Em formação de imagens por ultrassom de objetos com superfícies curvas, em imersão, se as velocidades de propagação no ruído e no objeto forem muito diferentes os efeitos de refração precisam ser compensados. Nesse caso a posição e a forma da superfície precisam ser conhecidas. Neste trabalho a superfície é detectada pelo mesmo array linear que captura os sinais para a formação de imagem.Dois métodos rápidos de detecção de superfície foram propostos, um baseado em técnicas de formação de imagem e outro que utiliza informações de tempo de percurso de ecos extraídas diretamente dos sinais de ultrassom.Os dois métodos foram comparados,e o método baseado em imagem apresentou uma maior tolerância a erros nos sinais, enquanto o método baseado em tempo de percurso mostrou-se mais rápido.Com a superfície detectada, a imagem foi formada por combinação de imagens por abertura sintética, que apresentou uma boa resolução. O uso conjunto dos métodos de detecção de superfície propostos e da formação de imagem resultou em uma taxa de imagens que permite ensaios interativos,com processamento em uma CPU de uso geral. / In ultrasound imaging ofobjects with curved surface, in immersion, the refraction eects must be compensated for if the propagation speed in the uid is very dierent from the speed in the object. In this case the surface position and shape must be known. In this work the surface is detected by the same linear array that captures the signals for the image formation. Two fast methods for surface detection were proposed,one is based on image formation techniques and another utilizes the echotime-of-ight information directly from the ultra\\sound signals.The two methods were compared,and the image-based method was more tolerant of signal errors, while the time-of-ight-based method was faster. After the surface detection, the image was formed by combination of synthetic aperture images,with a good resulting resolution. The utilization of the proposed surface detection methods together with the image formation resulted in an image rate that allows interactive testing, with processing on a general-purpose CPU.

Em formação de imagens por ultrassom de objetos com superfícies curvas, em imersão, se as velocidades de propagação no ruído e no objeto forem muito diferentes os efeitos de refração precisam ser compensados. Nesse caso a posição e a forma da superfície precisam ser conhecidas. Neste trabalho a superfície é detectada pelo mesmo array linear que captura os sinais para a formação de imagem.Dois métodos rápidos de detecção de superfície foram propostos, um baseado em técnicas de formação de imagem e outro que utiliza informações de tempo de percurso de ecos extraídas diretamente dos sinais de ultrassom.Os dois métodos foram comparados,e o método baseado em imagem apresentou uma maior tolerância a erros nos sinais, enquanto o método baseado em tempo de percurso mostrou-se mais rápido.Com a superfície detectada, a imagem foi formada por combinação de imagens por abertura sintética, que apresentou uma boa resolução. O uso conjunto dos métodos de detecção de superfície propostos e da formação de imagem resultou em uma taxa de imagens que permite ensaios interativos,com processamento em uma CPU de uso geral. / In ultrasound imaging ofobjects with curved surface, in immersion, the refraction eects must be compensated for if the propagation speed in the uid is very dierent from the speed in the object. In this case the surface position and shape must be known. In this work the surface is detected by the same linear array that captures the signals for the image formation. Two fast methods for surface detection were proposed,one is based on image formation techniques and another utilizes the echotime-of-ight information directly from the ultra\\sound signals.The two methods were compared,and the image-based method was more tolerant of signal errors, while the time-of-ight-based method was faster. After the surface detection, the image was formed by combination of synthetic aperture images,with a good resulting resolution. The utilization of the proposed surface detection methods together with the image formation resulted in an image rate that allows interactive testing, with processing on a general-purpose CPU.

The present research project is carried out at the von Karman Institute for Fluid Dynamics (Rhode-Saint-Genèse, Belgium) with the financial support of the European Space Agency. The first stage of spacecrafts (e.g. Ariane 5, Vega, Shuttle) generally consists of large solid propellant rocket motors (SRM), which often consist of segmented structure and incorporate a submerged nozzle. During the combustion, the regression of the solid propellant surrounding the nozzle integration part leads to the formation of a cavity around the nozzle lip. The propellant combustion generates liquefied alumina droplets coming from chemical reaction of the aluminum composing the propellant grain. The alumina droplets being carried away by the hot burnt gases are flowing towards the nozzle. Meanwhile the droplets may interact with the internal flow. As a consequence, some of the droplets are entrapped in the cavity forming an alumina puddle (slag) instead of being exhausted through the throat. This slag reduces the performances. The aim of the present study is to characterize the slag accumulation process in a simplified model of the MPS P230 motor using primarily optical experimental techniques. Therefore, a 2D-like cold-gas model is designed, which represents the main geometrical features of the real motor (presence of an inhibitor, nozzle and cavity) and allows to approximate non-dimensional parameters of the internal two-phase flow (e.g. Stokes number, volume fraction). The model is attached to a wind-tunnel that provides quasi-axial flow (air) injection. A water spray device in the stagnation chamber realizes the models of the alumina droplets, which are accumulating in the aft-end cavity of the motor. To be able to carry out experimental investigation, at first the the VKI Level Detection and Recording(LeDaR) and Particle Image Velocimetry (PIV) measurement techniques had to be adapted to the two-phase flow condition of the facility. A parametric liquid accumulation assessment is performed experimentally using the LeDaR technique to identify the influence of various parameters on the liquid deposition rate. The obstacle tip to nozzle tip distance (OT2NT) is identified to be the most relevant, which indicates how much a droplet passing just at the inhibitor tip should deviate transversally to leave through the nozzle and not to be entrapped in the cavity. As LeDaR gives no indication of the driving mechanisms, the flow field is analysed experimentally, which is supported by numerical simulations to understand the main driving forces of the accumulation process. A single-phase PIV measurement campaign provides detailed information about the statistical and instantaneous flow structures. The flow quantities are successfully compared to an equivalent 3D unsteady LES numerical model. Two-phase flow CFD simulations suggest the importance of the droplet diameter on the accumulation rate. This observation is confirmed by two-phase flow PIV experiments as well. Accordingly, the droplet entrapment process is described by two mechanisms. The smaller droplets (representing a short characteristic time) appear to follow closely the air-phase. Thus, they may mix with the air-phase of the recirculation region downstream the inhibitor and can be carried into the cavity. On the other hand, the large droplets (representing a long characteristic time) are not able to follow the air-phase motion. Consequently, a large mean velocity difference is found between the droplets and the air-phase using the two-phase flow measurement data. Therefore, due to the inertia of the large droplets, they may fall into the cavity in function of the OT2NT and their velocity vector at the level of the inhibitor tip. Finally, a third mechanism, dripping is identified as a contributor to the accumulation process. In the current quasi axial 2D-like set-up large drops are dripping from the inhibitor. In this configuration they are the main source of the accumulation process. Therefore, additional numerical simulations are performed to estimate the importance of dripping in more realistic configurations. The preliminary results suggest that dripping is not the main mechanism in the real slag accumulation process. However, it may still lead to a considerable contribution to the final amount of slag.

Antalet enheter som är uppkopplade till internet, Internet of Things (IoT), ökar hela tiden. År 2035 beräknas det finnas 1000 miljarder Internet of Things-enheter. Samtidigt som antalet enheter ökar, ökar belastningen på internet-nätverken som enheterna är uppkopplade till. Internet of Things-enheterna som finns i vår omgivning samlar in data som beskriver den fysiska tillvaron och skickas till molnet för beräkning. För att hantera belastningen på internet-nätverket flyttas beräkningarna på datan till IoT-enheten, istället för att skicka datan till molnet. Detta kallas för edge computing. IoT-enheter är ofta resurssnåla enheter med begränsad beräkningskapacitet. Detta innebär att när man designar exempelvis "machine learning"-modeller som ska köras med edge computing måste algoritmerna anpassas utifrån de resurser som finns tillgängliga på enheten. I det här arbetet har vi utvärderat olika multilayer perceptron-modeller för mikrokontrollers utifrån en rad olika experiment. "Machine learning"-modellerna har varit designade att detektera vägunderlag. Målet har varit att identifiera hur olika parametrar påverkar "machine learning"-systemen. Vi har försökt att maximera prestandan och minimera den mängd fysiskt minne som krävs av modellerna. Vi har även behövt förhålla oss till att mikrokontrollern inte haft tillgång till internet. Modellerna har varit ämnade att köras på en mikrokontroller "on the edge". Datainsamlingen skedde med hjälp av en accelerometer integrerad i en mikrokontroller som monterades på en cykel. I studien utvärderas två olika "machine learning"-system, ett som är en kombination av binära klassificerings modeller och ett multiklass klassificerings system som framtogs i ett tidigare arbete. Huvudfokus i arbetet har varit att träna modeller för klassificering av vägunderlag och sedan utvärdera modellerna. Datainsamlingen gjordes med en mikrokontroller utrustad med en accelerometer monterad på en cykel. Ett av systemen lyckas uppnå en träffsäkerhet på 93,1\% för klassificering av 3 vägunderlag. Arbetet undersöker även hur mycket fysiskt minne som krävs av de olika "machine learning"-systemen. Systemen krävde mellan 1,78kB och 5,71kB i fysiskt minne. / The number of devices connected to the internet, the Internet of Things (IoT), is constantly increasing. By 2035, it is estimated to be 1,000 billion Internet of Things devices in the world. At the same time as the number of devices increase, the load on the internet networks to which the devices are connected, increases. The Internet of Things devices in our environment collect data that describes our physical environment and is sent to the cloud for computation. To reduce the load on the internet networks, the calculations are done on the IoT devices themselves instead of in the cloud. This way no data needs to be sent over the internet and is called edge computing. In edge computing, however, other challenges arise. IoT devices are often resource-efficient devices with limited computing capacity. This means that when designing, for example, machine learning models that are to be run with edge computing, the models must be designed based on the resources available on the device. In this work, we have evaluated different multilayer perceptron models for microcontrollers based on a number of different experiments. The machine learning models have been designed to detect road surfaces. The goal has been to identify how different parameters affect the machine learning systems. We have tried to maximize the performance and minimize the memory allocation of the models. The models have been designed to run on a microcontroller on the edge. The data was collected using an accelerometer integrated in a microcontroller mounted on a bicycle. The study evaluates two different machine learning systems that were developed in a previous thesis. The main focus of the work has been to create algorithms for detecting road surfaces. The data collection was done with a microcontroller equipped with an accelerometer mounted on a bicycle. One of the systems succeeds in achieving an accuracy of 93.1\% for the classification of 3 road surfaces. The work also evaluates how much physical memory is required by the various machine learning systems. The systems required between 1.78kB and 5,71kB of physical memory.

The present research project is carried out at the von Karman Institute for Fluid Dynamics (Rhode-Saint-Genèse, Belgium) with the financial support of the European Space Agency.<p><p>The first stage of spacecrafts (e.g. Ariane 5, Vega, Shuttle) generally consists of large solid propellant rocket motors (SRM), which often consist of segmented structure and incorporate a submerged nozzle. During the combustion, the regression of the solid propellant surrounding the nozzle integration part leads to the formation of a cavity around the nozzle lip. The propellant combustion generates liquefied alumina droplets coming from chemical reaction of the aluminum composing the propellant grain. The alumina droplets being carried away by the hot burnt gases are flowing towards the nozzle. Meanwhile the droplets may interact with the internal flow. As a consequence, some of the droplets are entrapped in the cavity forming an alumina puddle (slag) instead of being exhausted through the throat. This slag reduces the performances.<p><p>The aim of the present study is to characterize the slag accumulation process in a simplified model of the MPS P230 motor using primarily optical experimental techniques. Therefore, a 2D-like cold-gas model is designed, which represents the main geometrical features of the real motor (presence of an inhibitor, nozzle and cavity) and allows to approximate non-dimensional parameters of the internal two-phase flow (e.g. Stokes number, volume fraction). The model is attached to a wind-tunnel that provides quasi-axial flow (air) injection. A water spray device in the stagnation chamber realizes the models of the alumina droplets, which are accumulating in the aft-end cavity of the motor.<p><p>To be able to carry out experimental investigation, at first the the VKI Level Detection and Recording(LeDaR) and Particle Image Velocimetry (PIV) measurement techniques had to be adapted to the two-phase flow condition of the facility.<p><p>A parametric liquid accumulation assessment is performed experimentally using the LeDaR technique to identify the influence of various parameters on the liquid deposition rate. The obstacle tip to nozzle tip distance (OT2NT) is identified to be the most relevant, which indicates how much a droplet passing just at the inhibitor tip should deviate transversally to leave through the nozzle and not to be entrapped in the cavity.<p><p>As LeDaR gives no indication of the driving mechanisms, the flow field is analysed experimentally, which is supported by numerical simulations to understand the main driving forces of the accumulation process. A single-phase PIV measurement campaign provides detailed information about the statistical and instantaneous flow structures. The flow quantities are successfully compared to an equivalent 3D unsteady LES numerical model.<p><p>Two-phase flow CFD simulations suggest the importance of the droplet diameter on the accumulation rate. This observation is confirmed by two-phase flow PIV experiments as well. Accordingly, the droplet entrapment process is described by two mechanisms. The smaller droplets (representing a short characteristic time) appear to follow closely the air-phase. Thus, they may mix with the air-phase of the recirculation region downstream the inhibitor and can be carried into the cavity. On the other hand, the large droplets (representing a long characteristic time) are not able to follow the air-phase motion. Consequently, a large mean velocity difference is found between the droplets and the air-phase using the two-phase flow measurement data. Therefore, due to the inertia of the large droplets, they may fall into the cavity in function of the OT2NT and their velocity vector at the level of the inhibitor tip.<p><p>Finally, a third mechanism, dripping is identified as a contributor to the accumulation process. In the current quasi axial 2D-like set-up large drops are dripping from the inhibitor. In this configuration they are the main source of the accumulation process. Therefore, additional numerical simulations are performed to estimate the importance of dripping in more realistic configurations. The preliminary results suggest that dripping is not the main mechanism in the real slag accumulation process. However, it may still lead to a considerable contribution to the final amount of slag.<p> / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished