Nanostructured ferritic alloy and silicon carbide composite materials (NFA-SiC) were sintered with spark plasma sintering (SPS) method and systematically investigated through X-ray diffraction (XRD), scanning electron microscopy (SEM), as well as density and Vickers hardness tests. Pure NFA, pure SiC, and their composites NFA-SiC with different compositions (2.5 vol% NFA-97.5 vol% SiC, 5 vol% NFA-95 vol% SiC, 97.5 vol% NFA-2.5 vol% SiC, and 95 vol% NFA-5 vol% SiC) were successfully sintered through SPS. In the high-NFA samples, pure NFA and NFA-SiC, minor gamma-Fe phase formation from the main alfa-Fe matrix occurred in pure NFA 950 degree C and 1000 degree C. The densities of the pure NFA and NFA-SiC composites increased with sintering temperature but decreased with SiC content. The Vickers hardness of the pure NFA and NFA-SiC composites was related to density and phase composition. In the high-SiC samples, NFA addition of 2.5 vol% can achieve full densification for the NFA-SiC samples at relative low temperatures. With the increase in sintering temperature, the Vickers hardness of the pure SiC and NFA-SiC composite samples were enhanced. However, the NFA-SiC composites had relative lower hardness than the pure SiC samples. A carbon layer was introduced in the NFA particles to prevent the reaction between NFA and SiC. Results indicated that the carbon layer was effective up to 1050 degree C sintering temperature. Green samples of gradient-structured NFA-SiC composites were successfully fabricated through slip casting of an NFA-SiC co-suspension. / Master of Science

Recent approval of emission standards for small utility engines by the California Air Resources Board suggested that substantial reductions in emissions from small utility engines will soon be required. Although the 1994 standards can be met by simple engine modifications, the 1999 standards may require the use of emission control technologies such as catalytic converters because they are more stringent. In this research catalytic control of individual hydrocarbons such as methane, ethylene, benzene, and toluene were evaluated. A platinum coated catalyst treated emissions from a 107cc, four-cycle gasoline engine loaded with a 1.4KW portable generator. Determination of emissions was performed at three different load levels: 0%, 50% and 92% of the engine rated load. Among the four hydrocarbons, toluene was oxidized as much as 60%, and benzene 40%, whereas ethylene remained unaffected by the catalyst. Also, a 5% to 10% methane oxidation occurred in one trial. / Master of Science

The motivation for this work stems from the enacting of stricter emissions requirements for the mid 1990's by the California Air Resources Board. It is foreseen that these requirements will favor the use of alcohol fuels in quantities comparable to the present usage of gasoline and diesel in order to reduce emissions of carbon monoxides (CO) and nitrogen oxides (NOx). The use of alcohol fuels at this level will substantially increase the amount of aldehyde emissions. This poses a problem in that aldehydes are odorants, components of photochemical smog, and volatile aldehydes are eye and respiratory tract irritants; therefore, it is only a matter of time before they too are strictly regulated. This thesis focuses on a systematic analysis of aldehyde emissions from alcohol fuels with respect to compression ratio. Compression ratio has been selected as the primary variable for this study, because alcohol-fueled vehicles are usually modified to have higher compression ratios than their gasoline-fueled counterparts in order to take advantage of alcohols' higher octane rating. The investigation is being conducted using a single-cylinder variable-compression ratio Waukesha-CFR engine. The aldehyde emissions are measured for various fuel alcohol percentages at different compression ratios and MBT timing. The effects on conventional vehicle emissions (Le. NOx, CO, unburned hydrocarbons) are also being measured so that tradeoffs between conventional emissions and aldehyde emissions can be determined. The goal of this research was to locate any trends between alcohol fuels and compression ratios which will allow for an optimization of these parameters to minimize aldehyde emissions. It was desired that this be achieved without sacrificing engine performance or increasing other regulated emissions. The variance of compression ratio was found to affect the pollutant formation process via its effects on temperature. The increasing expansion ratio, which accompanies increasing compression ratio, resulted in lower post .. expansion burned-gas temperatures. Temperature's influence on the rate of reactions was found to be the driving force in the formation of most of the pollutants. The experiment showed a definitive reduction in CO emissions with the use of alcohol fuels. The results also indicated an inherent tradeoff between NOx and formaldehyde emissions. / Master of Science

It is quite common practice in automotive and aircraft engine maintenance, operation, and research to employ any one of several types of instruments now on the market for determining the air-fuel ratio by exhaust gas analysis. It was the purpose of this investigation to determine the most important operating characteristics, especially range and accuracy, of each of these types of instruments. The theory underlying the operation of this kind of apparatus was studied critically, and certain tests were performed on representative makes in order to observe the operation of each type under service conditions. These tests consisted essentially of connecting the analyzers to the exhaust pipe of a single-cylinder engine and comparing the analyazer readings with the true air-fuel ratio determined by accurately measuring the air and fuel supplied to the engine while the instruments were being observed. This procedure was repeated for a number of different carburetor settings, all other factors being kept as nearly constant as possible during a given series of runs. The effect of variations in engine spark advance and the pressure of the gas supplied to the instruments was also investigated. The test revealed several interesting facts. Study of the operating principles of the several instruments indicated that they were limited te air fue1 ratios below about 14 to 1, and this has been conclusively proved by these experiments. This limitation applies to thermal conductivity, hot-wire catalytic, and relative density types. W While the most expensive makes of instruments were not tested, it was found that, in general, the limit of accuracy is not greater than one-half of one air-fue1 ratio, regardless of the operating principle employed. Large variations in the pressure and rate of flow of the exhaust supplied to the analyzers were found to cause considerable deviations in those instruments which did not employ some kind of device to insure a steady and uniform supply. Certain features of design and construction which effect the reliability of the various types of exhaust gas analyzers are also reviewed in this thesis, and some of the more important chemical methods of analysis are treated briefly. / Master of Science

近年來鏈結開放式資料 (Linked Open Data，簡稱LOD) 被認定含有大量潛在價值。如何蒐集與整合多元化的LOD並提供給資料分析人員進行資料的萃取與分析，已成為當前研究的重要挑戰。LOD資料是RDF (Resource Description Framework) 的資料格式。我們可以利用SPARQL來查詢RDF資料，但是目前對於大量RDF的資料除了缺少一個高性能且易擴展的儲存和查詢分析整合性系統之外，對於RDF大數據資料分析流程的研究也不夠完備。本研究以預測電影票房為例，使用DBpedia LOD資料集並連結外部電影資料庫 (例如：IMDb)，並在Spark大數據平台上進行巨量圖形的分析。首先利用簡單貝氏分類與貝氏網路兩種演算法進行電影票房預測模型實例的建構，並使用貝氏訊息準則 (Bayesian Information Criterion，簡稱BIC) 找到最佳的貝氏網路結構。接著計算多元分類的ROC曲線與AUC值來評估本案例預測模型的準確率。 / Recent years, Linked Open Data (LOD) has been identified as containing large amount of potential value. How to collect and integrate multiple LOD contents for effective analytics has become a research challenge. LOD is represented as a Resource Description Framework (RDF) format, which can be queried through SPARQL language. But large amount of RDF data is lack of a high performance and scalable storage analysis system. Moreover, big RDF data analytics pipeline is far from perfect. The purpose of this study is to exploit the above research issue. A movie box office sale prediction scenario is demonstrated by using DBpedia with external IMDb movie database. We perform the DBpedia big graph analytics on the Apache Spark platform. The movie box office prediction for optimal model selection is first evaluated by BIC. Then, Naïve Bayes and Bayesian Network optimal model’s ROC and AUC values are obtained to justify our approach.

Nanocompósitos de matriz de alumina (Al2O3) reforçada com uma segunda fase nanométrica apresentam melhores propriedades mecânicas, especialmente dureza, tenacidade à fratura e resistência ao desgaste, quando comparado à matriz monolítica. O carbeto de nióbio (NbC) possui propriedades que o tornam um material de reforço ideal em cerâmicas de matriz Al2O3, como alto ponto de fusão e dureza, baixa reatividade química e coeficiente de expansão térmica similar à Al2O3, prevenindo o aparecimento de trincas que diminuem a resistência do material. As maiores reservas de nióbio estão localizadas no Brasil e o estudo em torno do seu aproveitamento é importante para o país. Assim, o objetivo deste trabalho foi obter e caracterizar nanocompósitos de matriz de Al2O3 contendo 5% em volume de inclusões nanométricas de NbC obtidos por moagem reativa de alta energia, usando sinterização convencional, Spark Plasma Sintering (SPS) e micro-ondas. Para isso, os pós nanométricos precursores de Al2O3-NbC foram obtidos por moagem reativa de alta energia, realizada por 330 minutos em moinho tipo SPEX, desaglomerados, lixiviados com ácido clorídrico, adicionados à matriz de Al2O3 na proporção de 5% em volume e secos sob fluxo de ar. Os pós de Al2O3-5%vol.NbC foram sinterizados por diferentes métodos: convencional em atmosfera de argônio, micro-ondas e SPS, usando diferentes temperaturas. Os pós precursores foram caracterizados por difração de raios X (DRX), microscopia eletrônica de varredura (MEV) e medida de tamanho de partículas. Os nanocompósitos sinterizados convencionalmente e por SPS foram caracterizados quanto a sua microestrutura, densidade aparente em relação a densidade teórica e dureza por nanoindentação. Os nanocompósitos sinterizados por SPS foram caracterizados quanto ao módulo de Young por nanoindentação, tenacidade à fratura e resistência a flexão em três pontos. Os nanocompósitos sinterizados convencionalmente e por SPS foram caracterizados quanto à resistência ao desgaste por ensaios esfera no disco, usando esferas de WC-6%Co com cargas de 30 e 60 N e esferas de Al2O3 com cargas de 15 e 30 N. Os resultados mostram que a moagem reativa de alta energia foi completa e efetiva na obtenção de pós nanométricos, com tamanhos de cristalito iguais a 9,1 e 9,7 nm, para Al2O3 e NbC, respectivamente. Além disso, a desaglomeração, após o processo de moagem reativa de alta energia, foi eficaz na dispersão das inclusões de NbC na matriz de Al2O3. No entanto, não foi possível obter nanocompósitos de Al2O3-5%vol.NbC com alta densidade usando os processos de sinterização convencional (92-93 %DT) e micro-ondas (80-90 %DT). No processo de sinterização por SPS, os nanocompósitos apresentaram densidades próximas à teórica (99 %DT) e, consequentemente, melhores durezas e resistência ao desgaste, quando comparadas aos materiais obtidos em forno convencional. Os resultados obtidos na caracterização da resistência ao desgaste confirmaram que esta propriedade é influenciada por diversos fatores, como método e temperatura de sinterização, as esferas utilizadas como contra-materiais e cargas aplicadas durante o ensaio. Os resultados indicaram que nanocompósitos de Al2O3-5%vol.NbC sinterizados por SPS apresentam potencial para aplicações em diversos segmentos industriais, onde se exige materiais de alto desempenho mecânico e de desgaste. / Los nanocomposites de matriz alúmina (Al2O3) reforzados con una segunda fase nanométrica presentan mejores propiedades mecánicas, especialmente de dureza, tenacidad a la fractura y resistencia al desgaste, en comparación con el material monolítico de alúmina. Por otra parte, el carburo de niobio (NbC), como refuerzo de segunda fase, presenta propiedades que lo convierten en un material ideal para las cerámicas de matriz Al2O3, tales como alta temperatura de fusión, alta dureza, baja reactividad química y un coeficiente de expansión térmica similar al material de Al2O3, evitando así la aparición de grietas que disminuyen la resistencia del material. Actualmente, las mayores reservas de niobio se encuentran en Brasil y el estudio sobre su uso es un hito muy importante para el país. Por lo tanto, el objetivo de esta tesis es obtener y caracterizar nanocomposites de matriz de Al2O3 con una segunda fase del 5% en volumen de nanopartículas de NbC obtenidos por molienda reactiva de alta energía y, mediante la sinterización convencional, Spark Plasma Sintering (SPS) y microondas. Para ello, los polvos precursores nanométricos de Al2O3-NbC fueron obtenidos mediante molienda reactiva de alta energía, durante 330 minutos en molino SPEX, desaglomerados, lixiviados con ácido clorhídrico, añadidos a la matriz de Al2O3 en la proporción de 5% en volumen y secado bajo flujo de aire. Los polvos de Al2O3-5vol.%NbC fueron sinterizados por diferentes métodos: convencional bajo una atmósfera de argón, microondas y SPS usando diferentes temperaturas. Los polvos precursores se caracterizaron por difracción de rayos X (XRD), microscopía electrónica de barrido (SEM) y la medición del tamaño de partícula. Los nanocomposites sinterizados convencionalmente y mediante SPS se caracterizaron microestructuralmente, se estudió la densidad aparente y la dureza por nanoindentación. Los nanocomposites sinterizados mediante SPS fueron caracterizados mediante el módulo de Young por nanoindentación, la tenacidad a la fractura y la resistencia a la flexión en tres puntos. Por otra parte, los nanocomposites sinterizados convencionalmente y mediante SPS fueron caracterizados respecto a resistencia al desgaste mediante la técnica de \"ball-on-disc\", utilizando esferas de WC-6%Co con cargas 30 y 60 N y esferas de Al2O3 con cargas 15 y 30 N. Los resultados muestran que la molienda reactiva de alta energía ha sido completa y eficaz en la obtención de polvos nanométricos con tamaños de cristalito de 9,1 y 9,7 nm para la Al2O3 y NbC, respectivamente. Además, la desaglomeración, después del proceso de molienda reactiva de alta energía, fue eficaz en la dispersión de las inclusiones de NbC en la matriz de Al2O3. Sin embargo, no ha sido posible obtener nanocomposites de Al2O3-5vol.%NbC con alta densidad usando procesos de sinterización convencional (92-93 %DT) y microondas (80-90 %DT). En el proceso de sinterización mediante SPS, los nanocomposites presentaron densidades cercanas a la teórica (99 %DT) y, en con-secuencia, mejor dureza y resistencia al desgaste en comparación con los materiales obtenidos en un horno convencional. Los resultados correspondientes a la resistencia al desgaste han confirmado que esta propiedad está influenciada por varios factores tales como el método y temperatura de sinterización, las esferas utilizadas como contramaterial y las cargas aplicadas durante el test. Los resultados finales indicaron que los nanocomposites de Al2O3-5vol.%NbC obtenidos mediante SPS tienen un gran potencial para las distintas aplicaciones industriales, las cuales re-quieren materiales de alto rendimiento mecánico y al desgaste.

Comprendre les mécanismes de spéciation des Produits de Fission (PF) dans le combustible nucléaire est un enjeu majeur pour pouvoir estimer précisément le terme source d’un accident grave. Parmi les nombreux PF créés, certains sont très réactifs et peuvent avoir un impact radiologique important en cas de relâchement dans l’atmosphère. C’est notamment le cas du césium (Cs), du molybdène (Mo) et du baryum (Ba). C’est dans ce contexte que s’inscrit le travail de thèse qui propose d’apporter des données expérimentales sur l’effet du potentiel oxygène sur la spéciation du Cs, du Mo et du Ba dans des combustibles nucléaires, à différents stades d’un accident grave.Une approche thermodynamique a été utilisée en support à l’interprétation des données expérimentales obtenues dans le cadre de ce travail. Deux types d’échantillons ont été étudiés: des combustibles MOX irradiés et des matériaux simulant un combustible UO2 à fort taux de combustion, obtenus par frittage à haute température (SIMFuel). Les échantillons ont été traités thermiquement dans des conditions représentatives d’un accident grave survenant dans un Réacteur à Eau Pressurisée (REP). Les conditions expérimentales ont couvert une gamme de température allant de 400°C à 2530°C et des potentiels oxygène situés entre -470 kJ.mol(O2)-1 et -100 kJ.mol(O2)-1. Les échantillons ont été caractérisés finement avant et après chaque traitement à l’aide de techniques complémentaires comme la microscopie optique et électronique, la microsonde et le SIMS dans le cas de l’irradié. Des mesures de XANES sur synchrotron ont été réalisées sur SIMFuel et ont conduit à des résultats importants en termes de spéciation des PF. Enfin, la technique de Spark Plasma Sintering (SPS) a été explorée avec succès pour la fabrication de SIMFuel contenant du Cs, du Mo et du Ba sous des formes chimiques représentatives d’un combustible REP en fonctionnement nominal.Ce travail a permis de mettre en évidence l’effet de la température en conditions oxydantes sur le comportement du combustible et des PF. Une oxydation du Mo, initialement présent sous forme métallique dans les inclusions blanches du combustible, en MoO2 a été observée dès 1000°C en conditions oxydantes. Une interaction entre le MoO2 formé et le Ba contenu dans la phase oxyde a eu lieu dans les mêmes conditions, menant à la formation de BaMoO4. Le potentiel oxygène joue aussi un rôle important dans le phénomène d’interaction pastille-gaine, en favorisant la diffusion des espèces en conditions oxydantes, diminuant ainsi la température de fusion du combustible. / In the nuclear community, it is a top priority to gain in-depth understanding of fission product (FP) speciation mechanisms occurring in nuclear fuel in order to precisely estimate the source term of a severe accident. Among the FP produced, some are highly reactive and may have a strong radiological impact if released into the environment. This is particularly the case of cesium (Cs), molybdenum (Mo) and barium (Ba). In this context, the objective of this study is to provide experimental data on the effect of the oxygen potential on Cs, Mo and Ba speciation in nuclear fuels at different stages of a severe accident.A thermodynamic approach was coupled with the experimental work to support the interpretation of experimental data. Two types of samples were studied in detail: irradiated MOX fuels and simulated high burn-up UO2 fuels produced through sintering at high temperature (SIMFuel). The samples were submitted to thermal treatments in conditions representative of a pressurised water reactor (PWR) severe accident. This approach made it possible to cover a temperature range from 400°C up to 2530°C and oxygen potentials from -470 kJ.mol(O2)-1 to -100 kJ.mol(O2)-1. The samples were characterized before and after each test using complementary techniques like OM, SEM, EPMA and SIMS in the case of irradiated fuels. XANES measurements using synchrotron radiation facilities were performed on SIMFuels and provided valuable results on FP speciation. Moreover, spark plasma sintering (SPS) was successfully investigated for the production of SIMFuel samples containing Cs, Mo and Ba in a chemical state representative of PWR fuel in normal operating conditions.This work highlighted the effect of oxidizing severe accident conditions on the fuel and FP behavior. Oxidation of Mo initially contained in the fuel’s metallic inclusions into MoO2 was observed to take place around 1000°C in oxidizing conditions. An interaction between MoO2 and the oxide phase containing Ba took place in the same conditions, leading to the formation of BaMoO4. The oxygen potential also plays an important role in fuel-cladding interactions, enhancing the diffusion of species in oxidizing conditions and lowering the temperature at which fuel melting occurs.

Dans ce manuscrit, un travail multidisciplinaire, de la synthèse de précurseurs métalliques à la caractérisation des matériaux est présenté. La première porte sur l’élaboration de poudres d’oxyde TiO2 dopé Nb5+ et d’oxydes mixtes SnO2-TiO2 dopé Nb5+ par une approche Sol-Gel. Leur mise en forme sous forme de pastille par une méthode moderne de frittage flash (SPS) a permis d’étudier leurs propriétés physiques vibrationnelles et thermoélectriques. La deuxième partie de cette étude présente l’élaboration de nouveaux précurseurs d’étain et de tantale adaptés au procédé de dépôt de couches minces par DLI-MOCVD. Afin de répondre aux exigences de ce procédé, des dérivés moléculaires à base de ligand beta-aminoalcool fluoré ou pas ont été élaborées. Les complexes métalliques ont ensuite été entièrement caractérisés à l’état solide et en solution. Le bon comportement thermique (stabilité et volatilité) de certains de ces composés ont conduit à l’élaboration et à la caractérisation de couches minces de SnO2 et SnO2:F / In this manuscript, a multidisciplinary work, from the synthesis of metal precursors to the characterization of materials, is presented. The first concerns the development of Nb5+ doped TiO2 oxide powders and Nb5+ doped SnO2-TiO2 mixed oxides by a Sol-Gel approach. Their shaping into a pellet form by a modern flash sintering method (SPS) made it possible to study their physical, vibrational and thermoelectric properties. The second part of this study presents the development of new tin and tantalum precursors adapted to the DLI-MOCVD thin film deposition process. In order to meet the requirements of this process, molecular derivatives based on fluorinated or non-fluorinated beta-aminoalcohol ligand have been developed. The metal complexes were then fully characterized in solid state and in solution. The good thermal behaviour (stability and volatility) of some of these compounds has led to the development and characterization of thin layers of SnO2 and SnO2:F

In dieser Dissertation wird die Darstellung tetrelreicher Seltenerdmetall–Verbindungen in den Systemen SE:Tt (SE = La, Nd, Sm, Gd, Tb, Ho, Lu; Tt = Si, Ge) und die Charakterisierung ihrer Eigenschaften beschrieben. Diese Verbindungen, welche mittels der Hochdruck–Hochtemperatur–Methode dargestellt wurden, zeigen neuartige Verknüpfungsmuster in der Tetrel–Partialstruktur. Neben der Charakterisierung der Verbindungen hinsichtlich der thermischen Stabilität und der physikalischen Eigenschaften bei Normaldruck wurde bei den Germanium–reichen Seltenerdmetall–Verbindungen eine Untersuchung der Veränderungen der Kristallstruktur bei Variation des Drucks oder der Temperatur vorgenommen. Die dargestellten Seltenerdmetall–Trisilicide SESi3 (SE = Gd, Ho, Lu) kristallisieren tetragonal isotyp zu YbSi3. LuSi3 zeigt Supraleitung mit Tc = 7.0 K. In den Systemen Gd–Si und SE–Ge (SE = La, Nd, Sm, Gd, Tb) wurden die Verbindungen GdSi5 und SEGe5 (SE = La, Nd, Sm, Gd, Tb) synthetisiert. Sie kristallisieren orthorhombisch isotyp zu LaGe5. Durch in–situ Röntgenbeugungsexperimente bei erhöhten Temperaturen kann die Existenz metastabiler Germanium–ärmerer Verbindungen SE2Ge9 (SE = Nd, Sm) nachgewiesen werden. Es handelt sich um Defektvarianten der Verbindungen SEGe5 (SE = Nd, Sm). Die strukturelle Verwandtschaft zum Aristotyp SEGe5 wird anhand einer Gruppe–Untergruppe–Beziehung aufgezeigt. Eine alternative Synthesemethode zur Darstellung tetrelreicher Verbindungen mit Gerüststrukturen bei extremen Reaktionsbedingungen stellt das Spark–Plasma–Sinterverfahren (SPS) dar. Die Darstellung der Clathratphase Ba8Ga16±xGe30∓x (x = 0, 1) erfolgte mittels SPS aus den Precursoren BaGa2±x (x = 0, 0.125) und Germanium. Die Untersuchungen der thermoelektrischen Eigenschaften zeigen, dass durch Variation der nominellen Zusammensetzung sowohl n–leitende als auch p–leitende Eigenschaften erhalten werden können. / In this thesis the synthesis of tetrel–rich rare–earth metal compounds in the systems RE:Tt (RE = La, Nd, Sm, Gd, Tb, Ho, Lu; Tt = Si, Ge) and the characterization of their properties is described. These compounds, synthesized by means of high–pressure high–temperature method, show new structural motifs in the tetrel partial structure. The compounds were characterized with respect to their thermal stability and their physical properties at ambient pressure. In addition, the changes in the crystal structure of the germanium-rich rare–earth metal compounds by variation of pressure or temperature were investigated. The synthesized rare–earth trisilicides SESi3 (SE = Gd, Ho, Lu) crystallize tetragonal, isotypic to YbSi3. LuSi3 is a superconductor with Tc = 7.0 K. In the systems Gd–Si and SE–Ge (SE = La, Nd, Sm, Gd, Tb) the compounds GdSi5 and SEGe5 (SE = La, Nd, Sm, Gd, Tb) were synthesized. They crystallize orthorhombic isotypic to LaGe5. Using in-situ high–temperature X-ray experiments the metastable germanium-poorer compounds SE2Ge9 (SE = Nd, Sm) could be observed. These compounds are defect variants of the pentagermanides SEGe5 (SE = Nd, Sm). The structural relationship to the aristotype SEGe5 is given via a group-subgroup relation. An alternative synthesis route for the preparation of tetrel–rich compounds with framework structures at extreme reaction conditions is the spark plasma sintering method (SPS). The clathrate phase Ba8Ga16±xGe30∓x (x = 0, 1) was synthesized from the precursors BaGa2±x (x = 0, 0.125) and germanium by means of SPS. The investigation of the thermoelectric properties shows, that through variation of the nominal composition both n-type and p-type conduction properties can be obtained.

我國全民健保近年財務不佳，民國98年收支短絀達582億元。根據中央健康保險署資料，截至目前為止，特約醫事服務機構違規次數累積達13722次。在所有重大違規事件中，大部分是詐欺行為。 健保審查機制主要以電腦隨機抽樣，再由人工進行調查。然而，這樣的審查方式無法有效抽取到違規醫事機構之樣本，造成審查效果不彰。 Benford’s Law又稱第一位數法則，其概念為第一位數的值越小則該數字出現的頻率越大，反之相反。該方法被應用於會計、金融、審計及經濟領域中。楊喻翔(2012)將Benford’s Law相關指標應用於我國全民健保上，並結合機器學習演算法來進行健保異常偵測。 Zaharia et al. (2012)提出了一種具容錯的群集記憶內運算模式 Apache Spark，在相同的運算節點及資源下，其資料運算效率及速度可勝出Hadoop MapReduce 20倍以上。 為解決健保異常查核效果不彰問題，本研究將採用Benford’s Law，使用國家衛生研究院發行之健保資料計算成為Benford’s Law指標和實務指標，接著並使用支援向量機和邏輯斯迴歸來建構出異常查核模型。然而健保資料量龐大，為加快運算時間，本研究使用Apache Spark做為運算環境，並以Hadoop MapReduce作為標竿，比較運算效率。 研究結果顯示，本研究撰寫的Spark程式運算時間能較MapReduce快2倍；在分類模型上，支援向量機和邏輯斯迴歸所進行的住院資料測試，敏感度皆有80%以上；而所進行的門診資料測試，兩個模型的準確率沒有住院資料高，但邏輯斯迴歸測試結果仍保有一定的準確性，在敏感度仍有75%，整體正確率有73%。 本研究使用Apache Spark節省處理大量健保資料的運算時間。其次本研究建立的智慧型異常查核模型，確實能查核出違約的醫事機構，而模型所查核出可能有詐欺及濫用健保之醫事機構，可進行下階段人工調查，最終得改善健保查核效力。 / Financial condition of National Health Insurance (NHI) has been wretched in recent years. The income statement in 2009 indicated that National Health Insurance Administration (NHIA) was in debt for NTD $58.2 billion. According to NHIA data, certain medical institutions in Taiwan violated the NHI laws for 13722 times. Among all illegal cases, fraud is the most serious. In order to find illegal medical institutions, NHIA conducted random sampling by computer. Once the data was collected, NHIA investigators got involved in the review process. However, the way to get the samples mentioned above cannot reveal the reality. Benford's law is called the First-Digit Law. The concept of Benford’s Law is that the smaller digits would appear more frequently, while larger digits would occur less frequently. Benford’s Law is applied to accounting, finance, auditing and economics. Yang(2012) used Benford’s Law in NHI data and he also used machine learning algorithms to do fraud detection. Zaharia et al. (2012) proposed a fault-tolerant in-memory cluster computing -Apache Spark. Under the same computing nodes and resources, Apache Spark’s computing is faster than Hadoop MapReduce 20 times. In order to solve the problem of medical claims review, Benford’s Law was applied to this study. This study used NHI data which was published by National Health Research Institutes. Then, we computed NHI data to generate Benford’s Law variables and technical variables. Finally, we used support vector machine and logistics regression to construct the illegal check model. During system development, we found that the data size was big. With the purpose of reducing the computing time, we used Apache Spark to build computing environment. Furthermore, we adopted Hadoop MapReduce as benchmark to compare the performance of computing time. This study indicated that Apache Spark is faster twice than Hadoop MapReduce. In illegal check model, with support vector machine and logistics regression, we had 80% sensitivity in inpatient data. In outpatient data, the accuracy of support vector machine and logistics regression were lower than inpatient data. In this case, logistics regression still had 75% sensitivity and 73% accuracy. This study used Apache Spark to compute NHI data with lower computing time. Second, we constructed the intelligent illegal check model which can find the illegal medical institutions for manual check. With the use of illegal check model, the procedure of medical claims review will be improved.