Uma das tarefas mais desafiadoras do engenheiro na área da Geotecnia é a escolha dos valores de parâmetros geotécnicos obtidos de ensaios de campo ou laboratório e que serão utilizados nos modelos analíticos ou numéricos na fase de projeto de fundações. Diante das incertezas inerentes aos ensaios de SPT e da heterogeneidade de abordagens para a utilização dos valores de NSPT, é proposta neste estudo, a aplicação de um critério estatístico para obtenção de valores de NSPT, a partir da construção de intervalos de confiança de 95% de probabilidade em torno da reta ajustada de regressão linear simples entre a variável aleatória NSPT e a profundidade. Os valores obtidos de NSPT pelo critério aplicado foram utilizados na previsão da capacidade de carga de 19 estacas isoladas a partir da utilização de três métodos semi-empíricos: Aoki-Velloso (1975) com coeficientes alterados por Monteiro (1997), Décourt & Quaresma (1978) alterado pelo método de Décourt (1996) e Método de Alonso (1996). As cargas de ruptura dessas 19 estacas ensaiadas através de Provas de Carga Estática foram obtidas pelos métodos de extrapolação de Van Der Veen (1953) e Décourt (1996) e serviram para comparação e consequente validação do critério estatístico. Adicionalmente, com fulcro no item 6.2.1.2.1 da ABNT NBR 6122:2010 Resistência calculada por método semi-empírico, foram avaliados os fatores de segurança em relação às cargas de projeto, inclusive, também se utilizando da premissa de reconhecimento de regiões representativas, levando em conta o número de ensaios de SPT executados, fato que promove uma diminuição da incerteza dos parâmetros, apontando a um menor fator de segurança. A dissertação enfatiza as vantagens de um adequado tratamento estatístico dos parâmetros geotécnicos, a exemplo da recomendação já existente nas normas internacionais como Eurocódigo e outras. O critério construído permite e encoraja análises e decisões racionais no universo das partes interessadas consumidores, projetistas, fiscais de obras, contratantes e comunidade científica promovendo as discussões de forma mais objetiva e harmoniosa sobre o tema. / One of the most challenging aspects of geotechnical engineering is the selection of soil parameters from field and / or laboratory tests to be used in analytical or numerical models for foundation design. Due to known uncertainties in SPT tests and wide availability of criteria for NSPT interpretation, a proposed procedure is presented based on 95% confidence limits around a trend line defined by simple linear regression analysis expressing the variation of NSPT with depth. The NSPT values obtained by the proposed approach have been used to estimate the pile ultimate capacity of 19 isolated continuous flight auger piles using different semi-empirical methods, such as Aoki and Velloso (1975) with modified coefficients as proposed by Monteiro (1997), Décourt and Quaresma (1978) modified by Décourt (1996) and Alonso (1996). Static load tests of the same 19 piles have been extrapolated by Van Der Veen (1953) and Décourt (1996) methods, as an aid for comparison and validation of the statistical criterion. Additionally, were made with the fulcrum in item 6.2.1.2.1 of ABNT NBR 6122: 2010 - Resistance calculated by semi-empirical method, evaluations of safety factors in relation to load project, also including the premise of recognizing representative regions and taking into account the number of SPT tests, a fact that provides the decreased uncertainty of the parameters, indicating a lower FS. The dissertation emphasizes the advantages of an adequate statistical treatment of the geotechnical data, similar to what is recommended by the Eurocode. Such approach allows and encourages a more rational decision including all interested parties - consumers, designers, inspectors, contractors and scientific community providing more objective and harmonious discussions on this subject.

L’étude du transfert radiatif dans les gaz (atmosphères planétaires, chambres de combustion, etc.) se heurte à deux principales difficultés : les hétérogénéités et la dépendance spectrale des propriétés radiatives du milieu d’intérêt. Les travaux présentés dans ce manuscrit proposent, au travers d’une approche statistique du rayonnement, une solution à ces deux limites qui ne nécessite aucune approximation de modèle physique ou numérique. Cette approche conduira au développement d’algorithmes de Monte-Carlo considérés à l’heure actuelle comme méthodes de référence dans la communauté du transfert radiatif. La difficulté liée aux hétérogénéités du milieu participant sera traitée par une technique empruntée à d’autres disciplines de la physique du transport : les algorithmes à collisions nulles. Leur application au rayonnement consiste à ajouter arbitrairement aux évènements d’absorption et de diffusion, un troisième type de collision n’ayant aucun effet sur le transport de photons : les collisions nulles. Ainsi, le coefficient d’extinction résultant de ces trois types de collision pourra être assumé comme homogène. Ensuite, il sera montré comment cette même technique lève un second verrou permettant de repenser de façon statistique l’idée de coefficient d’absorption. Cela ouvrira la voie à des algorithmes de Monte-Carlo qui estiment directement une observable radiative à partir de paramètres de transitions répertoriés dans des bases de données spectroscopiques, sans avoir à précalculer rigoureusement le coefficient d’absorption. / Two major challenges are encountered when studying radiative transfer in gases (e.g. combustion chambers or planetary atmospheres): heterogeneity and spectral dependence of radiative properties. The work introduced in this manuscript, addresses this problem through a statistical approach of radiation that requires no model or numerical approximation. This approach leads to the development of Monte-Carlo methods, currently considered as reference solutions in the community of radiative transfer. The difficulty related to heterogeneity is handled by a technique borrowed from other fields of transport physics: null-collision algorithms. Their application to radiation consists in adding to the events of absorption and scattering a third arbitrary type of collision that has no effect on the photon transport. Thus, the extinction coefficient resulting from these three types of collisions can be assumed to be homogeneous. Then, it is shown how this very same technique opens the door to rethinking statistically the concept of absorption coefficient. This leads to Monte-Carlo algorithms that directly estimate radiative observables from transition parameters indexed in molecular spectroscopic databases, without the need of rigorously precomputing absorption coefficients.