The dissertation work focuses on the design and optimization of the castings and foundry technology and the production a particular model for the casting. The technology proposal and the pattern set building in 3D using CAD system. The casting simulation in CAE MAGMAsoft® system and evaluation of the proposed casting technology. Production of the pattern using CAD/CAM software and CNC machine tools with a focus on 3-axis and 5-axis machining. The economic benefits for the company.

Dans le domaine de la fabrication additive plusieurs technologies cohabitent et présentent des maturités et des applications différentes : le lit de poudre, la projection de poudre et le dépôt de fil pour ne citer que les principales. Nous avons étudié, dans le cadre de cette thèse, la réalisation de pièces de grandes dimensions du domaine aéronautique en alliage d’aluminium, par technologie WAAM (Wire Arc Additive Manufacturing) robotisée. Cette technologie repose sur l’utilisation un générateur de soudure à l'arc, d’un système de protection gazeuse et d’un système d'alimentation en métal d'apport sous forme de fil. Pour répondre à cette problématique, plusieurs voies de recherche ont été investiguées. La première traitait principalement de la génération de trajectoires : Plusieurs expérimentations ont permis de montrer l’intérêt et l’importance de la génération de trajectoires et notamment la maitrise de l’orientation outil pour la fabrication additive de pièces complexes en étudiant le respect de la géométrie souhaitée. La seconde concernait l’étude de la santé matière des pièces fabriquées. Des observations au niveau de la microstructure, mais aussi des caractéristiques mécaniques ont permis de mettre en évidence l’influence des paramètres opératoires sur la qualité de la matière déposée. Enfin, la réalisation de pièces fonctionnelles dans le cadre d’un projet financé par la DGA/DGAC et dont les partenaires étaient : STELIA, CONSTELLIUM, CT INGENIERIE et l’Ecole Centrale de Nantes, a permis de mettre en avant l’intérêt du procédé pour la fabrication de pièces aéronautiques. Un élément de structure aéronautique composé de raidisseurs a été fabriqué avec le procédé WAAM sur un substrat double courbure en alliage aluminium. Les difficultés accrues de réalisation ont pu être levées par l'emploi de la méthodologie développée dans le cadre de la thèse. / In the field of additive manufacturing (AM), several processes are present and have different applications and levels of development: the main technologies are powder-bed based AM, powder projection and Wire Additive Manufacturing (WAM). We have studied, in this PhD work, the manufacturing of large scale components in aluminum alloy for aircraft industry with Wire Arc Additive Manufacturing (WAAM). This technology is based on a welding generator, a shielding gas protection and a feedstock (wire in this case). To solve this issue, several ways of research were investigated. The first one dealt with toolpath generation: several experiments have highlighted the importance of tool path generation and the tool orientation to manufacture complex parts and improve the part accuracy. The second one was about the validation of the material quality after deposit. Microstructural observations and mechanical tests have demonstrated the effect of process parameters on the deposit quality. Finally, in the context of a DGA/DGAC funded research project, whose partners were STELIA, CT INGENIERIE, CONSTELLIUM and l’Ecole Centrale de Nantes, the manufacturing of functional part in aluminum alloy has shown the interest of the process for aircraft industry. A structural component based on a double curvature geometry has been manufactured with WAAM. The methodologies developed in this PhD work have enabled us to solve the issues to manufacture that type of component.

La réalisation de formes complexes comme les moules ou les prothèses médicales nécessite l’utilisation d’opérations de super finition pour obtenir de faibles défauts géométriques, pouvant aller jusqu’au poli-miroir. Ces opérations de pré-polissage et de polissage sont encore régulièrement réalisées manuellement. En effet, malgré des avantages en termes de répétabilité, de productivité et de qualité géométrique, les méthodes de polissage automatique sont peu utilisées car elles nécessitent une mise au point importante. Les travaux de recherche présentés participent à la maîtrise du procédé de polissage automatique tout en contrôlant la qualité géométrique des pièces. Pour parvenir à cette maîtrise, un outil de simulation de l’enlèvement de matière est mis en place. Cet outil permet de simuler l’enlèvement de matière au cours d’une opération de prépolissage réalisée sur centre d’usinage 5 axes. Il se base sur un modèle du contact obtenu entre l’outil de pré-polissage et la pièce, ainsi que sur un modèle du pouvoir abrasif intégrant l’usure et l’encrassement du disque. Cette simulation permet de vérifier la régularité de l’abrasion sur une surface et d’identifier les zones pouvant faire apparaitre des défauts macro-géométriques importants. Une méthode est également proposée pour compenser les variations du pouvoir abrasif au cours du temps. La compensation s’effectue en optimisant les consignes de vitesse d’avance et/ou de fréquence de broche le long de la trajectoire. Cette méthode de pilotage permet d’obtenir un taux d’enlèvement de matière plus constant et ainsi de réduire les défauts géométriques générés pendant une opération de prépolissage. / Complex shapes such as medical implants or injection molds require the use of super-finishing operations to minimize geometrical defects, down to mirror effect finish. These pre-polishing and polishing operations are still regularly performed manually by skilled workers. In spite of advantages in terms of repeatability, productivity and geometrical quality, automatic polishing methods are not widely used because they require systematic and significant developments. The present work contributes to enhance the automatic polishing process compared to the geometric quality of the parts. To achieve this control, a numerical simulation of material removal is implemented. This software simulates the material removal during a pre-polishing operation performed on 5-axis machining center. It is based on a contact model obtained between the pre-polishing tool and the part, as well as an abrasive model including wear of the disc. This simulation allows to check the uniformity of the material removal on the surface and to identify potential areas where macro-geometric defects appear. A method is also proposed to balance variations of the abrasive efficiency. The correction is performed by optimizing the federate and/or the spindle speed along the tool path. This method provides a constant material removal and reduces the geometrical deviations generated during pre-polishing operations.

Esta tese teve por objetivo desenvolver um algoritmo preciso e robusto capaz de compensar os erros térmicos volumétricos de um centro de usinagem 5 eixos em diferentes condições operacionais. O comportamento térmico da máquina foi modelado usando técnicas do método dos elementos finitos (MEF) com base na teoria do calor de atrito e calor de convecção, e validadas através dos vários campos de temperatura obtidos experimentalmente usando termopares e imagens térmicas. Os principais subsistemas da máquina foram inicialmente modelados, como o conjunto de fusos de esferas, guias lineares e motofuso, o que permitiu posteriormente a validação do comportamento termoelástico da máquina completa para onze ciclos de trabalho em vazio, seis ciclos de usinagem, nove ciclos de posicionamento e dois ciclos com temperatura ambiente variando, obtendo erros máximos inferiores a 9%, ao comparar os resultados numéricos com os resultados experimentais. A validação do modelo em elementos finitos permitiu usar os resultados obtidos para treinar e validar uma rede neural artificial (RNA) para prever os erros térmicos do centro de usinagem. Os desvios entre os erros térmicos previstos pela RNA e os calculados pelo MEF foram inferiores a 5%. Baseado nos resultados obtidos pelas medições das peças de trabalho usinadas foi possível formular e implementar um modelo de compensação dos erros térmicos no CNC do centro de usinagem, que obteve uma redução dos erros entre 62% e 100% nas peças usinadas com compensação. Foi também proposto um algoritmo de previsão e compensação dos erros térmicos para o centro de usinagem, baseado em todos os ciclos e simulações realizadas, e que se comparando com os resultados experimentais mostrou-se capaz de reduzir os erros térmicos entre 50% e 95%. Após sua validação, foi possível concluir que o algoritmo desenvolvido é uma ferramenta precisa e robusta para compensar os erros térmicos da máquina para várias condições de trabalho, podendo compensá-los mesmo com esta movendo-se a diferentes velocidades, em usinagem ou mesmo operando em temperatura ambiente variável. / This thesis aims to develop an accurate and robust algorithm capable of compensating the volumetric thermal errors of a 5-axis machining center under different operating conditions. The thermal behavior of the machine was first modeled using finite element method (FEM) techniques based on theory of friction heat and convection heat, and validated with the various experimentally raised temperature fields using thermocouples and thermal imaging. The main machine subsystems were initially modeled, such as the ball screw system, linear guides and spindle, which allowed for validating of the thermoelastic behavior of the entire machine for eleven no load duty cycles, six cycles of machining, nine cycles of positioning and two cycles with varying ambient temperature, obtaining errors lower than 9%, when comparing the numerical results with the experimental results. The validation of the finite element model allowed for the use of the results obtained to train and validate an artificial neural network (ANN) for predicting the thermal errors of the machining center. The deviations between the thermal errors predicted by ANN and the FEM simulation results were less than 5%. Based on the results obtained by the measurements of the machined workpieces, it was possible to formulate and implement a model of compensation of the thermal errors in the CNC of the machining center, which obtained a reduction of errors of 62% and 100% of the machined parts with compensation. It was also proposed a thermal error prediction and compensation algorithm for the machining center, based on all cycles and simulations performed, and that, comparing with the experimental results, it was able to reduce the thermal errors between 50% and 95%. After its validation, it was possible to conclude that the developed algorithm is an accurate and robust tool to compensate the thermal errors of the machine for various duty conditions, being able to compensate the errors even when it is moving at different speeds, in machining process or even operating in variable ambient temperature.

Planning of optimal toolpath, cutter orientation, and feed rate for 5-axis Computer Numerical Control (CNC) machining of curved surfaces using a flat-end mill is a challenging task, although the approach has a great potential for much improved machining efficiency and surface quality of the finished part. This research combines and introduces several key enabling techniques for curved surface machining using 5-axis milling and a flat end cutter to achieve maximum machining efficiency and best surface quality, and to overcome some of the key drawbacks of 5-axis milling machine and flat end cutter use. First, this work proposes an optimal toolpath generation method by machining the curved surface patch-by-patch, considering surface normal variations using a fuzzy clustering technique. This method allows faster CNC machining with reduced slow angular motion of tool rotational axes and reduces sharp cutter orientation changes. The optimal number of surface patches or surface point clusters is determined by minimizing the two rotation motions and simplifying the toolpaths. Secondly, an optimal tool orientation generation method based on the combination of the surface normal method for convex curved surfaces and Euler-Meusnier Sphere (EMS) method for concave curved surfaces without surface gouge in machining has been introduced to achieve the maximum machining efficiency and surface quality. The surface normal based cutter orientation planning method is used to obtain the closest curvature match and longest cutting edge; and the EMS method is applied to obtain the closest curvature match and to avoid local gouging by matching the largest cutter Euler-Meusnier sphere with the smallest Euler-Meusnier sphere of the machined surface at each cutter contact (CC) point. For surfaces with saddle shapes, selection of one of these two tool orientation determination methods is based on the direction of the CNC toolpath relative to the change of surface curvature. A Non-uniform rational basis spline (NURBS) surface with concave, convex, and saddle features is used to demonstrate these newly introduced methods. Thirdly, the tool based and the Tri-dexel workpiece based methods of chip volume and cutting force predictions for flat-end mills in 5-axis CNC machining have been explored for feed rate optimization to achieve the maximum material removal rate. A new approach called local parallel slice method which extends the Alpha Shape method - only for chip geometry and removal volume prediction has been introduced to predict instant cutting forces for dynamic feed rate optimization. The Tri-dexel workpiece model is created to get undeformed chip geometry, chip volume, and cutting forces by determining the intersections of the tool envelope and continuously updating the workpiece during machining. The comparison of these two approaches is made and several machining experiments are conducted to verify the simulation results. At last, the chip ploughing effects that become a more serious problem in micro-machining due to chip thickness not always being larger than the tool edge radius are also considered. It is a challenging task to avoid ploughing effects in micro-milling. A new model of 3D chip geometry is thus developed to calculate chip thickness and ploughing volume in micro 5-axis flat-end milling by considering the minimum chip thickness effects. The research forms the foundation of optimal toolpath, cutter orientation, cutting forces/volume calculations, and ploughing effects in 5-axis CNC machining of curved surfaces using a flat-end mill for further research and direct manufacturing applications. / Graduate / 0548 / luoshan@uvic.ca

Esta tese teve por objetivo desenvolver um algoritmo preciso e robusto capaz de compensar os erros térmicos volumétricos de um centro de usinagem 5 eixos em diferentes condições operacionais. O comportamento térmico da máquina foi modelado usando técnicas do método dos elementos finitos (MEF) com base na teoria do calor de atrito e calor de convecção, e validadas através dos vários campos de temperatura obtidos experimentalmente usando termopares e imagens térmicas. Os principais subsistemas da máquina foram inicialmente modelados, como o conjunto de fusos de esferas, guias lineares e motofuso, o que permitiu posteriormente a validação do comportamento termoelástico da máquina completa para onze ciclos de trabalho em vazio, seis ciclos de usinagem, nove ciclos de posicionamento e dois ciclos com temperatura ambiente variando, obtendo erros máximos inferiores a 9%, ao comparar os resultados numéricos com os resultados experimentais. A validação do modelo em elementos finitos permitiu usar os resultados obtidos para treinar e validar uma rede neural artificial (RNA) para prever os erros térmicos do centro de usinagem. Os desvios entre os erros térmicos previstos pela RNA e os calculados pelo MEF foram inferiores a 5%. Baseado nos resultados obtidos pelas medições das peças de trabalho usinadas foi possível formular e implementar um modelo de compensação dos erros térmicos no CNC do centro de usinagem, que obteve uma redução dos erros entre 62% e 100% nas peças usinadas com compensação. Foi também proposto um algoritmo de previsão e compensação dos erros térmicos para o centro de usinagem, baseado em todos os ciclos e simulações realizadas, e que se comparando com os resultados experimentais mostrou-se capaz de reduzir os erros térmicos entre 50% e 95%. Após sua validação, foi possível concluir que o algoritmo desenvolvido é uma ferramenta precisa e robusta para compensar os erros térmicos da máquina para várias condições de trabalho, podendo compensá-los mesmo com esta movendo-se a diferentes velocidades, em usinagem ou mesmo operando em temperatura ambiente variável. / This thesis aims to develop an accurate and robust algorithm capable of compensating the volumetric thermal errors of a 5-axis machining center under different operating conditions. The thermal behavior of the machine was first modeled using finite element method (FEM) techniques based on theory of friction heat and convection heat, and validated with the various experimentally raised temperature fields using thermocouples and thermal imaging. The main machine subsystems were initially modeled, such as the ball screw system, linear guides and spindle, which allowed for validating of the thermoelastic behavior of the entire machine for eleven no load duty cycles, six cycles of machining, nine cycles of positioning and two cycles with varying ambient temperature, obtaining errors lower than 9%, when comparing the numerical results with the experimental results. The validation of the finite element model allowed for the use of the results obtained to train and validate an artificial neural network (ANN) for predicting the thermal errors of the machining center. The deviations between the thermal errors predicted by ANN and the FEM simulation results were less than 5%. Based on the results obtained by the measurements of the machined workpieces, it was possible to formulate and implement a model of compensation of the thermal errors in the CNC of the machining center, which obtained a reduction of errors of 62% and 100% of the machined parts with compensation. It was also proposed a thermal error prediction and compensation algorithm for the machining center, based on all cycles and simulations performed, and that, comparing with the experimental results, it was able to reduce the thermal errors between 50% and 95%. After its validation, it was possible to conclude that the developed algorithm is an accurate and robust tool to compensate the thermal errors of the machine for various duty conditions, being able to compensate the errors even when it is moving at different speeds, in machining process or even operating in variable ambient temperature.

Content of the thesis is the proposal of new technology using machining centres, including five controlled axis. Benefit of implementation the new nechnology is investigated at machining example, which is the gear housing of a single stage gearbox, manufactured in the company. In the first part of the thesis, the general properties of gearboxes and their housings are analyzed, as well as the procedure of the current proces of the machining representative. In practical part of the thesis is proposed necessary tool equipment and jig equipment, needed for machining of the part. After that are analyzed and filled in time studies, provided by manufacturers of machining centres. On the bases of these studies, were calculated operating times and operating costs per one piece of the housing. In the part of technical and economical evaluation is calculated the return of initial investment, including costs for byuing new machining cente and cutting tools equipment. Finaly are evaluated the overal benefits of the new technology, given the needs of the company.

Dans le contexte d'inspection du procédé d'usinage, l'intégration des opérations de mesure au processus de fabrication permet de réaliser les actions correctives à appliquer aux opérations de fabrication et ainsi de diminuer le temps de cycle de fabrication. Nous nous sommes intéressés dans ces travaux à la réalisation de la mesure on-machine par capteur laser plan dans une MOCN 5 axes. La problématique liée à la mesure on-machine concerne la proposition d'une méthode de planification de trajectoires qui résulte d'un compromis entre le temps de mesure et la qualité des données acquises.Nous proposons dans un premier temps un format de description des paramètres de la trajectoire du capteur permettant d'intégrer l'orientation de la broche dans la machine. Comme la détermination de point de vue dans notre contexte est adoptée pour le capteur laser plan, les contraintes de visibilité et de qualité sont appliquées à la position de ligne laser sur la surface. Nous proposons de réaliser d'un simulateur de trajectoires de numérisation. Ce simulateur nous permet d'identifier les zones de la pièce correctement numérisées en termes des critères de visibilité et de qualité. Il permet à la fois de valider la trajectoire avant l'exécution de la numérisation sur le système physique et d'aider à la génération de trajectoire de numérisation.Pour contrôler le temps et la qualité de numérisation, une méthode de planification de la trajectoire de numérisation pour une structure 6 axes, ISOvScan, est proposée. Dans cette méthode, la zone de recouvrement entre deux passes consécutives est constante, l'orientation du capteur en chaque point piloté est gérée afin d'assurer la qualité de numérisation et de maximiser le taux de couverture du laser. La méthode repose sur l'utilisation de la transformation conforme au sens des moindres carrés (Least Squares Conformal Maps-LSCM) pour transformer la surface 3D en un espace 2D. Le calcul de la trajectoire de numérisation peut alors être effectué plus simplement dans cet espace 2D.Comme la cinématique de la machine-outil 5 axes ne permet pas d'appliquer directement la trajectoire 6 axes créée par ISOvScan, les méthodes pour générer une trajectoire du capteur laser adaptée à la machine-outil 5-axes sont proposées dans ces travaux. Une application de numérisation sur la machine-outil 5-axes est réalisé pour vérifier la validité de la trajectoire de numérisation 5 axes. / In the context of the inspection of the machining process, the integration of the measurement operations into the manufacturing process allows to realize the corrective actions applied to the manufacturing operations and thus to reduce the manufacturing cycle time. Our objective in this work is to realize an on-machine measurement (OMM) by laser plane sensor in a 5-axis machine-tool. The problem related to OMM concerns the proposition of a scan path planning method that results from a compromise between the measurement time and the quality of the data acquired.First of all, we propose a description format of the parameters of the sensor path allowing to integrate the orientation of the spindle in the machine. As the determination of view point in our context is adopted for the laser plane sensor, the visibility and quality constraints are applied to the digitizing laser line position on the surface. We propose to realize a simulator of digitizing path. This simulator allows us to identify correctly the areas of the digitized part in terms of the scanning visibility and quality. It allows to validate the scan path before executing the scan on the physical system and to help the scan path generation.To control the scan time and quality, a scan path planning method for a 6-axis structure, ISOvScan, is proposed. In this method, the overlap zone between two consecutive passes is constant, the orientation of the sensor at each driven point is managed in order to ensure the scanning quality and to maximize the coverage rate of the laser. The method relies on the use of the Least Squares Conformal Maps (LSCM) to transform the 3D surface into a 2D space. The calculation of the scan path can then be done more simply in this 2D space.Since the kinematics of the machine tool does not enable to directly apply the 6 axes scan path created by ISOvScan, the methods for generating a laser scan path adapted to 5-axis machine-tool are proposed in this work. An application on the 5-axis machine-tool is performed to validate this adapted scan path.

La qualité des surfaces des veines fluides fermées des pièces tournantes de turbomachine participe au rendement de la turbomachine. Il est donc essentiel de maîtriser la finition des veines en usinage 5 axes avec une fraise boule. L'alliage de titane Ti6Al4V est l'un des matériaux utilisés et souffre d'une faible usinabilité. Le choix des paramètres de coupe conditionne la qualité de surface et la durée de vie de la fraise. Pour maîtriser le fraisage 5 axes, des méthodes de caractérisation de la coupe sont développées pour des trajectoires linéaire et circulaire. Les diamètres effectifs et l'épaisseur coupée sont à l'origine de plusieurs phénomènes associés à la coupe tels que la vitesse de coupe, la vitesse d'évolution de l'usure d'outil, des modes d'usinage et des efforts de coupe. Des essais font le lien entre les mesures d'efforts de coupe et d'état de surface avec les méthodes de caractérisation de la coupe. Une fois l'usinage 5 axes en fraise boule caractérisé, deux stratégies de finition multiaxes de veines fermées sont développées en gardant constantes la vitesse d'avance du point générateur et l'orientation relative de l'axe de l'outil avec la normale de la surface locale. La stratégie de tréflage se caractérise par une trajectoire continue en courbure. La stratégie de contournage hélicoïdal met en évidence des discontinuités en tangence de la trajectoire. Une méthode de lissage local de trajectoire est alors développée pour assurer un comportement cinématique et dynamique raisonnable de la machine. / The surface quality of closed fluid veins rotating parts of turbo machines participates in the machine output. It is therefore essential to control the finishing of veins in 5-axis machining with a ball end mill. The titanium alloy Ti6Al4V is one of the materials used and suffers from a poor machinability. The choice of cutting parameters affects the surface quality and the life of the cutter. In order to control the 5-axis milling, characterization methods of cutting are developed for linear and circular paths. Effective diameters and the uncut chip thickness is responsible for several phenomena associated with the cut such as the cutting speed, the speed of evolution of the tool wear, the milling modes and cutting forces. Tests are the link between measures of cutting forces and surface quality and characterization methods of cutting. Once the 5-axis machining with ball end mill characterized, two strategies of finishing closed veins in multiaxis are developed keeping constant the feed speed of the contact and the relative orientation of the tool axis with the normal the local surface. The plunge milling strategy is characterized by a curvature continuous trajectory. The helical milling strategy reveals tangent discontinuities of the trajectory. A method of local smoothing trajectory is then developed to provide a reasonable kinematics and dynamics behavior of machine.

La conception des engrenages spiro-coniques reste encore très complexe de nos jours car la géométrie des dentures, et donc les performances cinématiques, découle du mode de fabrication de ce type d’engrenage. Le taillage est lié à deux constructeurs principaux : Gleason et Klingelnberg. De nombreux paramètres de réglage des machines influencent directement les surfaces de denture, leur optimisation n’est donc pas intuitive. Avec les progrès réalisés cette dernière décennie par les machines d’usinage à commande numérique et la FAO (Fabrication Assistée par Ordinateur), il devient possible de fabriquer des engrenages spiro-coniques de bonne qualité sur une machine 5 axes. Un modèle numérique a été développé pour générer une géométrie simplifiée de type Gleason, usinée par la suite avec une machine 5 axes. Une étude de métrologie, permettant de comparer les dents usinées avec les modèles CAO, a ensuite été réalisée pour prouver que l’usinage par une machine 5 axes peut être une alternative aux méthodes de taillage classiques. De nouveaux types de géométrie peuvent donc être proposés, qui ne pouvaient pas être envisagés par les moyens de fabrication classiques. Une géométrie basée sur la théorie des développantes sphériques, combinée à une spirale logarithmique a été développée, puis usinée. De plus, des corrections de bombé ou de profil peuvent être définies afin d’éviter les contacts en bords de denture. Ce type de géométrie analytique offre des possibilités plus simples d’optimisation de l’engrènement. L'optimisation des surfaces peut être réalisée à l’aide du modèle d’engrènement quasi-statique sous charge développé dans le cadre de cette thèse. L’environnement de l’engrenage est pris en compte dans la simulation : déformation des arbres, des dentures et de leurs supports (jantes et voiles) ainsi que les déformations locales de contact. La méthode des coefficients d’influence est utilisée pour résoudre le partage des charges entre toutes les dents instantanément en contact. Une méthode originale, utilisant sur un seul calcul élément finis et la définition de bases de fonctions, permet de calculer rapidement les flexions de denture dans leur environnement. Les déformations de contacts sont, quant à elles, obtenues par une méthode analytique, basée sur les théories de Boussinesq. De plus, des défauts d’assemblage peuvent être intégrés entre le pignon et la roue spiro-conique. Afin de valider les modèles numériques développés, un banc d’essai a été mis en place, permettant la mesure de l'erreur de transmission et la visualisation des portées. Le banc d’essai est intégré dans une fraiseuse numérique 3 axes : le pignon est monté dans la broche de la fraiseuse, le reste du banc étant bridé sur son plateau. Ainsi, des défauts de montage peuvent être appliqués facilement et précisément. / The design of spiral bevel gears is still very complex nowadays because the tooth geometry, and thus the kinematic performance, come from the manufacturing process of this type of gear. The cutting is related to two major manufacturers: Gleason and Klingelnberg. Many machine settings drive directly the shape of teeth surfaces, their optimization is therefore not intuitive. Due to the progress made during the last decade by the CNC machines and the CAM (Computer Aided Manufacturing) softwares, it becomes possible to manufacture spiral bevel gears of quite correct quality on a 5-axis milling machine. A numerical model was developed in order to generate a simplified type Gleason geometry. This last was then manufactured with a 5-axis milling machine. A metrological study, comparing the teeth obtained with the CAD models, was then carried out to prove that the manufacturing by 5-axis milling machine can be an alternative to conventional cutting methods. New types of geometry can be then proposed, which could not be considered by the conventional methods of manufacturing. Geometry based on the theory of spherical involutes, combined with a logarithmic spiral was developed and then manufatured. In addition, profile and crowning modifications can be defined to avoid the tooth edge contacts. This type of analytical geometry offers simpler possibilities for optimizing the meshing. The surface optimization can be achieved using the quasi-static meshing model under load developed in the context of this thesis. The surroundings of the gear are taken into account in the simulation: deformation of the shafts, of the gears and their supports (rims for example) as well as the local contact deformations. The influence coefficient method is used to solve the load sharing between all the teeth instantaneously in contact. An original method, using only one finite element computation and the definition of a set of functions, can quickly calculate the teeth bending, taking into account their surroundings. The contact deformations are obtained with an analytical method, based on Boussinesq theories. In addition, meshing defects can be integrated between the spiral beval pinion and gear. To validate the numerical model, a test bench was achieved, allowing the measurement of the loaded transmission error and the visualization of the contact patterns. The test bench is integrated inside a numerical 3-axis milling machine: the pinion is mounted in the spindle of the milling machine, when the base of the bench is clamped on its plate. Thus, assembly errors can be imposed easily and accurately.