Tesis sobre el tema "Structure porale complexe"

Dans un contexte de changements globaux et de transitions écologiques et énergétiques, le changement climatique induit une sècheresse récurrente et de crise de la ressource en eau dans plusieurs régions et continents. Il est donc crucial d'étudier les aquifères d'eaux douces afin d'accompagner ces transitions et assurer une gestion et une utilisation durable et raisonnée de cette ressource vitale. En l'occurrence, la zone vadose qui joue un rôle important dans la recharge de ces aquifères et le transfert d'éventuels polluants et d'intrants. La plupart du temps, cette zone non saturée en eau, se caractérise par la présence d'hétérogénéités (structure de pores, fractures, variation minéralogique) multi-échelles notamment lorsqu'il s'agit d'un milieu calcaire. Ces hétérogénéités sont complexifiées par les processus diagenétiques liés principalement à l'altération physico-chimique et minéralogique. La surimposition de ces phénomènes primaires et secondaires rend difficile l'estimation des propriétés pétrophysiques (porosité, perméabilité, saturation en eau) à partir des méthodes géophysiques. Parmi ces méthodes couramment utilisées, les méthodes acoustique et électrique sont bien adaptées car il existe, à priori, un lien étroit entre les hétérogénéités et les propriétés mesurées.Les travaux de recherche menés dans le cadre de cette thèse s'appuient ainsi sur une approche multi-géophysique permettant de mieux caractériser un réservoir carbonaté complexe en se basant notamment sur une analyse pétrophysique combinée à une description microstructurale fine.Dans ce travail, nous avons démontré l'influence de la microstructure sur la prédiction et la modélisation des propriétés pétroacoustiques ce qui conduit à une bonne discrimination de certains faciès, utile pour améliorer les modèles de simulation d'écoulements. De plus, ce travail a aussi permis de montrer la pertinence de la conductivité complexe dans la caractérisation des calcaires et la prédiction de leur perméabilité. Cependant, des développements supplémentaires sont nécessaires concernant la problématique du changement d'échelles des réservoirs complexes et hétérogènes<br>In a context of global changes and ecological and energy transitions, climate change induces recurrent drought and water resource crisis in several regions and continents. It is crucial to study groundwaters in order to support these transitions and ensure effective management and use of this vital resource. In particular, the vadose zone which plays an important role in the recharge of these groundwaters and the transfer of possible pollutants and inputs. Most of the time this unsaturated zone is characterized by multi-scale heterogeneities (e.g., pore structure, fractures, mineralogical variation) particularly in a limestone environment. These heterogeneities are complexified by diagenetic processes linked mainly to physicochemical and mineralogical alteration which leads to uncertainty in reservoir property (e.g., porosity, permeability, water saturation) estimation from geophysical methods. Among these methods, acoustic and electrical methods are well suited because of the strong relation between heterogeneities and the measured properties.This thesis relies on a multi-geophysical approach in order to better characterize a complex carbonate reservoir using petrophysical measurements combined with microstructural descriptions. Based on this approach, we demonstrated the influence of rock structure on the prediction and modeling of petroacoustic properties. This work leads to a good discrimination of some facies, which can be used to improve simulation and flow models. In addition, we demonstrate the relevance of complex conductivity measurements in limestone characterization and permeability prediction. However, additional developments are needed to understand the upscaling problematic for heterogeneous and complex reservoirs

With the advent of high-throughput technology there has been a large increase in the availability of biological data, such as interaction data of genes, proteins and metabolites. It is therefore necessary to develop ways in which these data can be efficiently modelled and analysed. Networks over a natural modelling framework for complex biological systems and as such, network theory and related computational approaches have proven important in bioinformatics. A particular facet of network theory that has been employed to analyse biological networks is community structure detection. Community structure is a modular network topology where communities are defined as groups nodes with dense intra-community connections and less dense inter-community connections. Methods to uncover such communities in complex biological networks have the potential to contribute towards a better understanding of the underlying organisation of a system. Consequently, this thesis focuses on the development of a series of mathematical programming models to address various manifestations of the community structure detection problem. The aim is to produce more information-rich models that can accurately represent the features of biological systems; with weighted and unweighted interactions, disjoint and overlapping communities and network dynamics all being considered. First, the detection of disjoint communities in unweighted networks is approached through a two-stage procedure, known as iMod. A mixed integer nonlinear programming (MINLP) model optimises modularity to find an initial partition which is then improved by iteratively solving a mixed integer quadratic programming (MIQP) model. A comparative analysis shows that iMod finds globally optimal solutions for networks of up to 512 nodes and outperforms all other methods tested when applied to larger networks. Subsequently, the MINLP model is generalised to weighted networks, known accordingly as WeiMod. Competitive results are found when WeiMod is compared with several other well known methods from the literature. Next, the work on disjoint community structure is extended to _nd overlapping modules. An MINLP model, known as OverMod, transforms disjoint to overlapping communities through the optimisation of the community strength metric. OverMod is compared with similar methods from the literature and is further assessed on protein-protein interaction (PPI) networks to test the method's ability to extract meaningful biological results. It is shown that proteins assigned to more than one module exhibit topological and functional properties indicative of their strategically important role in the organisation of the PPI networks. The work on disjoint and overlapping community structure concludes with the investigation of a network generated from sequence, protein interaction and co-expression data, for the fungal pathogen, Fusarium graminearum. The functional coherence of communities, properties of multiclustered genes and aspects of virulence-associated genes are all explored in an attempt to link topological and functional features. Finally, the concept of community structure in dynamic networks is explored. Consensus clustering is tackled; de_ned as detecting a single partition of a dynamic network that is relevant across multiple snapshots. This is addressed by extending the previously proposed MIQP and MINLP models such that average modularity across network snapshots is now optimised. A comparison is made with a similar method from the literature showing that the proposed approaches achieve competitive results for small to medium sized networks. Overall, this thesis demonstrates that the exible nature of mathematical programming lends itself well to developing versatile solution procedures for community structure detection. The method evaluations show the proposed algorithms to be comparable to other approaches from the literature and able to detect meaningful results in biological applications. Finally, the methods described in this thesis have the potential to infer important topological-functional relationships and help to provide insight into the organisation and evolution of biological systems.