No description available.

Au cours des dernières années, le cycles organique de Rankine (ORC) ont reçu un grand intérêt de la communauté scientifique et technique en raison de sa capacité à récupérer de l'énergie à partir de sources de chaleur faible. Dans certaines applications, comme la récupération de chaleur des déchets (WHR), les plantes ORC doivent être aussi le plus compact possible en raison de contraintes géométriques et de poids. Récemment, ces questions ont été étudiées dans le but de promouvoir la technologie ORC pour moteur à combustion interne (ICE). L'idée de récupérer ce résidu d'énergie est pas nouvelle et dans les années 1970 la crise énergétique a encouragé le développement de petite ORC plants (1-10 kWe). En raison de la complexité moléculaire du fluides de travail , fort effets de gaz réel doivent être pris en compte en raison de la haute pression et la densité, si on le compare à un gaz idéal. Dans ces conditions, le fluide est connu comme gaz dense. Les gaz denses sont définis comme des vapeurs monophasés, caractérisé par des molécules complexes et avec importantes masses moléculaires. Le rôle de gaz dense dans la gaz dynamique des flux transsonique interne a été largement étudié pour son importance dans les turbomachines. Récemment, l'attention a été concentrée sur des turbines axiales, qui réduisent au minimum la taille du système, en comparaison avec les solutions radiales dans les mêmes rapports de pression et la chute d'enthalpie. Dans ce travail, une nouvelle méthodologie de conception de turbines ORC supersonique est proposé. Elle consiste dans un design à deux dimensions rapide et précise qui est réalisée pour stator et rotor avec une metode de caractéristique (MOC) étendue à une équation d'etat générique. Les effets visqueux sont pris en compte par l'introduction d'une correction turbulente appropriée de la couche limite compressible. Étant donné que les sources de chaleur proposées pour turbines ORC comprennent typiquement des sources d'énergie variables, comme la WHR des procédés industriels ou des applications automobiles, pour améliorer la faisabilité de cette technique, la résistance à des conditions variables d'entrée est prise en compte. L'optimisation numérique sous incertitudes est appelé Optimisation robuste (RO) et il surmonte la limitation de l'optimisation déterministe qui néglige l'effet des incertitudes dans les variables de design et / ou des paramètres de design. Pour mesurer la robustesse d'un nouveau design, les statistiques (la moyenne et la variance, ou écart-type) d'une réponse sont calculées dans le processus RO. Dans ce travail, la conception MOC des ORC aubes supersoniques est utilisé pour créer une profil de référence. Cela est optimisé grâce à une boucle RO. L'optimiseur stochastique est basée sur un modèle de krigeage bayésien de la réponse du système aux paramètres incertains, utilisé pour l'approximation des statistiques de la sortie du système, couplé à une algorithme genetique multi-objectif (NSGA). Une forme optimale qui maximise la moyenne et minimise la variance de l'efficacité isentropique est recherché. L'efficacité isentropique est évaluée au moyen de simulations RANS (Reynolds Average Navier-Stokes) de l'aube. Le comportement thermodynamique du fluide de travail est modélisée au moyen de l'équation d'etat de Peng-Robinson-stryjek-Vera. La forme de l'aube est paramétrée au moyen d'une approche Free Form Deformation. Pour accélérer le RO processus, une modèle de krigeage supplémentaire est construit pour la fonction multi-objectifs et une stratégie adaptif de remplissage basée sur le Multi Objective Expected Improvement es prise en compte afin d'améliorer la précision de krigeage à chaque génération de la NSGA. La forme robuste optimisé d'aube ORC est comparé aux résultats fournis par le MOC et l'optimiseur déterministe. / In recent years, the Organic Rankine Cycle (ORC) technology has received great interest from the scientific and technical community because of its capability to recover energy from low-grade heat sources. In some applications, as the Waste Heat Recovery (WHR), ORC plants need to be as compact as possible because of geometrical and weight constraints. Recently, these issues have been studied in order to promote the ORC technology for Internal Combustion Engine (ICE) applications. The idea to recover this residual energy is not new and the 1970s energy crisis encouraged the development of feasible ORC small-scale plants (1-10 kWe). Due to the molecular complexity of the working fluids, strong real gas effects have to be taken into account because of the high pressures and densities, if compared to an ideal gas. In these conditions the fluid is known as dense gas. Dense gases are defined as single phase vapors, characterized by complex molecules and moderate to large molecular weights. The role of dense gas dynamics in transonic internal flows has been widely studied for its importance in turbomachinery applications involved in low-grade energy exploitation, such as the ORC. Recently, the attention has been focused on axial turbines, which minimize the system size, if compared with radial solutions at the same pressure ratios and enthalpy drops. In this work, a novel design methodology for supersonic ORC axial impulse turbine stages is proposed. It consists in a fast, accurate two-dimensional design which is carried out for the mean-line stator and rotor blade rows of a turbine stage by means of a method of characteristic (MOC) extended to a generic equation of state. The viscous effects are taken into account by introducing a proper turbulent compressible boundary layer correction to the inviscid design obtained with MOC. Since proposed heat sources for ORC turbines typically include variable energy sources such as WHR from industrial processes or automotive applications, as a result, to improve the feasibility of this technology, the resistance to variable input conditions is taken into account. The numerical optimization under uncertainties is called Robust Optimization (RO) and it overcomes the limitation of deterministic optimization that neglects the effect of uncertainties in design variables and/or design parameters. To measure the robustness of a new design, statistics such as mean and variance (or standard deviation) of a response are calculated in the RO process. In this work, the MOC design of supersonic ORC nozzle blade vanes is used to create a baseline injector shape. Subsequently, this is optimized through a RO loop. The stochastic optimizer is based on a Bayesian Kriging model of the system response to the uncertain parameters, used to approximate statistics of the uncertain system output, coupled to a multi-objective non-dominated sorting genetic algorithm (NSGA). An optimal shape that maximizes the mean and minimizes the variance of the expander isentropic efficiency is searched. The isentropic efficiency is evaluated by means of RANS (Reynolds Average Navier-Stokes) simulations of the injector. The fluid thermodynamic behavior is modelled by means of the well-known Peng-Robinson-Stryjek-Vera equation of state. The blade shape is parametrized by means of a Free Form Deformation approach. In order to speed-up the RO process, an additional Kriging model is built to approximate the multi-objective fitness function and an adaptive infill strategy based on the Multi Objective Expected Improvement for the individuals is proposed in order to improve the surrogate accuracy at each generation of the NSGA. The robustly optimized ORC expander shape is compared to the results provided by the MOC baseline shape and the injector designed by means of a standard deterministic optimizer.

The use of hazardous toxic substances is very common in the industrial sector. The substances are often stored in tanks in storage compartments or transported between industrial premises. In case of an accident involving these substances, severe harm can affect both population and the environment. This leaves a demand for an accurate prediction of the substance concentration distribution to mitigate the risks as much as possible and in advance create suitable safety measures. Toxic gases and vapors are often denser than air making it affected by negative buoyancy forces. This will make the gas descend and spread horizontally when reaching the ground. Swedish Defence Research Agency (FOI) carries today a model called LillPello for simulating the dispersion of gases, yet it does not account for the specific case of a dense gas. Therefore, this thesis aims to implement the necessary effects needed to accurately simulate the dispersion of a dense gas. These effects were implemented in Fortran 90 by solving five conservation equations for energy, momentum (vertical and horizontal) and mass. The model was compared against experimental data of a leak of ammonia (NH3). By analyzing the result of the simulations in this thesis, we can conclude that the overall result is satisfactory. We can notice a small concentration underestimation at all measurement points and the model produced a concentration power law coefficient which lands inside the expected range. Two out of the three statistical quantities Geometric Mean (MG), Geometric Variance (VG) and Factor of 2 (FA2) produced values within the ranges of acceptable values. The drawback of the model as it is implemented today is its efficiency, so the main priority for the future of this thesis is to improve this. The model should also be analyzed on more experiments to further validate its accuracy. / Användandet av giftiga ämnen är vanligt inom den industriella sektorn. Ämnena är oftast lagrade i behållare positionerade i lagringsutrymmen eller så transporteras ämnena mellan industrilokaler. I samband med en olycka innehållande dessa substanser kan stora skador drabba både befolkning och miljön. Detta leder till ett behov av att noggrant kunna förutspå koncentrationsfördelningen för att minska riskerna, samt i förväg kunna skapa lämpliga säkerhetsåtgärder. Giftiga gaser och ångor är oftast tyngre än luft vilket gör att gasen blir påverkad av negativ bärkraft. Detta gör att gasen sjunker och sprids horisontalt när den når marken. Totalförsvarets Forskningsinstitut (FOI) besitter idag en modell kallad LillPello som simulerar spridning av gaser, men den hanterar inte det specifika fallet av en tunggas. Därför siktar detta projekt på att, in i LillPello, implementera de nödvändiga effekterna som behövs för att korrekt kunna simulera spridningen av en tunggas. Dessa effekter är implementerad i Fortran 90 genom att lösa fem konserveringsekvationer för energi, momentum (vertikal och horisontell) samt massa. Modellen jämfördes mot data från ett fältexperiment där ammoniak (NH3) släpptes ut. Genom att analysera resultatet från simuleringar kan vi dra slutsatsen att det övergripande resultatet är tillfredsställande. Vi kan notera en underskattning för alla koncentrationsmätningar i simuleringarna och modellen producerade en potenslagsexponent vars värde hamnade innanför den accepterade gränsen. Två utav de tre beräknade statistiska kvantiteterna: Geometriskt medelvärde (MG), Geometrisk varians (VG) och Faktor av 2 (FA2) producerade värden inom de acceptabla gränserna. Största nackdelen med modellen är dess effektivitet och därför är största prioritet för det fortsatta arbetet inom detta projekt att effektivisera implementeringen. Modellen ska även bli vidare analyserad mot fler experiment för att validera dess noggrannhet.

The applications of dense gas technology (DGT) in modern particle engineering have shown promising results in producing submicron particles with uniform particle morphology. In this study, two configurations of dense gas antisolvent processes were employed for the micronization, encapsulation and co-precipitation of pharmaceutical compounds. The encapsulation of superparamagnetic iron oxide nanoparticles (SPIONs) by a pH-responsive polymer (Eudragit?? S100) was successfully performed using the supercritical antisolvent (SAS) process. Nanocomposites of less than 200nm in diameter with encapsulated SPIONs content as high as 16 wt% were achieved. Magnetic characterization of the product was also performed and the data were fitted by the Langevin equation. The superparamagnetic properties of the composites were preserved and the effective magnetic size was about 10 nm. The magnetically and pH-responsive nanocomposites can be potentially utilized as magnetic resonance imaging contrast agents and drug carriers. Screening experiments of 8 active pharmaceutical ingredients and 5 pharmaceutical excipients were performed using the recently patented atomized rapid injection solvent extraction (ARISE) process. Candidates with promising product morphology and recovery were selected for co-precipitation studies. The co-precipitation of the anti-cancer drug 5-fluorouracil (5FU) and poly l-lactic acid (PLLA) was conducted to develop a controlled release system. Experiments were designed based on a two-level, three-factor factorial design, in order to investigate the effects of processing parameters on product characteristics. Submicron PLLA-5FU composites (diameter<0.8 ??m) with a drug loading of 7.4 wt% were produced.

No description available.

The use of dense CO2 in polymer processing can provide a response to the need for more environmentally-friendly industrial processes. Products with high-purity, sterility, and porosity can be achieved using dense gas technology (DGT). Currently, DGT has been used in different aspects of polymer processing including polymerization, micronization, and impregnation. Due to its solubility in polymers, CO2 can penetrate and plasticize polymers, while impregnating them with low-molecular weight CO2 -soluble compounds. Biodegradable polymers and other medical-grade polymers have benefited from the application of DGT. Dense CO2 processing properties of inertness, non-toxicity, and affinity for various therapeutic compounds are specifically advantageous to the medical and biomedical industries. In this work, the different applications of DGT in polymer processing are revised, then implemented. The polymerization of polycarbonate (PC) and polycaprolactone (PCL) in dense CO2 are presented. The syntheses of both polymers were successful and were aided by the use of dense CO2 . A multi-stage approach using dense CO2 as a sweep fluid to extract the PC polymerization by-product phenol is reported. Polycaprolactone was synthesized with varying temperatures and dense CO pressures, then impregnated with a CO2 -soluble therapeutic agent. The impregnated PCL acted as a drug reservoir with a drug-loading of 27wt% and a sustained drug release profile was observed for all samples over several days. Polymer blends of PC/PCL have potential industrial and biomedical applications both in vivo and in vitro. The applicability of PCL can be extended by enhancing its mechanical properties by creating a bio-blend with a stronger polymer such as PC. In this work, PC/PCL nonporous and porous blends were produced. Three novel dense CO2 blending techniques were used. The macroporous PC/PCL blend was impregnated with a therapeutic agent using CO2 as the carrier. A drug loading of 20wt% was achieved and sustained drug release was observed over 3 days. The applicability of dense CO2 in polymer processing was further demonstrated by sterilizing macroporous PC/PCL blends and soft hydrogels with dense CO2 . The PC/PCL blends and hydrogels were inoculated with vegetative bacteria and bacterial endospores. Industrial standard sterilization levels were achieved.

A percepção crescente da sociedade em relação aos perigos inerentes às instalações industriais que manipulam grandes inventários de substâncias perigosas faz com que a ferramenta análise quantitativa de risco ganhe importância na complexa discussão sobre a viabilidade destes empreendimentos, no intuito de promover a ocupação adequada do solo na área urbana e prevenir a ocorrência do chamado acidente maior. Contudo, para se chegar à expressão de risco de uma determinada instalação industrial deve-se aplicar um conjunto de técnicas e de modelos matemáticos, entre os quais estão os modelos de dispersão atmosférica, usados para se estimar a área afetada na vizinhança da mesma por liberações acidentais que levam à formação de nuvens de substâncias químicas na atmosfera. Em decorrência da complexidade inerente ao próprio processo de dispersão atmosférica, especialmente no que tange aos denominados gases densos, existe uma diversidade de modelos que podem ser aplicados no escopo da análise de risco, o que leva a seus usuários, naturalmente, ao questionamento sobre a suscetibilidade dos resultados finais ao tipo de modelagem adotada. Neste sentido, este trabalho estuda o processo de dispersão atmosférica de nuvens densas formadas em liberações acidentais, identificando as principais possibilidades de modelagem deste processo e, ao final, apresenta um estudo de caso demonstrando que diferentes modelagens desta dispersão, comumente empregadas em análise de risco de instalações industriais, podem produzir variações na estimativa do risco de uma mesma instalação e, portanto, influenciar as decisões baseadas em risco. / The concern of the society about the risks posed by activities that deal with hazardous substances has increased in an environment strongly industrialized and with high population density in view of the inherent potential hazards of them as well as the impact of recent accidental episodes, even though their benefits provided. In this context the quantitative risk analysis is presented as an essential tool to assess the risk of these activities and compose a complex discussion about its feasibility. Some of these accident scenarios may involve the formation of a hazardous product cloud and its subsequent air dispersion in the off-site region when an accidental released take place and one should apply the so-called atmospheric dispersion models for estimating the consequences of the releases. Due to the complexity involved in this atmospheric dispersion process, there is a wide variety of mathematical models that can be applied for estimating the offsite consequences of the accidental releases leading, naturally, to one wonder whether the final risk expression of a facility is susceptible to these differences. Often in the world of industrial use of hazardous materials, toxic or flammable there is a possibility that these accidental releases produce clouds that are denser than air, a situation that demands even more attention in terms of risk aspects involved. Then, this dissertation studies the process of atmospheric dispersion of heavier-than-air clouds produced after an accidental release, identifying the main ways of modelling the process and presents a case study comparing different dispersion models that demonstrates that the final expression of risk of a typical installation can be different when it is used different dispersion model in the process.

Com o atual desenvolvimento industrial e tecnológico da sociedade, a presença de substâncias inflamáveis e/ou tóxicas aumentou significativamente em um grande número de atividades. A possível dispersão de gases perigosos em instalações de armazenamento ou em operações de transporte representam uma grande ameaça à saúde e ao meio ambiente. Portanto, a caracterização de uma nuvem inflamável e/ou tóxica é um ponto crítico na análise quantitativa de riscos. O objetivo principal desta tese foi fornecer novas perspectivas que pudessem auxiliar analistas de risco envolvidos na análise de dispersões em cenários complexos, por exemplo, cenários com barreiras ou semi-confinados. A revisão bibliográfica mostrou que, tradicionalmente, modelos empíricos e integrais são usados na análise de dispersão de substâncias tóxicas / inflamáveis, fornecendo estimativas rápidas e geralmente confiáveis ao descrever cenários simples (por exemplo, dispersão em ambientes sem obstruções sobre terreno plano). No entanto, recentemente, o uso de ferramentas de CFD para simular dispersões aumentou de forma significativa. Estas ferramentas permitem modelar cenários mais complexos, como os que ocorrem em espaços semi-confinados ou com a presença de barreiras físicas. Entre todas as ferramentas CFD disponíveis, consta na bibliografia que o software FLACS® tem bom desempenho na simulação destes cenários. Porém, como outras ferramentas similares, ainda precisa ser totalmente validado. Após a revisão bibliográfica sobre testes de campo já executados ao longo dos anos, alguns testes foram selecionados para realização de um exame preliminar de desempenho da ferramenta CFD utilizado neste estudo. Foram investigadas as possíveis fontes de incertezas em termos de capacidade de reprodutibilidade, de dependência de malha e análise de sensibilidade das variáveis de entrada e parâmetros de simulação. Os principais resultados desta fase foram moldados como princípios práticos a serem utilizados por analistas de risco ao realizar análise de dispersão com a presença de barreiras utilizando ferramentas CFD. Embora a revisão bibliográfica tenha mostrado alguns dados experimentais disponíveis na literatura, nenhuma das fontes encontradas incluem estudos detalhados sobre como realizar simulações de CFD precisas nem fornecem indicadores precisos de desempenho. Portanto, novos testes de campo foram realizados a fim de oferecer novos dados para estudos de validação mais abrangentes. Testes de campo de dispersão de nuvem de propano (com e sem a presença de barreiras obstruindo o fluxo) foram realizados no campo de treinamento da empresa Can Padró Segurança e Proteção (em Barcelona). Quatro testes foram realizados, consistindo em liberações de propano com vazões de até 0,5 kg/s, com duração de 40 segundos em uma área de descarga de 700 m2. Os testes de campo contribuíram para a reavaliação dos pontos críticos mapeados durante as primeiras fases deste estudo e forneceram dados experimentais para serem utilizados pela comunidade internacional no estudo de dispersão e validação de modelos. Simulações feitas utilizando-se a ferramenta CFD foram comparadas com os dados experimentais obtidos nos testes de campo. Em termos gerais, o simulador mostrou bom desempenho em relação às taxas de concentração da nuvem. O simulador reproduziu com sucesso a geometria complexa e seus efeitos sobre a dispersão da nuvem, mostrando claramente o efeito da barreira na distribuição das concentrações. No entanto, as simulações não foram capazes de representar toda a dinâmica da dispersão no que concerne aos efeitos da variação do vento, uma vez que as nuvens simuladas diluíram mais rapidamente do que nuvens experimentais. / With the industrial and technological development of the present-day society, the presence of flammable and toxic substances has increased in a growing number of activities. Dispersion of hazardous gas releases occurring in transportation or storage installations represent a major threat to health and environment. Therefore, forecasting the behaviour of a flammable or toxic cloud is a critical challenge in quantitative risk analysis. The main aim of this dissertation has been to provide new insights that can help technological risks analysts when dealing with complex dispersion modelling problems, particularly those problems involving dispersion scenarios with barriers or semi-confined. A literature survey has shown that, traditionally, empirical and integral models have been used to analyse dispersion of toxic/flammable substances, providing fast estimations and usually reliable results when describing simple scenarios (e.g. unobstructed gas flows over flat terrain). In recent years, however, the use of CFD tools for simulating dispersion accidents has significantly increased, as they allow modelling more complicated gas dispersion scenarios, like those occurring in complex topographies, semi-confined spaces or with the presence of physical barriers. Among all the available CFD tools, FLACS® software is envisaged to have high performance when simulating dispersion scenarios, but, as other codes alike, still needs to be fully validated. This work contributes to the validation of FLACS software for dispersion analysis. After a literature review on historical field tests, some of them have been selected to undertake a preliminary FLACS performance examination, inspecting all possible sources of uncertainties in terms of reproducibility capacity, grid dependence and sensitivity analysis of input variables and simulation parameters. The main outcomes of preliminary FLACS investigations have been shaped as practical guiding principles to be used by risk analysts when performing dispersion analysis with the presence of barriers using CFD tools. Although the literature survey has shown some experimental data available, none of the works include detailed exercises giving new insights of how to perform accurate CFD simulations nor giving precise rates of FLACS performance. Therefore, new experiments have been performed in order to offer new sets of cloud dispersion data for comprehensive validation studies. Propane cloud dispersion field tests (unobstructed and with the presence of a fence obstructing the flow) have been designed and undertaken at Can Padró Security and Safety training site (Barcelona) by which intensive data on concentration has been acquired. Four tests were performed, consisting on releases up to 0.5 kg/s of propane during 40 seconds in a discharge area of 700 m2. The field tests have contributed to the reassessment of the critical points raised in the guiding principles and have provided experimental data to be used by the international community for dispersion studies and models validation exercises. FLACS software has been challenged against the experimental data collected during the field tests. In general terms, the CFD-based simulator has shown good performance when simulating cloud concentration. FLACS reproduces successfully the presence of complex geometry and its effects on cloud dispersion, showing realistic concentration decreases due to cloud dispersion obstruction by the existence of a fence. However, simulated clouds have not represented the whole complex accumulation dynamics due to wind variation, since they have diluted faster than experimental clouds.

A percepção crescente da sociedade em relação aos perigos inerentes às instalações industriais que manipulam grandes inventários de substâncias perigosas faz com que a ferramenta análise quantitativa de risco ganhe importância na complexa discussão sobre a viabilidade destes empreendimentos, no intuito de promover a ocupação adequada do solo na área urbana e prevenir a ocorrência do chamado acidente maior. Contudo, para se chegar à expressão de risco de uma determinada instalação industrial deve-se aplicar um conjunto de técnicas e de modelos matemáticos, entre os quais estão os modelos de dispersão atmosférica, usados para se estimar a área afetada na vizinhança da mesma por liberações acidentais que levam à formação de nuvens de substâncias químicas na atmosfera. Em decorrência da complexidade inerente ao próprio processo de dispersão atmosférica, especialmente no que tange aos denominados gases densos, existe uma diversidade de modelos que podem ser aplicados no escopo da análise de risco, o que leva a seus usuários, naturalmente, ao questionamento sobre a suscetibilidade dos resultados finais ao tipo de modelagem adotada. Neste sentido, este trabalho estuda o processo de dispersão atmosférica de nuvens densas formadas em liberações acidentais, identificando as principais possibilidades de modelagem deste processo e, ao final, apresenta um estudo de caso demonstrando que diferentes modelagens desta dispersão, comumente empregadas em análise de risco de instalações industriais, podem produzir variações na estimativa do risco de uma mesma instalação e, portanto, influenciar as decisões baseadas em risco. / The concern of the society about the risks posed by activities that deal with hazardous substances has increased in an environment strongly industrialized and with high population density in view of the inherent potential hazards of them as well as the impact of recent accidental episodes, even though their benefits provided. In this context the quantitative risk analysis is presented as an essential tool to assess the risk of these activities and compose a complex discussion about its feasibility. Some of these accident scenarios may involve the formation of a hazardous product cloud and its subsequent air dispersion in the off-site region when an accidental released take place and one should apply the so-called atmospheric dispersion models for estimating the consequences of the releases. Due to the complexity involved in this atmospheric dispersion process, there is a wide variety of mathematical models that can be applied for estimating the offsite consequences of the accidental releases leading, naturally, to one wonder whether the final risk expression of a facility is susceptible to these differences. Often in the world of industrial use of hazardous materials, toxic or flammable there is a possibility that these accidental releases produce clouds that are denser than air, a situation that demands even more attention in terms of risk aspects involved. Then, this dissertation studies the process of atmospheric dispersion of heavier-than-air clouds produced after an accidental release, identifying the main ways of modelling the process and presents a case study comparing different dispersion models that demonstrates that the final expression of risk of a typical installation can be different when it is used different dispersion model in the process.

Com o atual desenvolvimento industrial e tecnológico da sociedade, a presença de substâncias inflamáveis e/ou tóxicas aumentou significativamente em um grande número de atividades. A possível dispersão de gases perigosos em instalações de armazenamento ou em operações de transporte representam uma grande ameaça à saúde e ao meio ambiente. Portanto, a caracterização de uma nuvem inflamável e/ou tóxica é um ponto crítico na análise quantitativa de riscos. O objetivo principal desta tese foi fornecer novas perspectivas que pudessem auxiliar analistas de risco envolvidos na análise de dispersões em cenários complexos, por exemplo, cenários com barreiras ou semi-confinados. A revisão bibliográfica mostrou que, tradicionalmente, modelos empíricos e integrais são usados na análise de dispersão de substâncias tóxicas / inflamáveis, fornecendo estimativas rápidas e geralmente confiáveis ao descrever cenários simples (por exemplo, dispersão em ambientes sem obstruções sobre terreno plano). No entanto, recentemente, o uso de ferramentas de CFD para simular dispersões aumentou de forma significativa. Estas ferramentas permitem modelar cenários mais complexos, como os que ocorrem em espaços semi-confinados ou com a presença de barreiras físicas. Entre todas as ferramentas CFD disponíveis, consta na bibliografia que o software FLACS® tem bom desempenho na simulação destes cenários. Porém, como outras ferramentas similares, ainda precisa ser totalmente validado. Após a revisão bibliográfica sobre testes de campo já executados ao longo dos anos, alguns testes foram selecionados para realização de um exame preliminar de desempenho da ferramenta CFD utilizado neste estudo. Foram investigadas as possíveis fontes de incertezas em termos de capacidade de reprodutibilidade, de dependência de malha e análise de sensibilidade das variáveis de entrada e parâmetros de simulação. Os principais resultados desta fase foram moldados como princípios práticos a serem utilizados por analistas de risco ao realizar análise de dispersão com a presença de barreiras utilizando ferramentas CFD. Embora a revisão bibliográfica tenha mostrado alguns dados experimentais disponíveis na literatura, nenhuma das fontes encontradas incluem estudos detalhados sobre como realizar simulações de CFD precisas nem fornecem indicadores precisos de desempenho. Portanto, novos testes de campo foram realizados a fim de oferecer novos dados para estudos de validação mais abrangentes. Testes de campo de dispersão de nuvem de propano (com e sem a presença de barreiras obstruindo o fluxo) foram realizados no campo de treinamento da empresa Can Padró Segurança e Proteção (em Barcelona). Quatro testes foram realizados, consistindo em liberações de propano com vazões de até 0,5 kg/s, com duração de 40 segundos em uma área de descarga de 700 m2. Os testes de campo contribuíram para a reavaliação dos pontos críticos mapeados durante as primeiras fases deste estudo e forneceram dados experimentais para serem utilizados pela comunidade internacional no estudo de dispersão e validação de modelos. Simulações feitas utilizando-se a ferramenta CFD foram comparadas com os dados experimentais obtidos nos testes de campo. Em termos gerais, o simulador mostrou bom desempenho em relação às taxas de concentração da nuvem. O simulador reproduziu com sucesso a geometria complexa e seus efeitos sobre a dispersão da nuvem, mostrando claramente o efeito da barreira na distribuição das concentrações. No entanto, as simulações não foram capazes de representar toda a dinâmica da dispersão no que concerne aos efeitos da variação do vento, uma vez que as nuvens simuladas diluíram mais rapidamente do que nuvens experimentais. / With the industrial and technological development of the present-day society, the presence of flammable and toxic substances has increased in a growing number of activities. Dispersion of hazardous gas releases occurring in transportation or storage installations represent a major threat to health and environment. Therefore, forecasting the behaviour of a flammable or toxic cloud is a critical challenge in quantitative risk analysis. The main aim of this dissertation has been to provide new insights that can help technological risks analysts when dealing with complex dispersion modelling problems, particularly those problems involving dispersion scenarios with barriers or semi-confined. A literature survey has shown that, traditionally, empirical and integral models have been used to analyse dispersion of toxic/flammable substances, providing fast estimations and usually reliable results when describing simple scenarios (e.g. unobstructed gas flows over flat terrain). In recent years, however, the use of CFD tools for simulating dispersion accidents has significantly increased, as they allow modelling more complicated gas dispersion scenarios, like those occurring in complex topographies, semi-confined spaces or with the presence of physical barriers. Among all the available CFD tools, FLACS® software is envisaged to have high performance when simulating dispersion scenarios, but, as other codes alike, still needs to be fully validated. This work contributes to the validation of FLACS software for dispersion analysis. After a literature review on historical field tests, some of them have been selected to undertake a preliminary FLACS performance examination, inspecting all possible sources of uncertainties in terms of reproducibility capacity, grid dependence and sensitivity analysis of input variables and simulation parameters. The main outcomes of preliminary FLACS investigations have been shaped as practical guiding principles to be used by risk analysts when performing dispersion analysis with the presence of barriers using CFD tools. Although the literature survey has shown some experimental data available, none of the works include detailed exercises giving new insights of how to perform accurate CFD simulations nor giving precise rates of FLACS performance. Therefore, new experiments have been performed in order to offer new sets of cloud dispersion data for comprehensive validation studies. Propane cloud dispersion field tests (unobstructed and with the presence of a fence obstructing the flow) have been designed and undertaken at Can Padró Security and Safety training site (Barcelona) by which intensive data on concentration has been acquired. Four tests were performed, consisting on releases up to 0.5 kg/s of propane during 40 seconds in a discharge area of 700 m2. The field tests have contributed to the reassessment of the critical points raised in the guiding principles and have provided experimental data to be used by the international community for dispersion studies and models validation exercises. FLACS software has been challenged against the experimental data collected during the field tests. In general terms, the CFD-based simulator has shown good performance when simulating cloud concentration. FLACS reproduces successfully the presence of complex geometry and its effects on cloud dispersion, showing realistic concentration decreases due to cloud dispersion obstruction by the existence of a fence. However, simulated clouds have not represented the whole complex accumulation dynamics due to wind variation, since they have diluted faster than experimental clouds.