Ce mémoire est centré sur l’application de la technique d’optimisation de type Space Mapping dans le cadre du dimensionnement d’actionneurs électriques pris en compte par des modélisations multi-physiques. L’intérêt particulièrement recherché de ce type de méthode est la réduction potentiellement forte du coût du dimensionnement optimal. Cette volonté de réduction du coût de l’approche optimale s’explique par plusieurs considérations. En premier lieu, la modélisation des actionneurs tend à considérer de plus en plus de phénomènes physiques (tels que les phénomènes magnétiques, électriques, thermiques, mécaniques …) afin de décrire au mieux les phénomènes observés et mesurés. En second lieu il devient alors nécessaire de tenir compte des couplages entre ces physiques afin de traduire au plus juste l’interdépendance de ces phénomènes. Dans ce cadre, un travail particulier a été réalisé concernant la prise en compte des aspects thermiques dans les machines électriques. C’est ainsi qu’un modèle thermique à constantes localisées d’une machine synchrone à aimants permanents a été construit. Pour valider les résultats de calcul et préciser la définition de certain de ses éléments, une démarche expérimentale a été réalisée. Tous ces points, traduits dans le plan numérique, haussent le coût de l’évaluation des performances des actionneurs, et donc celui de leurs dimensionnements. De là, l’utilisation des techniques d’optimisation basées sur des modèles substituts permet d’envisager des réductions significatives des coûts de dimensionnement. La technique de Space Mapping est utilisée dans ce travail comme solution pour trouver un compromis entre la qualité des solutions trouvées et le temps de calcul. Plus particulièrement, elle est utilisée pour résoudre un problème de dimensionnement optimal d’une machine synchrone à aimants permanents assurant la fonction de démarreur dans une application de véhicule hybride. L’approche d’optimisation par Space Mapping a été comparée à celle, plus classique, n’utilisant qu’une seule modélisation de l’actionneur à dimensionner, c’est-à-dire sans modèle substitut. Il est montré que les techniques de Space Mapping sont à même de trouver des solutions de dimensionnement similaires à celles issues d’une approche classique, mais de manière beaucoup plus efficace, i.e. en utilisant un nombre plus faible d’évaluations de la modélisation multi-physique de l’actionneur. / This thesis focuses on the application of the Space Mapping optimization technique in the case of the sizing of electrical actuators taking into account a multi-physical modeling. The main interest in this type of optimization method is to considerably reduce the cost of optimal sizing. The need to use such optimization approach is due to several considerations. First, electrical actuators modeling tends to increasingly require the consideration of several physical phenomena (such as magnetic, electrical thermal and mechanical phenomena) in order to better describe observed and measured phenomena. Besides, it becomes necessary to take into account couplings between the different physical phenomena to precisely calculate the interdependencies between these phenomena. In this context, taking into account the thermal aspect in the case of electrical machines is particularly highlighted. A lumped parameter model of a permanent magnet synchronous machine is built. An experimental procedure has been followed to validate calculation results and define some elements of the proposed model. When implemented numerically, all points mentioned above increase the cost of the calculation of the performances of the electrical actuator, and then the cost of the optimal sizing. Thus, the use of an optimization technique based on surrogate models permits to reduce the optimal sizing cost. Space Mapping technique was used in this work as a solution to find a compromise between the quality of the found results and the calculation time. It is particularly used to solve an optimal sizing problem of a permanent magnet synchronous machine used as starter in a hybrid vehicle application. The Space Mapping optimization approach was compared to a classical one using a unique modeling of sized the electrical actuator : no surrogate model is used in the classical approach. Il is demonstrated that the Space Mapping techniques find optimization results that are similar to those found by the classical approach, yet, in a much more efficiently. Space Mapping techniques require only few calculations of the multi-physical model of the actuator.

As climate change escalates and the demands for energy resources increase, modern geotechnical engineering must tackle critical challenges to ensure sustainable development and enhance the resilience of infrastructure in society. The coupled chemo-hydro-mechanical processes in multiphase materials present significant challenges in geotechnical engineering, particularly for applications like carbon sequestration, geological disposal of nuclear waste, and hydraulic fracturing with reactive fluids, all of which involve highly heterogeneous and strongly anisotropic multiphysics environments. This dissertation introduces a multiphysical computational framework specifically designed to address the challenges associated with these unconventional applications. In this dissertation, we consider not only the local multiphysical coupling effects in the constitutive model but also the nonlocal effects arising from pore fluid flow, chemical species convection and diffusion, chemical reactions occurring in both solid and fluid constituents, and damage due to fluid pressure acting on fractures in the solid. We have integrated all these physical processes and developed a single unified model capable of handling the complex hydro-chemo-mechanical responses of geomaterials under varying geochemical conditions, confining pressures, and external loading scenarios. This computational framework offers a comprehensive simulation tool to investigate the long-term stability of geomaterials, which is determined by the intensity of chemical reactions under specific temperature and pressure conditions (assuming an isothermal condition in this dissertation), as well as the sustainability of geotechnical infrastructure in erosive environments driven by both mechanical and chemical processes. Three key aspects of engineering applications related to the effects of chemical reactions in geotechnical engineering are addressed. Firstly, we have integrated a complete calcite reaction system into poromechanics to couple pore geochemistry with poroelasticity theory. This integration is capable of predicting the geomechanical response essential for long-term stability analysis in \ch{CO2} sequestration engineering. Key features of this model include a multi-field finite element approach, local-equilibrium explicit geochemistry characterization of the calcite dissolution/precipitation reaction system, a robust algorithm for sequentially coupling pore geochemistry with poromechanics, and strategies to enhance the computational efficiency of solvers. Secondly, for applications involving acid working fluids in hydraulic fracturing, we have extended and adapted previous models within the phase field method framework. This extended integration effectively addresses the effects of chemically assisted fracturing in hydraulic fracturing operations. The key innovations of this model are the implementation of the phase field method to capture crack behaviors with poromechanics, the modeling of acid fluid transport in porous media and fractures, and its application to multiple mineral reaction systems. Thirdly, we have proposed a constitutive model that incorporates pore geochemistry and the pressure dissolution effect into internal variables, effectively capturing the chemical reactions contributing to softening in geomaterials. This model effectively illustrates and predicts chemically induced weathering or damage in granular porous media, such as sinkholes and subsidence. Derivations of a thermodynamically-based degradation index consider the influences of pore geochemistry and contact forces between grains and bonds. The model also proposes cross-scale relationships that consider reaction effects from individual particle sizes to particle aggregates. Furthermore, these relationships are incorporated into classical Cam-Clay-type models, along with the derivation of a consistent tangent modulus. / Dissertation / Doctor of Philosophy (PhD) / This thesis presents the comprehensive behaviors of geomaterials under mechanical, fluid, and chemical interactions, which result in displacement and cracking. Since there is no existing software or simulation tool that includes all the physical behaviors considered in this dissertation, the development and implementation of these physical mechanisms, followed by testing and analysis for engineering problems, constitutes the main contribution of this work. The newly developed simulation tool ranges from simulating the mechanical behavior of porous media saturated with water and reactive fluid to modeling the seepage of water/reactive fluid that triggers damage (cracks) in the porous media. This simulation tool can effectively analyze engineering problems that focus on the interactions between the working fluid and the host solid matrix under complex solution conditions. Examples include modeling carbon sequestration in saline aquifers and the storage of nuclear waste in subsurface repositories etc. The simulation tool proposed in this thesis incorporates rigorous mathematical derivations, efficient and accurate multiscale discretization techniques, robust non-iterative and iterative numerical coupling strategies, and thorough comparisons between numerical results and experimental/laboratory data. Simultaneously, it is important to recognize the model's limitations. Although the model assumes local equilibrium and interactions between physical mechanisms, it cannot fully capture all behaviors under these assumptions due to the restrictions in our understanding and potential constraints of numerical methods.

La maîtrise de la gestion thermique du moteur à combustion interne permet de répondre à des problématiques telles que la réduction de la consommation de carburant et des émissions de polluants. Cette gestion peut être réalisée par des systèmes mécatroniques, plus précisément grâce à une vanne électromécanique multivoie, appelée ACT valve (Active Cooling thermo-management valve), permettant de mettre en œuvre une stratégie de gestion thermique lors du contrôle des températures dans différentes branches du circuit de refroidissement du moteur thermique.L’objectif du travail est d’améliorer la robustesse du processus de conception de cette vanne en tenant compte des contraintes fonctionnelles multi-physiques telles que la déformation, l’usure, mais aussi de contraintes de fabrication et de tolérancement géométrique liées au processus d’injection des pièces thermoplastiques. Ces incertitudes doivent être prises en compte dès la phase de la conception pour assurer l'efficacité et la fiabilité de ces vannes jusqu'à la fin de leurs vies.Ces travaux de thèses proposent tout d’abord une nouvelle méthode de conception de ces vannes qui se base sur des modèles numériques multi-physiques permettant à tous les métiers d’avoir une base de données commune. Validés expérimentalement, Ces modèles ont permis de mettre en évidence la sensibilité de certains paramètres géométriques sur le couple développé par l’actionneur de la vanne et de s’assurer de la fiabilité du système par la prédiction de l’usure sur un des éléments clés de l’ACT valve.Bien que l’analyse de sensibilité nous ait permis de comprendre l’influence de certains paramètres sur le système, nous avons proposé une nouvelle technique d’identification des configurations optimales du design de cette vanne en utilisant une méthode d’optimisation méta-heuristique multi-objectifs. Les suggestions de conception offertes par cette méthode permettent de réduire le couple résistant sur l’actionneur de la vanne ainsi que l’encombrement global du système. / The thermal management of the internal combustion engine can solve issues related to fuel consumption reduction and pollutant emissions. This management can be applied using mechatronic systems, more precisely thanks to a multi-way electromechanical valve, called ACT valve (Active Cooling thermo-management valve), that presents a thermal management strategy when controlling temperatures in different branches of the engine cooling circuit.The aim of this work is to improve the robustness of the design process of this valve taking into account the multi-physical functional constraints such as deformation, wear, but also geometrical tolerances constraints related to the thermoplastic parts manufacturing process. These uncertainties must be taken into account in the first steps of the design phase to ensure the effectiveness and reliability of this valve over its lifetime.This work first proposes a new method of designing these valves, which is based on multi-physical modeling, allowing the product designers to have a common database. Experimentally validated, these models made it possible to highlight the sensitivity of certain geometrical parameters on the torque developed by the actuator of the valve and to make sure of the reliability of the system with wear prediction on one of the key elements of ACT valve.Although the sensitivity analysis allowed us to understand the influence of certain parameters on the system, we proposed a new technique for identifying optimal configuration configurations of this valve using a metaheuristic, multi-objective optimization method. The design suggestions offered by this method can reduce the resistive torque on the valve DC-actuator as well as the overall packaging.