1 |
Design, Analysis, Modeling and Testing of a Micro-scale Refrigeration SystemGuo, Dongzhi 01 September 2014 (has links)
Chip scale refrigeration system is critical for the development of electronics with the rapid increase of power consumption and substantial reduction of device size, resulting in an emergent demand on novel cooling technologies with a high efficiency for the thermal management. In this thesis, active refrigeration devices based on Stirling cycle and an electrocaloric material, are designed and investigated to achieve a high cooling performance. Firstly, a new Stirling micro-refrigeration system composed of arrays of silicon MEMS cooling elements is designed and evaluated. The cooling elements are fabricated in a stacked array on a silicon wafer. A regenerator is placed between the compression (hot side) and expansion (cold side) diaphragms, which are driven electrostatically. Under operating conditions, the hot and cold diaphragms oscillate sinusoidally and out of phase such that heat is extracted to the expansion space and released from the compression space. A first-order of thermodynamic analysis is performed to study the effect of geometric parameters. Losses due to regenerator non-idealities and chamber heat transfer limitation are estimated. A multiphysics computational approach for analyzing the system performance that considers compressible flow and heat transfer with a large deformable mesh is demonstrated. The optimal regenerator porosity for the best system COP (coefficient of performance) is identified. To overcome the computational complexity brought about by the fine pillar structure in the regenerator, a porous medium model is used to allow for modeling of a full element. The analysis indicates the work recovery of the system and the diaphragm actuation are main challenges for this cooler design.The pressure drop and friction factor of gas flow across circular silicon micro pillar arrays fabricated by deep reactive ion etch (DRIE) process are investigated. A new correlation that considers the coupled effect of pillar spacing and aspect ratio, is proposed to predict the friction factor in a Reynolds v number range of 1-100. Silicon pillars with large artificial roughness amplitudes is also fabricated, and the effect of the roughness is studied in the laminar flow region. The significant reduction of pressure drop and friction factor indicates that a large artificial roughness could be built for pillar arrays in the regenerator to enhance the micro-cooler efficiency. The second option is to develop a fluid-based refrigeration system using an electrocaloric material poly(vinylidene fluoride-trifluoroethylene-chlorofluoroethylene) [P(VDF-TrFE-CFE)] terpolymer. Each cooling element includes two diaphragm actuators fabricated in the plane of a silicon wafer, which drive a heat transfer fluid back and forth across terpolymer layers that are placed between them. Finite element simulations with an assumption of sinusoidal diaphrahm motions are conducted to explore the system performance detailedly, including the effects of the applied electric field, geometric dimensions, operating frequency and externally-applied temperature span. Multiphysics modeling coupled with solid-fluid interaction, heat transfer, electrostatics, porous medium and moving mesh technique is successfully performed to verify the thermal modeling feasibility. The electrocaloric effect in thin films of P(VDF-TrFE-CFE) terpolymer is directly measured by infrared imaging at ambient conditions. At an electric field of 90 V/μm, an adiabatic temperature change of 5.2 °C is obtained and the material performance is stable over a long testing period. These results suggest that application of this terpolymer is promising for micro-scale refrigeration.
|
2 |
Assessing the late Holocene14C reservoir age of theChukchi Sea with the AniakchakCFE II tephra 3.6 kyr BPGeels, Alexis January 2019 (has links)
Tephrochronology is a powerful tool to correlate and improve the chronology of sedimentaryarchives in the Arctic Ocean. The Aniakchak Caldera Forming Eruption (CFE) in Alaska at3.6 cal kyr BP ejected ash that were found in a widespread layer in Alaska, and as cryptotephrain the Chukchi Sea, Newfoundland, and Greenland. This study presents data from the coreSWERUS-L2-4-PC1 (4PC) taken at a water depth of 120 m in the Chukchi Sea. The sharp peakin tephra shards concentration permitted to clearly place the isochron. Unfortunately, the microprobeanalyses were unsuccessful, however measurements of trace elements were performedwith Laser Ablated-Inductively Coupled Plasma-Mass Spectrometry (LA-ICP-MS). The geochemicalsignature of the Aniakchak 3.6 eruption was ensured with significant trace elementratios. The isochron of the eruption combined with the radiocarbon dates from 4PC permittedto calculate the local marine radiocarbon reservoir age offset DR=36446 years. This value isrelatively low compared to recent estimates in the Chukchi Sea, especially to the neighbouringcore SWERUS-L2-2-PC1 were DR=47760. The DR value of this study is explained by theinfluence of the "young" Atlantic water mixing with the "old" Pacific water at the depth wherethe core was taken.
|
3 |
Radômes actifs utilisant des matériaux et structures à propriétés électromagnétiques contrôléesLunet, Guillaume 28 October 2009 (has links)
Les recherches que nous présentons dans ce mémoire s'inscrivent dans le cadre du développement de nouvelles structures et de l'étude de matériaux accordables en vue d'une intégration industrielle comme radôme actif.Plus particulièrement, ils consistent en la réalisation d'un dispositif micro-onde permettant à la fois un filtrage et une agilité fréquentiels en espace libre. Des structures basées sur des surfaces sélectives en fréquences, pour l'aspect filtrage, et sur des matériaux de type ferroélectrique, pour l'aspect accordabilité, sont développées. Des modélisations et des simulations électromagnétiques montrent que le changement de permittivité du matériau, obtenu par application d'un champ électrique externe, permet le pilotage fréquentiel de la transmission de la structure. Une mise en oeuvre expérimentale complète ces travaux, au cours de laquelle des prototypes ont été fabriqués par des techniques de photolithographie, puis caractérisés en espace libre grâce à un banc ABmm. Les mesures micro-ondes valident ainsi les résultats de simulations menées en amont et montrent les possibilités de contrôler la fréquence de transmission du radôme. / The research we present in this memory registers within the framework to develop new structures and to study tunable materials for an industrial integration as an active radome. Specifically, they consist of achieving a free space microwave device for both a filtering behaviour and a frequency agility behaviour. Structures based on frequency selective surfaces, for the filtering aspect, and on ferroelectric materials for the tuning aspect, are developed. Modeling and simulations show that the change of the material permittivity, obtained by applying an external electric field, enable piloting the transmission frequency of the structure. An experimental implementation complete this work and prototypes have been fabricated by photolithography techniques and then characterized in free space with a bench ABmm. Thus, microwave measurements validate the results of simulations and show the possibility to control the frequency transmission of the radome.
|
4 |
Aplicação de transistores orgânicos na fabricação de inversores lógicos digitais / Organic transistors and their application in organic logic invertersCardoso, Lilian Soares 09 December 2016 (has links)
Esta tese tem por objetivo o desenvolvimento de metodologias eficientes e de baixo custo para ajustar as propriedades elétricas de OFETs de canal p e de canal n, a fim de possibilitar a fabricação do circuito complementar orgânico, semelhante a uma estrutura CMOS. O desempenho do circuito complementar fabricado foi otimizado, e também foi confeccionado por impressão um OFETs de canal operando em baixas tensões. Para a fabricação do CMOS orgânico foi proposto um método baseado na seleção adequada do solvente da camada dielétrica para ajustar o desempenho elétrico dos OFETs de canal p e de canal n. Os solventes, MEK, nBA e DMSO foram selecionados para a dissolução do PMMA por apresentarem diferenças nos valores de momento de dipolo, de ponto de ebulição e de graus de ortogonalidade em relação as camadas semicondutoras de P3HT e de P(NDI2OD-T2) dos OFETs. A análise dos resultados dos OFETs de canal p e de canal n demonstrou que a metodologia proposta é adequada tanto para o ajuste das propriedades elétricas destes dispositivos quanto para a otimização do desempenho dos mesmos. Os melhores desempenhos elétricos para os OFETs de canal p e de canal n foram obtidos quando utilizados o DMSO e o MEK como solventes do PMMA, respectivamente, devido à perfeita ortogonalidade destes solventes em relação às camadas semicondutoras. Os OFETs de canal p que utilizaram o DMSO e os OFETs de canal n que utilizaram o nBA foram os que apresentaram desempenhos elétricos semelhantes, sendo portanto aplicados na fabricação do CMOS. Valores de ganho entre 6,8 e 7,8 e de margem de ruído entre 28,3 V e 34,5 V foram obtidos para inversores complementares fabricados nesta etapa do trabalho. OFETs de canal p utilizando uma blenda de PTAA: diF TES ADT como camada semicondutora, o PEDOT:PSS como eletrodos dreno/fonte e o P(VDF-TrFE-CFE) como camada dielétrica também foram fabricados neste trabalho. A técnica de blade-coating foi utilizada para a deposição dos eletrodos dreno/fonte e da camada semicondutora, ao passo que a técnica de spray-coating foi utilizada para a deposição da camada dielétrica. Da análise dos resultados foi possível inferir que a utilização de um dielétrico com elevada constante dielétrica (K), como o P(VDF-TrFE-CFE), possibilita o funcionamento dos transistores a baixas tensões (≤ 8 V), porém com valores de mobilidade reduzidos devido à elevada desordem dipolar na interface provocada por este dielétrico. Para minimizar esses efeitos, uma fina camada de um polímero fluorado foi depositada entre a camada semicondutora e a dielétrica pela técnica de blade-coating, constituindo assim uma bicamada dielétrica nos OFETs. Dos resultados das medidas elétricas dos OFETs constituídos pela bicamada dielétrica foi observada permanência do funcionamento destes dispositivos a tensões inferiores a 8 V com desempenho elétricos superiores a resultados já publicados na literatura. Por fim, inversores lógicos unipolares com transistores de carga foram fabricados com os OFETs que utilizaram a bicamada dielétrica, sendo obtidos valores de ganho entre 1,2 e 1,6 e de margem de ruído entre 56% e 68,5% de ½ VDD. / This thesis aimed to develop an efficient and low cost method to adjust the electrical properties of p- and n-channel OFETs to allow us to build an organic CMOS and the optimization of printed p-channel OFETs to work at low voltages. We proposed a method to fabricate the organic CMOS, based on the careful selection of dielectric solvent, which was adjusted to obtain the best performance of p- and n-channel OFETs. The dielectric solvents as MEK, nBA and DMSO were selected to dissolve the PMMA dielectric polymer due their different physical properties as dipole moment and boiling point and because they showed slightly different degrees of orthogonality to the P3HT and P(NDI2OD-T2) semiconductor layers of the OFETs. The results showed that the careful selection of the dielectric solvent not only allows to tune the electrical characteristics of the p- and n-channel OFETs, but also to improve the performance of these devices. The best performances were achieved when DMSO and MEK were used as dielectric solvents of the p and n-channel OFETs, respectively, as result of the perfectly orthogonality of these solvents to the semiconductor layers. P-channel OFETs using DMSO and n-channel OFETs using nBA showed similar electrical characteristics and thus, they were used to construct the organic CMOS. The organic complementary inverters showed high gain and noise margin values in the range of 6,8 to 7,8 and 28,3 V to 34,5 V, respectively. Printed p-channel OFETs were also fabricated, in which the blend PTAA:diF TES ADT was used as semiconductor channel, PEDOT:PSS as the drain/source electrodes and P(VDF-TrFE-CFE) as the dielectric layer. The blade-coating technique was used to deposit the source/drain electrodes and the semiconductor layer, while the spray-coating technique was used to deposit the dielectric layer. It was observed that using high-k dielectric as P(VDF-TrFE-CFE) enable to reduce the operating voltage of the OFETs (≤8 V), however, this high-k dielectric also reduced the field effect mobility due the dipolar disorder at the semiconductor/dielectric interface. To minimize the dipolar issue at the interface, we inserted a thin fluoropolymer dielectric layer by blade-coating between the semiconductor and the high-k dielectric layers, thus constituting a dielectric bilayer on the OFETs. From the electrical measurements of the OFETs with the dielectric bilayer, it was observed that the devices were still working at 8 V and they also showed better performance in comparison to results already published. Finally, organic unipolar inverters with load transistors were fabricated using the p-channel OFETs with the dielectric bilayer and they showed reasonable performance, with gain and noise margin in the range of 1,2 to 1,6 and 56% e 68,5% of ½ VDD, respectively.
|
5 |
Aplicação de transistores orgânicos na fabricação de inversores lógicos digitais / Organic transistors and their application in organic logic invertersLilian Soares Cardoso 09 December 2016 (has links)
Esta tese tem por objetivo o desenvolvimento de metodologias eficientes e de baixo custo para ajustar as propriedades elétricas de OFETs de canal p e de canal n, a fim de possibilitar a fabricação do circuito complementar orgânico, semelhante a uma estrutura CMOS. O desempenho do circuito complementar fabricado foi otimizado, e também foi confeccionado por impressão um OFETs de canal operando em baixas tensões. Para a fabricação do CMOS orgânico foi proposto um método baseado na seleção adequada do solvente da camada dielétrica para ajustar o desempenho elétrico dos OFETs de canal p e de canal n. Os solventes, MEK, nBA e DMSO foram selecionados para a dissolução do PMMA por apresentarem diferenças nos valores de momento de dipolo, de ponto de ebulição e de graus de ortogonalidade em relação as camadas semicondutoras de P3HT e de P(NDI2OD-T2) dos OFETs. A análise dos resultados dos OFETs de canal p e de canal n demonstrou que a metodologia proposta é adequada tanto para o ajuste das propriedades elétricas destes dispositivos quanto para a otimização do desempenho dos mesmos. Os melhores desempenhos elétricos para os OFETs de canal p e de canal n foram obtidos quando utilizados o DMSO e o MEK como solventes do PMMA, respectivamente, devido à perfeita ortogonalidade destes solventes em relação às camadas semicondutoras. Os OFETs de canal p que utilizaram o DMSO e os OFETs de canal n que utilizaram o nBA foram os que apresentaram desempenhos elétricos semelhantes, sendo portanto aplicados na fabricação do CMOS. Valores de ganho entre 6,8 e 7,8 e de margem de ruído entre 28,3 V e 34,5 V foram obtidos para inversores complementares fabricados nesta etapa do trabalho. OFETs de canal p utilizando uma blenda de PTAA: diF TES ADT como camada semicondutora, o PEDOT:PSS como eletrodos dreno/fonte e o P(VDF-TrFE-CFE) como camada dielétrica também foram fabricados neste trabalho. A técnica de blade-coating foi utilizada para a deposição dos eletrodos dreno/fonte e da camada semicondutora, ao passo que a técnica de spray-coating foi utilizada para a deposição da camada dielétrica. Da análise dos resultados foi possível inferir que a utilização de um dielétrico com elevada constante dielétrica (K), como o P(VDF-TrFE-CFE), possibilita o funcionamento dos transistores a baixas tensões (≤ 8 V), porém com valores de mobilidade reduzidos devido à elevada desordem dipolar na interface provocada por este dielétrico. Para minimizar esses efeitos, uma fina camada de um polímero fluorado foi depositada entre a camada semicondutora e a dielétrica pela técnica de blade-coating, constituindo assim uma bicamada dielétrica nos OFETs. Dos resultados das medidas elétricas dos OFETs constituídos pela bicamada dielétrica foi observada permanência do funcionamento destes dispositivos a tensões inferiores a 8 V com desempenho elétricos superiores a resultados já publicados na literatura. Por fim, inversores lógicos unipolares com transistores de carga foram fabricados com os OFETs que utilizaram a bicamada dielétrica, sendo obtidos valores de ganho entre 1,2 e 1,6 e de margem de ruído entre 56% e 68,5% de ½ VDD. / This thesis aimed to develop an efficient and low cost method to adjust the electrical properties of p- and n-channel OFETs to allow us to build an organic CMOS and the optimization of printed p-channel OFETs to work at low voltages. We proposed a method to fabricate the organic CMOS, based on the careful selection of dielectric solvent, which was adjusted to obtain the best performance of p- and n-channel OFETs. The dielectric solvents as MEK, nBA and DMSO were selected to dissolve the PMMA dielectric polymer due their different physical properties as dipole moment and boiling point and because they showed slightly different degrees of orthogonality to the P3HT and P(NDI2OD-T2) semiconductor layers of the OFETs. The results showed that the careful selection of the dielectric solvent not only allows to tune the electrical characteristics of the p- and n-channel OFETs, but also to improve the performance of these devices. The best performances were achieved when DMSO and MEK were used as dielectric solvents of the p and n-channel OFETs, respectively, as result of the perfectly orthogonality of these solvents to the semiconductor layers. P-channel OFETs using DMSO and n-channel OFETs using nBA showed similar electrical characteristics and thus, they were used to construct the organic CMOS. The organic complementary inverters showed high gain and noise margin values in the range of 6,8 to 7,8 and 28,3 V to 34,5 V, respectively. Printed p-channel OFETs were also fabricated, in which the blend PTAA:diF TES ADT was used as semiconductor channel, PEDOT:PSS as the drain/source electrodes and P(VDF-TrFE-CFE) as the dielectric layer. The blade-coating technique was used to deposit the source/drain electrodes and the semiconductor layer, while the spray-coating technique was used to deposit the dielectric layer. It was observed that using high-k dielectric as P(VDF-TrFE-CFE) enable to reduce the operating voltage of the OFETs (≤8 V), however, this high-k dielectric also reduced the field effect mobility due the dipolar disorder at the semiconductor/dielectric interface. To minimize the dipolar issue at the interface, we inserted a thin fluoropolymer dielectric layer by blade-coating between the semiconductor and the high-k dielectric layers, thus constituting a dielectric bilayer on the OFETs. From the electrical measurements of the OFETs with the dielectric bilayer, it was observed that the devices were still working at 8 V and they also showed better performance in comparison to results already published. Finally, organic unipolar inverters with load transistors were fabricated using the p-channel OFETs with the dielectric bilayer and they showed reasonable performance, with gain and noise margin in the range of 1,2 to 1,6 and 56% e 68,5% of ½ VDD, respectively.
|
6 |
Nuage hypermassif, chocs et efficacité de formation stellaire / Hypermassive cloud, shock and stellar formation efficiencyLouvet, Fabien 22 September 2014 (has links)
Les étoiles massives, de type O ou B, sont d'une importance capitale pour le budget énergétique des galaxies et l'enrichissement du milieu interstellaire. Néanmoins, leur formation, contrairement à celle des étoiles de type solaire reste sujet à débats, sinon une énigme. Les toutes premières étapes de la formation des étoiles massives ainsi que la formation de leur nuage parent sont des thèmes qui stimulent une grande activité sur les plans théorique et observationnel depuis une décennie. Il semble maintenant acquis que les étoiles massives naissent dans des cœurs denses massifs, qui se forment au travers de processus dynamiques, tels que les flots de gaz collisionnels. Au cours de ma thèse, j'ai mené une étude approfondie de la formation des cœurs denses et des étoiles massives au sein de la structure hypermassive W43-MM1, localisée à 6~kpc du soleil. Dans un premier temps, j'ai montré une corrélation directe entre l'efficacité à former des étoiles et la densité volumique des nuages moléculaires, en décalage avec un certain nombre d'études précédentes. En effet, la distribution spatiale et de masse des cœurs denses massifs en formation au sein de W43-MM1 suggère que ce filament hypermassif est en phase de flambée de formation d'étoiles, flambée d'autant plus grande que l'on se rapproche de son cœur. J'ai comparé ces résultats observationnels aux modèles numériques et analytiques d'efficacité de formation stellaire les plus récents. Cette confrontation permet non seulement d'apporter de nouvelles contraintes sur la formation des filaments hypermassifs, mais suggère aussi que la compréhension de la formation d'étoiles dans les nuages hypermassifs nécessite une description fine de la structure de ces objets exceptionnels. En second lieu, ayant montré que la formation des étoiles massives est fortement dépendante des propriétés des filaments qui les forment, je me suis naturellement intéressé aux processus de formation de ces filaments, grâce à une étude de leur dynamique globale. Plus précisément, j'ai utilisé un traceur de chocs (la molécule de SiO) pour discerner les chocs dûs aux processus locaux de formation des étoiles (jets et flots bipolaires), des chocs dûs aux processus permettant la formation du nuage. J'ai ainsi pu, via une étude sans précédent alliant observations et modélisation de chocs dans une région formant de nombreuses étoiles, montrer l'existence de chocs à basse vitesse, première signature directe de la formation du nuage moléculaire dans lequel les étoiles massives se forment. Ces résultats constituent une étape importante reliant, via des processus dynamiques, la formation des nuages moléculaires à la formation des étoiles massives. / O and B types stars are of paramount importance in the energy budget of galaxies and play a crucial role enriching the interstellar medium. However, their formation, unlike that of solar-type stars, is still subject to debate, if not an enigma. The earliest stages of massive star formation and the formation of their parent cloud are still crucial astrophysical questions that drew a lot of attention in the community, both from the theoretical and observational perspective, during the last decade. It has been proposed that massive stars are born in massive dense cores that form through very dynamic processes, such as converging flows of gas. During my PhD, I conducted a thorough study of the formation of dense cores and massive stars in the W43-MM1 supermassive structure, located at ~ 6 kpc from the sun. At first, I showed a direct correlation between the star formation efficiency and the volume gas density of molecular clouds, in contrast with scenarii suggested by previous studies. Indeed, the spatial distribution and mass function of the massive dense cores currently forming in W43-MM1 suggests that this supermassive filament is undergoing a star formation burst, increasing as one approaches its center. I compared these observational results with the most recent numerical and analytical models of star formation. This comparison not only provides new constraints on the formation of supermassive filaments, but also suggests that understanding star formation in high density, extreme ridges requires a detailed portrait of the structure of these exceptional objects. Second, having shown that the formation of massive stars depends strongly on the properties of the ridges where they form, I studied the formation processes of these filaments, thanks of the characterization of their global dynamics. Specifically, I used a tracer of shocks (SiO molecule) to disentangle the feedback of local star formation processes (bipolar jets and outflows) from shocks tracing the pristine formation processes of the W43-MM1 cloud. I was able, via an unprecedented study combining observations and modeling of shocks in a starbust region, to show the existence of widespread low velocity shocks, that are the first direct signature of the formation of the massive molecular cloud from which massive stars form.These results are an important step connecting, via dynamical processes, the formation of molecular clouds to the formation of massive stars.
|
7 |
A Unified, Configurable, Non-Iterative Guidance System For Launch VehiclesRajeev, U P 12 1900 (has links)
A satellite launch vehicle not subjected to any perturbations, external or internal, could be guided along a trajectory by following a stored, pre-computed steering program. In practice, perturbations do occur, and in order to take account of them and to achieve an accurate injection, a closed loop guidance system is required. Guidance algorithm is developed by solving the optimal control problem. Closed form solution is difficult because the necessary conditions are in the form of Two Point Boundary Value Problems (TBVP) or Multi Point Boundary Value Problems (MPBVP). Development of non-iterative guidance algorithm is taken as a prime objective of this thesis to ensure reliable on-board implementation. If non-iterative algorithms are required, the usual practice is to approximate the system equations to derive closed form solutions. In the present work, approximations cannot be used because the algorithm has to cater to a wide variety of vehicles and missions. Present development adopts an alternate approach by splitting the reconfigurable algorithm development in to smaller sub-problems such that each sub-problem has closed form solution. The splitting is done in such a way that the solution of the sub-problems can be used as building blocks to construct the final solution. By adding or removing the building blocks, the algorithm can be configured to suit specific requirements.
Chapter 1 discusses the motivation and objectives of the thesis and gives a literature survey. In chapter 2, Classical Flat Earth (CFE) guidance algorithm is discussed. The assumptions and the nature of solution are closely analyzed because CFE guidance is used as the baseline for further developments. New contribution in chapter 2 is the extension of CFE guidance for a generalized propulsion system in which liquid and solid engines are present.
In chapter 3, CFE guidance is applied for a mission with large pitch steering angles. The result shows loss of optimality and performance. An algorithm based on regular perturbation is developed to compensate for the small angle approximation. The new contribution in chapter 3 is the development of Regular Perturbation based FE (RPFE) guidance as an extension of CFE guidance. RPFE guidance can be configured as CFE guidance and FEGP.
Algorithms presented up to chapter 3 are developed to inject a satellite in to orbits with unspecified inertial orientation. Communication satellite missions demand injection in to an orbit with a specific inertial orientation defined by argument of perigee. This problem is formulated using Calculus of Variations in chapter 4. A non-iterative closed form solution (Predicted target Flat Earth or PFE guidance) is derived for this problem.
In chapter 5, PFE guidance is extended to a multi-stage vehicle with a constraint on the impact point of spent lower stage. Since the problem is not analytically solvable, the original problem is split in to three sub-problems and solved.
Chapter 6 has two parts. First part gives theoretical analysis of the sub-optimal strategies with special emphasis to guidance. Behavior of predicted terminal error and control commands in presence of plant approximations are theoretically analyzed for a class of optimal control problems and the results are presented as six theorems. Chapter 7 presents the conclusions and future works.
|
8 |
Predicting Drag Polars For Micro Air VehiclesLuke, Mark Elden 03 November 2003 (has links) (PDF)
Drag polars for three Micro Air Vehicles (MAVs) were measured at Reynolds numbers of 70,000, 50,000, 30,000, and 10,000 and compared to predictions generated using the classical approach. The MAVs tested had different configurations and aspect ratios varying from 1.2 to 1.6 and ratios of wetted surface area to planform area from 2.6 to 3.9. A force balance was used to measure the lift and drag on the MAVs at angles of attack ranging from -5 degrees (or -10 degrees) to 10 degrees. The force balance allowed the MAVs to rotate in the pitching axis. The MAV angle of attack was set using an elevator installed on the MAV and controlled using a standard radio control used by RC plane enthusiasts. Uncertainty analysis performed on the data showed the uncertainty for high Reynolds numbers was dominated by velocity uncertainty, and uncertainty for the lower Reynolds numbers was dominated by uncertainty in the force measurements. Agreement between measured and predicted drag polars was good with the measured drag never being more than two times the predicted drag. For the majority of the tests, the drag coefficients followed the expected Reynolds number trend: increasing with decreasing Reynolds number.
|
9 |
Computationally Efficient Explainable AI: Bayesian Optimization for Computing Multiple Counterfactual Explanantions / Beräkningsmässigt Effektiv Förklarbar AI: Bayesiansk Optimering för Beräkning av Flera Motfaktiska FörklaringarSacchi, Giorgio January 2023 (has links)
In recent years, advanced machine learning (ML) models have revolutionized industries ranging from the healthcare sector to retail and E-commerce. However, these models have become increasingly complex, making it difficult for even domain experts to understand and retrace the model's decision-making process. To address this challenge, several frameworks for explainable AI have been proposed and developed. This thesis focuses on counterfactual explanations (CFEs), which provide actionable insights by informing users how to modify inputs to achieve desired outputs. However, computing CFEs for a general black-box ML model is computationally expensive since it hinges on solving a challenging optimization problem. To efficiently solve this optimization problem, we propose using Bayesian optimization (BO), and introduce the novel algorithm Separated Bayesian Optimization (SBO). SBO exploits the formulation of the counterfactual function as a composite function. Additionally, we propose warm-starting SBO, which addresses the computational challenges associated with computing multiple CFEs. By decoupling the generation of a surrogate model for the black-box model and the computation of specific CFEs, warm-starting SBO allows us to reuse previous data and computations, resulting in computational discounts and improved efficiency for large-scale applications. Through numerical experiments, we demonstrate that BO is a viable optimization scheme for computing CFEs for black-box ML models. BO achieves computational efficiency while maintaining good accuracy. SBO improves upon this by requiring fewer evaluations while achieving accuracies comparable to the best conventional optimizer tested. Both BO and SBO exhibit improved capabilities in handling various classes of ML decision models compared to the tested baseline optimizers. Finally, Warm-starting SBO significantly enhances the performance of SBO, reducing function evaluations and errors when computing multiple sequential CFEs. The results indicate a strong potential for large-scale industry applications. / Avancerade maskininlärningsmodeller (ML-modeller) har på senaste åren haft stora framgångar inom flera delar av näringslivet, med allt ifrån hälso- och sjukvårdssektorn till detaljhandel och e-handel. I jämn takt med denna utveckling har det dock även kommit en ökad komplexitet av dessa ML-modeller vilket nu lett till att även domänexperter har svårigheter med att förstå och tolka modellernas beslutsprocesser. För att bemöta detta problem har flertalet förklarbar AI ramverk utvecklats. Denna avhandling fokuserar på kontrafaktuella förklaringar (CFEs). Detta är en förklaringstyp som anger för användaren hur denne bör modifiera sin indata för att uppnå ett visst modellbeslut. För en generell svarta-låda ML-modell är dock beräkningsmässigt kostsamt att beräkna CFEs då det krävs att man löser ett utmanande optimeringsproblem. För att lösa optimeringsproblemet föreslår vi användningen av Bayesiansk Optimering (BO), samt presenterar den nya algoritmen Separated Bayesian Optimization (SBO). SBO utnyttjar kompositionsformuleringen av den kontrafaktuella funktionen. Vidare, utforskar vi beräkningen av flera sekventiella CFEs för vilket vi presenterar varm-startad SBO. Varm-startad SBO lyckas återanvända data samt beräkningar från tidigare CFEs tack vare en separation av surrogat-modellen för svarta-låda ML-modellen och beräkningen av enskilda CFEs. Denna egenskap leder till en minskad beräkningskostnad samt ökad effektivitet för storskaliga tillämpningar. I de genomförda experimenten visar vi att BO är en lämplig optimeringsmetod för att beräkna CFEs för svarta-låda ML-modeller tack vare en god beräknings effektivitet kombinerat med hög noggrannhet. SBO presterade ännu bättre med i snitt färre funktionsutvärderingar och med fel nivåer jämförbara med den bästa testade konventionella optimeringsmetoden. Både BO och SBO visade på bättre kapacitet att hantera olika klasser av ML-modeller än de andra testade metoderna. Slutligen observerade vi att varm-startad SBO gav ytterligare prestandaökningar med både minskade funktionsutvärderingar och fel när flera CFEs beräknades. Dessa resultat pekar på stor potential för storskaliga tillämpningar inom näringslivet.
|
Page generated in 0.0276 seconds