The primary purpose of this study is to evaluate the use of Doppler Lidar sensors for assessing track weakening that would indicate early stages of track instability. Such track weakening could lead to gage widening or track buckling due to rail thermal expansion. A series of tests are performed at the Transportation Technology Center's High Tonnage Loop, where two sections of track are "doctored" to have weaker lateral strength, one on a tangent and another one in a curve. Multiple tests are performed at speeds ranging from 10 – 40 mph, during which the lateral and vertical deflections of the rail are measured under the weight of the passing wheels of a heavily-loaded gondola. The track weakness is created by removing the rail spikes from eight consecutive ties. The measurements from the soft sections are compared with a track section on a tangent that is determined to have nominally sufficient ("good") stiffness. The measurement system consists of four Doppler Lidar units, two oriented toward the rail gage face to measure lateral rail movement, and two directed to the top of the rail to measure vertical rail movement. The combination of the vertical and lateral measurements is used as an indicator of a lack of rail stability if larger-than-normal movement of the rail is detected in either direction. The data collected is analyzed through various methods designed to differentiate sections of track including Gaussian Mixture Model sorting algorithms, inspection via Short Time Fourier Transforms, Discrete Wavelet Transforms, and manual inspection. None of the methods can be done automatically; they each require a different amount of setup and pre-processing before the raw data can be made suitable for the analysis offered by each. The pre-processing can account for dropped data and can be used to identify some false positives such as switches or lubricators. The test results indicate that the system provides a distinctly different measurement in the sections that are doctored to have less track stability than the section with nominally sufficient stiffness. The detection of the loose track in the tangent sections, however, proves to be less reliable. For those, a mostly ad hoc approach is necessary to match the measured data with video images to pinpoint the exact location of the measurements. It is not clear to what extent such approaches would be feasible in practice. Further evaluations of the test data may be used to shed more light on practical analysis methods—possibly wavelets—that are more automated and less ad hoc. They can also provide alternative system setups or designs of experiments for future tests at TTC or on revenue service tracks. / Master of Science / The purpose of this study is to evaluate the effectiveness of a set of Doppler Lidar sensors for their ability to determine the locations of weaker sections of railroad track. These weaker sections could cause damage to the track or passing trains by deforming or buckling under load. A set of tests are performed at the Transportation Technology Center's High Tonnage Loop to evaluate these capabilities. The track had two sections, one of curved track the other of straight track, where the rail was purposefully weakened by removing retaining spikes from the railroad ties. The weakened sections were created by removing the vertical retaining spikes in eight consecutive ties. The tests were conducted at speeds of between 10 to 40 mph, and the sensors measured both the vertical and lateral movement of both rails. The results of these measurements were compared with the unaffected rail. The collected data is analyzed using various data processing techniques. These techniques included using a sorting algorithm to find sections of track with different characteristics as well as inspecting the time and frequency content of the data. None of these methods are automated, and each requires specific setup and adjustment to be effective. The data also needs to be prepared by correcting for any missing or incorrect data points. The tests indicate that the system is able to differentiate between the purposefully weakened track and the rest of the track, however the clearest results of this were for the weakened track in the curve. The straight track results were able to be found with the addition of aligning the video, Lidar, and GPS data sets. It is not clear whether the system could be improved to detect this type of weakness in straight track in practice. Additional testing and evaluation could serve to expand the range of data analysis methods used in differentiating the track conditions and could serve to automate the process. Additionally, alternative test setups could provide further information as to the capabilities of the sensors to detect different types of weakened track.

Avec l'évolution des dispositifs d'acquisition 3D, les nuages de points sont maintenant devenus une représentation essentielle des scènes numérisées. Les systèmes récents sont capables de capturer plusieurs centaines de millions de points en une seule acquisition. Comme plusieurs acquisitions sont nécessaires pour capturer la géométrie de scènes de grande taille, un site historique par exemple, nous obtenons des nuages de points massifs, i.e., composés de plusieurs milliards de points. Dans cette thèse, nous nous intéressons à la structuration et à la manipulation de nuages de points issus d'acquisitions générées à partir de LiDARs terrestres. A partir de la structure de chaque acquisition, des graphes, représentant chacun la connectivité locale de la surface numérisée, sont construits. Les graphes créés sont ensuite liés entre eux afin d'obtenir une représentation globale de la surface capturée. Nous montrons que cette structure est particulièrement adaptée à la manipulation de la surface sous-jacente aux nuages de points massifs, même sur des ordinateurs ayant une mémoire limitée. Notamment, nous montrons que cette structure permet de traiter deux problèmes spécifiques à ce type de données. Un premier lié au ré-échantillonnage de nuages de points, en générant des distributions de bonne qualité en termes de bruit bleu grâce à un algorithme d'échantillonnage en disques de Poisson. Un autre lié à la construction de diagrammes de Voronoï centroïdaux, permettant l'amélioration de la qualité des distributions générées, ainsi que la reconstruction de maillages triangulaires. / With the evolution of 3D acquisition devices, point clouds have now become an essential representation of digitized scenes. Recent systems are able to capture several hundreds of millions of points in a single acquisition. As multiple acquisitions are necessary to capture the geometry of large-scale scenes, a historical site for example, we obtain massive point clouds, i.e., composed of billions of points. In this thesis, we are interested in the structuration and manipulation of point clouds from acquisitions generated by terrestrial LiDARs. From the structure of each acquisition, graphs, each representing the local connectivity of the digitized surface, are constructed. Created graphs are then linked together to obtain a global representation of the captured surface. We show that this structure is particularly adapted to the manipulation of the underlying surface of massive point clouds, even on computers with limited memory. Especially, we show that this structure allow to deal with two problems specific to that kind of data. A first one linked to the resampling of point clouds, by generating distributions of good quality in terms of blue noise thanks to a Poisson disk sampling algorithm. Another one connected to the construction of centroidal Voronoi tessellations, allowing to enhance the quality of generated distributions and to reconstruct triangular meshes.

abstract: Environmental remote sensing has seen rapid growth in the recent years and Doppler wind lidars have gained popularity primarily due to their non-intrusive, high spatial and temporal measurement capabilities. While lidar applications early on, relied on the radial velocity measurements alone, most of the practical applications in wind farm control and short term wind prediction require knowledge of the vector wind field. Over the past couple of years, multiple works on lidars have explored three primary methods of retrieving wind vectors viz., using homogeneous windfield assumption, computationally extensive variational methods and the use of multiple Doppler lidars. Building on prior research, the current three-part study, first demonstrates the capabilities of single and dual Doppler lidar retrievals in capturing downslope windstorm-type flows occurring at Arizona’s Barringer Meteor Crater as a part of the METCRAX II field experiment. Next, to address the need for a reliable and computationally efficient vector retrieval for adaptive wind farm control applications, a novel 2D vector retrieval based on a variational formulation was developed and applied on lidar scans from an offshore wind farm and validated with data from a cup and vane anemometer installed on a nearby research platform. Finally, a novel data visualization technique using Mixed Reality (MR)/ Augmented Reality (AR) technology is presented to visualize data from atmospheric sensors. MR is an environment in which the user's visual perception of the real world is enhanced with live, interactive, computer generated sensory input (in this case, data from atmospheric sensors like Doppler lidars). A methodology using modern game development platforms is presented and demonstrated with lidar retrieved wind fields. In the current study, the possibility of using this technology to visualize data from atmospheric sensors in mixed reality is explored and demonstrated with lidar retrieved wind fields as well as a few earth science datasets for education and outreach activities. / Dissertation/Thesis / Doctoral Dissertation Mechanical Engineering 2017