L’analyse des paysages, de ses dynamiques et ses processus environnementaux, nécessite d’acquérir régulièrement des données des sites, notamment pour le bilan glaciaire au Spitsberg et en haute montagne. A cause des mauvaises conditions climatiques communes aux latitudes polaires et à cause de leur coût, les images satellites journalières ne sont pas toujours accessibles. De ce fait, les événements rapides comme la fonte de la neige ou l'enneigement ne peuvent pas être étudiés à partir des données de télédétection à cause de leur fréquence trop faible. Nous avons complété les images satellites par un ensemble de de stations photo automatiques et autonomes qui prennent 3 photos par jour. L’acquisition de ces photos génère une grande base de données d’images. Plusieurs traitements doivent être appliqués sur les photos afin d’extraire l’information souhaitée (modifications géométriques, gestion des perturbations atmosphériques, classification, etc). Seule l’informatique est à même de stocker et gérer toutes ces informations. Le cloud computing offre en tant que services des ressources informatiques (puissance de calcul, espace de stockage, applications, etc). Uniquement le stockage de la masse de données géographique pourrait être une raison d’utilisation du cloud computing. Mais en plus de son espace de stockage, le cloud offre une simplicité d’accès, une architecture scalable ainsi qu’une modularité dans les services disponibles. Dans le cadre de l’analyse des photos in situ, le cloud computing donne la possibilité de mettre en place un outil automatique afin de traiter l’ensemble des données malgré la variété des perturbations ainsi que le volume de données. A travers une décomposition du traitement d’images en plusieurs tâches, implémentées en tant que web services, la composition de ces services nous permet d’adapter le traitement aux conditions de chacune des données. / Analyzing landscape, its dynamics and environmental evolutions require regular data from the sites, specifically for glacier mass balanced in Spitsbergen and high mountain area. Due to poor weather conditions including common heavy cloud cover at polar latitudes, and because of its cost, daily satellite imaging is not always accessible. Besides, fast events like flood or blanket of snow is ignored by satellite based studies, since the slowest sampling rate is unable to observe it. We complement satellite imagery with a set of ground based autonomous automated digital cameras which take 3 pictures a day. These pictures form a huge database. Each picture needs many processing to extract the information (geometric modifications, atmospheric disturbances, classification, etc). Only computer science is able to store and manage all this information. Cloud computing, being more accessible in the last few years, offers as services IT resources (computing power, storage, applications, etc.). The storage of the huge geographical data could, in itself, be a reason to use cloud computing. But in addition to its storage space, cloud offers an easy way to access , a scalable architecture and a modularity in the services available. As part of the analysis of in situ images, cloud computing offers the possibility to set up an automated tool to process all the data despite the variety of disturbances and the data volume. Through decomposition of image processing in several tasks, implemented as web services, the composition of these services allows us to adapt the treatment to the conditions of each of the data.

The purpose of this work is to develop and evaluate electroding methods for a reverse osmosis (RO) membrane that results in an in situ sensor able to detect RO membrane protein fouling. Four electroding techniques were explored: i) gold exchange-reduction, ii) encapsulated carbon grease, iii) "direct assembly process" (DAP), and iv) platinized polymer graft. The novel platinized polymer graft method involves chemically modifying the RO membrane surface to facilitate platinization based on the hypothesis that deposition of foulant on the platinized surface will affect platinum/foulant/solution interfacial regions, thus sensor impedance. Platinized polymer graft sensors were shown to be sensitive to protein fouling. Electrodes were characterized by their electrical properties, SEM and XPS. Assembled sensors were evaluated for sensitivity to electrolyte concentration and protein fouling. Micrographs showed coating layers and pre-soak solution influence gold exchange-reduction electrode formation. High surface resistance makes gold exchange-reduction an unsuitable method. Concentration sensitivity experiments showed carbon grease and DAP electroding methods produce unusable sensors. Carbon grease sensors have time-dependent impedance response due to electrolyte diffusion within the micro-porous polysulfone support. DAP electroded sensors proved quite fragile upon hydration; their impedance response is transient and lacks predictable trends with changes in concentration. A parametric study of the platinized polymer graft method shows amount of grafted monomer correlates to grafting time, and deposited platinum is a function of exchange-reduction repetitions and amount of grafted monomer. Platinized polymer graft sensors were fouled in both dead-end and cross-flow RO systems, and their impedance trends, while varying between sensors, indicate protein-fouling sensitivity. / Master of Science

The layer-by-layer nature of additive manufacturing (AM) allows for toolless fabrication of highly complex geometries that cannot be made via traditional processes. AM is unique in its ability to precisely define both the material properties and geometric shape throughout the volume of a part, giving designers unmatched freedom in the creation of new components. However, this freedom of design also creates numerous challenges in the qualification of these parts. As AM processes primitive material in real time to produce each voxel of part volume, manufacturing defects may be dispersed anywhere throughout the part. Many part designs may have complex geometries or material parameters that are challenging for traditional qualification and inspection techniques to inspect for such embedded errors. Even more troubling, this freedom of design also extends to malicious actors, who would then be able to embed intentional targeted defects within the volume of the part. As the AM process is driven almost entirely by computer controlled machines and cyber-domain data, the AM process is uniquely at risk of nearly undetectable cyber-physical attacks, or cyber attacks that can cause physical damage. Additionally, as much of the valuable intellectual property associated with the design and material parameters of parts are stored in digital form, theft of these design files could result in mass replication of lower quality counterfeit parts, putting the supply chain of these AM parts at risk. In order to mitigate these vulnerabilities in the AM process, prior works have focused on in situ monitoring of the manufacturing process in order to ensure the part is constructed as expected. Typically for in situ monitoring, the constructed geometry is compared to the design files associated with the part in question using a monitoring system connected to either the AM machine or the larger network. However, such methods trust the validity of both the design files and monitoring systems used for verification, when either or both may have also been attacked. Therefore, a valid in situ monitoring method needs secure access to a provable set of validation data, while also isolating or air-gapping itself from the network to prevent cyber attacks on the monitoring system itself. Similarly, other works have focused on mitigating the risk of counterfeiting by novel means of part identification tailored for the AM process. Many of these identification methods leverage stochastic or prescribed features, such as surface patterns measured via visible or ultraviolet scanning, or internal porosity features measured via x-ray computed tomography (CT) scanning. However, these surface features are not impacted by alterations or damage to the part in areas away from the specific features being measured, possibly preventing the detection of attacks or damage to other areas of the part in transit. CT scanning can be used to detect damage or alterations to more areas of the part and incorporate this measurement into the identification mechanism, but may be prohibitively expensive while also possibly failing to properly penetrate and measure a sufficiently complex AM part. In this work, efforts to expand the cyber-physical security of the AM process are explored, including (1) a novel method of in situ process validation by means of covertly transmit- ting process quality information to an otherwise air-gapped monitoring system, (2) a novel method of metal AM part identification via a low-cost piezoelectric sensor-actuator able to record a part frequency response that is dependent on the geometry and material properties of the part as a whole, (3) an exploration of part-to-part variation across AM processes, again measured via a piezoelectric sensor-actuator, and (4) a novel means of using the same piezoelectric sensor-actuator for detecting the presence of remaining powder in metal AM parts. / Doctor of Philosophy / Additive manufacturing (AM), or 3D printing, allows for the creation of highly complex parts. AM machines do this by building parts layer-by-layer, processing (e.g., selectively melting metal powder) and placing each segment of a part from the bottom up, allowing it to make internal features which would be impossible with traditional manufacturing processes, such as machining.. However, because these parts can be so complicated, it is difficult to validate that a part is "good", i.e., is free from defects. As the entire volume of the part is built layer- by-layer, any layer anywhere in the part could be defective, with very few techniques being capable of detecting the defect from the outside. Worse, because the AM process is driven by digital design files and other data, cyber attacks have the ability to maliciously change the design of a part before it is made, resulting in physical damage. These cyber-physical attacks can similarly affect existing validation methods, allowing these attacked parts to slip through undetected. Alternatively, part designs can be stolen, allowing the thieves to produce unauthorized and possibly subpar counterfeits. These dangers require new means of validating the AM process and the parts it can produce. In order to detect a cyber-physical attack, previous studies have looked to recording and monitoring the physical actions of the AM process in order to ensure the part is built layer- by-layer as expected. Typically, the part design files are sent to the network-connected monitoring system, which then compares the files to the as-built geometry being recorded. However, in this case, the design files can themselves be attacked, as can the monitoring system recording and comparing the part geometry. In order to detect bad parts without exposing the system to cyber attacks, the monitoring system needs a way to validate the AM part without relying on the part design files directly or being connected to the network. To determine is a part is counterfeit or not, previous studies have tried to create "fingerprints" for parts, allowing a unique part to be identified. However, many of these techniques require changes to the part in question, or rely on features that could be duplicated (i.e. copying the fingerprint) by a skilled attacker. Certain methods using x-ray computed tomography (CT) scanning, while effective at fingerprinting small parts, can be very expensive, and may not work for parts which are too large or complex for x-rays to cleanly pass through. To be successful, a fingerprint needs to be simple to measure, and dependent on the entirety of the part itself, not just a handful of manufactured features. This can be done using the frequency response of the part, or how much the part vibrates over a range of frequencies. This response is dependent on the entire part, including the geometry and the material properties, and can be measured using low-cost equipment, allowing it to be used for a variety of different purposes. In this work, several methods to enhance the cyber-physical security of the AM process are explored. These include (1) a method of validating the AM process by covertly transmitting information to a network disconnected monitoring system, (2) a method of identifying metal AM parts identification using the parts frequency response as a fingerprint, (3) an exploration of part frequency response for fingerprinting across other AM processes, including both metal and polymer parts, and (4) a means of using the frequency response of a part for detecting the presence of residual powder from powder-based AM processes.