Le travail présenté dans cette thèse concerne la modélisation et la commande robuste en force de robots manipulateurs industriels à articulations flexibles utilisés pour le procédé FSW. Afin de réduire les temps de calcul et l'occupation de la mémoire, une approche basée sur la méthode par intervalle est proposée en vue de la simplification des modèles dynamiques des robots industriels, et contribue à identifier les paramètres d'inertie qui sont négligeables. Des études de cas sur trois types de trajectoires de test et l’analyse des couples moteurs ont démontré l'efficacité et les bonnes performances de la méthode de simplification. Ensuite, la modélisation dynamique et l'identification des paramètres du procédé FSW ont été effectuées. Les paramètres des modèles linéaires et non-linéaires de forces axiales sont identifiés. Sur la base de la modélisation du procédé FSW qui considère simultanément la cinématique du système complet, le modèle de déplacement du robot rigide, les flexibilités des articulations et le modèle dynamique de la force axiale, un contrôleur robuste en force est obtenu par la méthode de réglage fréquentielle. En outre, un simulateur du procédé FSW robotique est développé et les résultats de simulation montrent les bonnes performances du contrôleur en force. L'oscillation de la force axiale dans le procédé FSW peut être simulée en utilisant un modèle de perturbation de la position verticale de référence. / The work presented in this thesis focuses on the modeling and robust force control of flexible joints industrial robot manipulators used for FSW process. In order to reduce computation time and memory occupation, a novel interval-based approach for dynamic model simplification of industrial robots is proposed, which applies to arbitrary trajectories of whole robot workspace and contributes to obtaining negligible inertia parameters. Cases studies have been carried out on three kinds of test trajectories and torques analysis of robot dynamic equation, demonstrating the effectiveness and good performance of the simplification method. Then, the dynamic modeling and identification of robotic FSW process is performed, and the parameters of linear and nonlinear dynamic axial force process models are identified by using the plunge depth and its derivative. On the basis of the modeling of robotic FSW process which simultaneously considers the complete kinematics, the rigid robot displacement model, the joint flexibility and the dynamic axial force process model, a robust force controller can be obtained by using the frequency response approach. Besides, a simulator of robotic FSW process is developed and simulation results show good performance of the force controller. The oscillation of axial force in FSW process can be simulated when a disturbance model of initial vertical reference position is proposed and used in the simulation.

This project explores the feasibility of the solid-state welding method direct-drive friction welding to be used as a joining method for the nitrogen alloyed steel Uddeholm Vanax SuperClean, produced via processes based on powder metallurgy. Vanax SuperClean cannot be welded using fusion welding methods where the base material melts, due to nitrogen escaping the material, resulting in inferior quality welds. The cost of the material motivates the use of Vanax SuperClean for critical parts in applications, combined with a less costly material for the remaining parts, causing alternative joining methods to be examined. Vanax SuperClean is friction welded to itself and to Uddeholm steel types Stavax ESR and UHB 11. Samples are prepared for a number of examinations. Microstructures of the samples are examined using microscopy, microhardness testing is carried out per the Vickers principle, retained austenite is measured using X-ray diffraction and tensile testing of the welded samples is performed. Defect-free welds are produced in all examined samples, showing that the method is suitable for Vanax SuperClean and that no preheating or slow cooling of workpieces are necessary. The possibility of using friction stir welding as a joining method for Vanax SuperClean is discussed. / Detta projekt undersöker möjligheten att använda trycksvetsningsmetoden friktionssvetsning som sammanfogningsmetod för det kvävelegerade pulvermetallurgiskt framställda stålet Uddeholm Vanax SuperClean. Vanax SuperClean kan inte svetsas med smältsvetsmetoder där grundmaterialet smälter, på grund av kvävgasbildning som resulterar i undermåliga svetsfogar. Kostnaden för materialet motiverar användandet av Vanax SuperClean för kritiska delar i applikationer, kombinerat med ett mindre kostsamt material till övriga delar, vilket föranleder undersökning av alternativa sammanfogningsmetoder. Vanax SuperClean friktionssvetsas mot sig själv, såväl som mot Uddeholmsstålen Stavax ESR och UHB 11. Prov tas fram för ett antal undersökningar. Mikrostruktur undersöks med mikroskopi, mikrohårdhetsprovning utförs enligt Vickersprincipen, restaustenitnivåer mäts med röntgendiffraktion och dragprovning utförs. Lyckade svetsfogar fås i alla undersökta prover, vilket visar att svetsmetoden är lämplig för Vanax SuperClean och att varken förvärmning eller långsamt svalnande av arbetsstycken krävs. Möjligheten att använda friktionsomrörningssvetsning som sammanfogningsmetod för Vanax SuperClean diskuteras.

<p>Within the framework of presented PhD, friction stir spot welding (FSSW) of<br />multiple ultrathin sheets of AA 5754 &ndash; H111 (AlMg3) alloy 0.3 mm thick, was<br />studied. The influence of tool geometry and process parameters such as rotational<br />speed and axial load have been analysed using numerous techniques. It has been<br />understood that during the welding at low rotational speeds weld zone undergoes<br />strain hardening, while at high rotational speeds weld zone undergoes thermal<br />softening. It was observed that during FSSW at low rotational speeds a complex<br />layer at weld interface is present, which causes delamination when welded samples<br />are subjected to load.</p> / <p>У оквиру ове докторске дисертације испитивано је тачкасто заваривање трењем са мешањем ултратанких лимова дебљине 0.3 mm од легуре АА 5754 &ndash; H111 (AlMg3). Утицај геометрије алата и параметара као што су угаона брзина и аксијално оптерећење су детаљно анализирани уз помоћ бројних техника. Установљено је да приликом заваривања ниским угаоним брзинама долази до деформационог ојачавања, док на високим угаоним брзинама долази до термичког омекшавања зоне завара. Код узорка завареног са најмањим бројем обртаја долази до формирања комплексног слоја на међуконтактној површини који изазива деламинацију приликом испитивања механичких особина.</p> / <p>U okviru ove doktorske disertacije ispitivano je tačkasto zavarivanje trenjem sa mešanjem ultratankih limova debljine 0.3 mm od legure AA 5754 &ndash; H111 (AlMg3). Uticaj geometrije alata i parametara kao što su ugaona brzina i aksijalno opterećenje su detaljno analizirani uz pomoć brojnih tehnika. Ustanovljeno je da prilikom zavarivanja niskim ugaonim brzinama dolazi do deformacionog ojačavanja, dok na visokim ugaonim brzinama dolazi do termičkog omekšavanja zone zavara. Kod uzorka zavarenog sa najmanjim brojem obrtaja dolazi do formiranja kompleksnog sloja na međukontaktnoj površini koji izaziva delaminaciju prilikom ispitivanja mehaničkih osobina.</p>

Les simulations numériques représentent un outil fondamental pour la conception et l'optimisation de procédés industriels de fabrication tels que le soudage. Malgré le développement impressionnant des méthodes numériques et des moyens de calcul utilisables, la complexité des procédés de fabrication et les nouvelles exigences des industries les plus avancées obligent à repenser les méthodes, les stratégies et les algorithmes de simulation disponibles. Dans cette thèse, de nouvelles méthodes numériques avec une approche de Réduction des Modèles sont proposées, une discipline consolidée qui a fourni des solutions étonnantes dans différentes applications, comme les procédés de fabrication avancés. Tout d'abord, différentes stratégies sont proposées pour la simulation efficace des procédés de soudage conventionnel, à cet effet, l'utilisation de Computational Vademecums est introduite. L’introduction de ces abaques numériques améliorent des méthodes telles que : les Éléments Finis Généralisés pour le calcul thermique, l'approche local-global pour le calcul mécanique et enfin, la construction directe des abaques numériques utiles pour la phase de pré-design. En second lieu, un solveur PGD efficace est présenté pour les simulations thermo-mécaniques de soudage par friction-malaxage. Cette thèse montre comment la réduction des modèles,en plus d'être une fin en soi, peut être un excellent ingrédient pour améliorer l'efficacité des méthodes numériques traditionnelles. Cela représente un grand intérêt pour l'industrie. / Numerical simulations represent a fundamental tool for the design and optimization of industrial manufacturing processes such as welding. Despite the impressive development of the numerical methods and the means of calculation, the complexity of these processes and the new demands of the more advanced industries make it necessary to rethink the available methods, strategies and simulation algorithms. In this thesis, we propose new numerical methods with a Model Order Reduction approach, a consolidated discipline that has provided surprising solutions indifferent applications, such as advanced manufacturing processes. First, different strategies for the efficient simulation of conventional welding processes are proposed. To this end, the use of Computational Vademecums is introduced for the improvement of methods such as the Generalized Finite Element for thermal calculation, the local-global approach for the mechanical calculation or the direct construction of vademecums useful for predesign phases. Then, an efficient PGD solver for thermomechanical simulations for friction stir welding is presented. This thesis shows how Model Reduction, besides being an end, it can be an excellent ingredient to improve the efficiency of traditional numerical methods, with great interest for the industry.

Friction stir welding (FSW) of steel is often performed with an insert made of polycrystalline cubic boron nitride (PCBN). Specifically, MS80 is a grade of PCBN made by Smith MegaDiamond that has been optimized for the FSW process. The PCBN insert is attached to a tungsten carbide (WC) shank by a compression fitting. However, FSW tools manufactured by this method inevitably fail by fracture in the PCBN. Permanently bonding PCBN to WC would likely solve the fracturing problem and increase the life of PCBN FSW tools to be economically viable. Partial transient liquid phase (PTLP) bonding, a process used to join ceramics with thin metallic interlayers, was proposed as a method to permanently bond PCBN to WC. PTLP bonding is often performed using three layers of pure elements. On heating, the two thin outer interlayers melt and bond to the ceramics. Concurrently, these liquid layers diffuse into the thicker refractory core until solidification has occurred isothermally. A procedure was developed to reduce the number of possible three-layer PTLP bonding setups to a small set of ideal setups using logical filters. Steps in this filtering method include a database of all existing binary systems, sessile drop testing of 20 elements, and a routine that calculates maximum interlayer thicknesses. Results of sessile drop testing showed that the PCBN grade required for this research could only be bonded with an alloy of Ti, Cu, Mg, and Sb. Two PTLP bond setups were tested using this special coating on the PCBN, but a successful bond could not be achieved. However, a PTLP bond of WC to WC was successful and proved the usefulness of the filtering procedure for determining PTLP bond setups. This filtering procedure is then set forth in generalized terms that can be used to PTLP bond any material. Also, recommendations for future research to bond this grade of PCBN, or some other grade, to WC are presented.