency vibrations, something which was not possible by relying on wind or people to vibrate the building. Although these new approaches are not used to inspect for damage, they have laid the foundation for improving the early-warning capabilities of SHM systems. This could mean that buildings and other structures can be repaired sooner, remain in operation longer, and cost the owners less money in the end!

Wind power is a fast-growing source of non-polluting, renewable energy with vast potential. However, current wind technology must be improved before the potential of wind power can be fully realized. One of the key components in improving wind turbines is their blades. Blade failure is very costly because blade failure can damage other blades, the wind turbine itself, and possibly other wind turbines. A successful structural health monitoring (SHM) system incorporated into wind turbines could extend blade life and allow for less conservative designs. Impedance-based SHM is a method which has shown promise on a wide variety of structures. The technique utilizes small piezoceramic (PZT) patches attached to a structure as self-sensing actuators to both excite the structure with high-frequency excitations, and monitor any changes in structural mechanical impedance. By monitoring the electrical impedance of the PZT, assessments can be made about the integrity of the mechanical structure. Recent advances in hardware systems with onboard computing, including actuation and sensing, computational algorithms, and wireless telemetry, have improved the accessibility of the impedance method for in-field measurements. The feasibility of implementing impedance-based SHM on wind turbine blades is investigated in this work. Experimentation was performed to determine the capability of the method to detect damage on blades. First, tests were run to detect both indirect and actual forms of damage on a section of an actual wind turbine blade provided by Sandia National Laboratories. Additional tests were run on the same blade section using a high-frequency response function method of SHM for comparison. Finally, based on the results of the initial testing, the impedance method was utilized in an attempt to detect damage during a fatigue test of an experimental wind turbine blade at the National Renewable Energy Laboratory's National Wind Technology Center. / Master of Science