L’émergence du concept d’un avion plus électrique implique une refonte en profondeur des réseaux embarqués. L’axe principalement développé repose sur une augmentation de la tension du réseau et un passage au bus continu. Le contrôle des actionneurs électriques embarqués passe alors par l’utilisation massive de convertisseurs électroniques qui vont imposer des fronts de tension très raides et des contraintes particulières aux enroulements des machines. Les oscillations pseudopériodiques qui suivent ces fronts raides sont à l’origine de champs électriques intenses qui peuvent dépasser le seuil d’apparition des décharges partielles (PDIV) et de ce fait provoquer une dégradation rapide des isolants classiques qui sont à base de polymères. Le travail effectué dans cette thèse est centré sur l’étude d’une nouvelle méthode de conception des bobinages permettant la réduction des problèmes de décharges partielles. Un dispositif expérimental mis au point dans ce cadre reproduit les contraintes imposées par des convertisseurs modernes et permet de mesurer les tensions inter-spires d’une bobine. L’étude expérimentale met en évidence les phénomènes àprendre en compte avec ce type d’alimentation et donne une première idée des axes d’optimisation envisageables. Un modèle prédictif capable de donner la distribution des tensions inter-spires, en réponse à un front raide de tension, aide à identifier les points critiques. Les paramètres constitutifs de ce modèle ont été déterminés en utilisant des méthodes analytiques, expérimentales et numériques. Compte tenu de la complexité du modèle, un outil numérique automatisé a été élaboré afin de simuler le modèle des bobines étudiées. Plusieurs types de bobinage ont été analysés et optimisés avec succès. Cette optimisation consiste à diminuer la tension inter-spires en agissant sur l’arrangement des fils dans les encoches. L’utilisation des bobines optimisées avec cette méthode permet d’envisager une nette amélioration de la fiabilité des machines électriques. / The emergence of the concept of a more electric aircraft requires a major overhaul of onboard grid. The main developed axis is based on an increase in the grid voltage and a switch to DC bus. In this context, the onboard electric actuators are supplied through power electronic converters that will impose very fast voltage fronts and particular constraints in the windings of the machines. The transient over-voltages that follow the steep-fronted pulses are the source of intense electric fields which may exceed the partial discharge inception voltage (PDIV) and cause a rapid deterioration of polymer based insulators. This thesis focuses on the study of a new design method of the windings in order to reduce problems of partial discharges. An experimental device developed in this framework reproduces the constraints of modern converters and allows measuring the turn-to-turn voltages of a coil. The experimental study shows the phenomena to be considered with this type of power supply and gives an idea of possible areas for optimization. A predictive model able to give the distribution of turn-to-turn voltages, in response to a steep-fronted voltage, helps to identify critical points. The constitutive parameters of this model were determined using analytical, experimental and numerical methods. Given the complexity of the model, an automated numerical tool has been developed in order to simulate the model of studied coils. Several types of coil have been analyzed and optimized successfully. This optimization consists in reducing the turn-to-turn voltages by acting on the arrangement of the wires in the slots. The use of optimized coils with this method allows envisaging a significantimprovement in the reliability of electrical machines.

Within the framework of this thesis, the electron beam melting of pure copper for the purpose of producing electrical windings was examined. The conventional manufacturing process of copper windings was investigated, and the potential advantages offered by the design freedom inherent in the electron beam melting process were explored. A comprehensive parameter study was conducted to optimize the existing production parameters for the electron beam melting of pure copper, with a specific focus on achieving the desired material properties suitable for electric motor windings. Moreover, according to additive manufacturing possibilities, conceptual winding models were developed and subsequently were fabricated in a laboratory setting using electron beam melting. The investigation revealed a notable correlation between the achievable wall thickness and the focus offset. A focused electron beam, which was achieved with a focus offset of 17 mA, allowed for the production of thinner walls with reduced surface roughness. Additionally, the study highlighted that the density of the manufactured parts decreased as the hatch offset increased, particularly posing a more critical impact on smaller cross-sectional areas than larger ones. Furthermore, implementing a double-scanned contour for the hatch contributed to diminishing roughness while simultaneously increasing the maximum density to 99.6 %. Influencing the electrical conductivity of printed copper samples through heat treatment was successfully demonstrated, resulting in a conductivity of 91.7% ± 1.8% IACS post-heat treatment. / Inom ramen för denna avhandling undersöktes elektronstrålesmältningen av ren koppar i syfte att producera elektriska lindningar. Den konventionella tillverkningsprocessen av kopparlindningar undersöktes, och de potentiella fördelarna som erbjuds av designfriheten som är inneboende i elektronstrålesmältningsprocessen undersöktes. En omfattande parameterstudie genomfördes för att optimera befintliga produktionsparametrar för elektronstrålesmältning av ren koppar, med särskilt fokus på att uppnå önskade materialegenskaper lämpliga för elmotorlindningar. Dessutom, i enlighet med additiva tillverkningsmöjligheter, utvecklades konceptuella lindningsmodeller och tillverkades därefter i laboratoriemiljö med användning av elektronstrålesmältning. Undersökningen avslöjade en anmärkningsvärd korrelation mellan den möjliga väggtjockleken och fokusförskjutningen. En fokuserad elektronstråle, som uppnåddes med en fokusförskjutning på 17 mA, möjliggjorde produktion av tunnare väggar med minskad ytjämnhet. Dessutom visade studien att densiteten hos de tillverkade delarna minskade när avståndet mellan smältlinjerna ökade, vilket i synnerhet innebar en mer kritisk påverkan på mindre tvärsnittsareor än större. Dessutom bidrog implementeringen av en dubbelskannad kontur för luckan till att minska grovheten samtidigt som den maximala densiteten ökade till 99,6 %. Att påverka den elektriska ledningsförmågan hos tryckta kopparprover genom värmebehandling demonstrerades framgångsrikt, vilket resulterade i en ledningsförmåga på 91,7 % ± 1,8 % IACS efter värmebehandling.

MULTILEVEL inverters are the preferred choice of converters for electronic power conversion for high power applications. They are gaining popularity in variety of industrial applications including electric motor drives, transportation, energy management, transmission and distribution of power. A large portion of energy conversion systems comprises of multilevel inverter fed induction motor drives. The multilevel inverters are ideal for such applications, since the switching frequency of the devices can be kept low. In conventional two level inverters, to get nearly sinusoidal phase current waveform, the switching frequency of the inverter is increased and the harmonics in the currents are pushed higher in the frequency spectrum to reduce the size and cost of the filters. But higher switching frequency has its own drawbacks – in particular for medium voltage, high power applications. They cause large dv_/ dt stresses on the motor terminals and the switching devices, leading to increased electromagnetic interference (EMI) problems and higher switching losses. Harmonics in the motor currents can further be minimized by adopting dodecagonal voltage space vector (SV) switching (12-sided polygon). In case of dodecagonal SV switching, the fifth and seventh order (6n , 1, n = odd) harmonics are completely eliminated for the full modulation range including over modulation and twelve step operation in the motor phase voltages and currents. In addition to low order harmonic current suppression, the linear modulation range for dodecagonal SV switching is also more by 6% when compared to that of the conventional hexagonal SV switching. The dodecagonal voltage SV structure is made possible by connecting two inverters with DC-link voltages Vd and 0:366Vd on either side of an open-end winding induction motor. The dodecagonal space vector switching can be used to produce better quality phase voltage and current waveforms and overcome the problem of low order fifth and seventh harmonic currents and to improve the range for linear modulation while reducing the switching frequency of the inverters when compared to that of the conventional hexagonal space vector based inverters. This thesis focuses on three aspects of multilevel dodecagonal space vector structures (i) Two new power circuit topologies that generate a multilevel dodecagonal voltage space vector structure with symmetric triangles, (ii) A multilevel dodecagonal SV structure with nineteen concentric dodecagons, (iii) Pulse width modulation (PWM) timing calculation methods for a general N-level dodecagonal SV structure. (i) Two new power circuit topologies capable of generating multilevel dodecagonal voltage space vector structure with symmetric triangles with minimum number of DC link power supplies and floating capacitor H-bridges are proposed. The first power topology is composed of two hybrid cascaded five level inverters connected to either side of an open end winding induction machine. Each inverter consists of a three level neutral point clamped (NPC) inverter, cascaded with an isolated capacitor fed H-bridge making it a five level inverter. The second topology is a hybrid topology for a normal induction motor (star or delta connected), where the power is fed to the motor only from one side. The proposed scheme retains all the advantages of multilevel topologies as well the advantages of the dodecagonal voltage space vector structure. Both topologies have inherent capacitor balancing for floating H-bridges for all modulation indices including transient operations. The proposed topologies do not require any pre-charging circuitry for startup. PWM timing calculation method for space vector modulation is also explored in this chapter. Due to the symmetric arrangement of congruent triangles within the voltage space vector structure, the timing computation requires only the sampled reference values and does not require any iterative searching, off-line computation, look-up tables or angle estimation. Experimental results for steady state operation and transient operation are also presented to validate the proposed concept. (ii) A multilevel dodecagonal voltage space vector structure with nineteen concentric do-decagons is proposed for the first time. This space vector structure is achieved by connecting two sets of asymmetric hybrid five level inverters on either side of an open-end winding induction motor. The dodecagonal structure is made possible by proper selection of DC-link voltages and switching states of the inverters. In addition to that, a generic and simple method for calculation of PWM timings using only sampled reference values (v and v ) is proposed. This enables the scheme to be used for any closed loop application like vector control. Also, a new switching technique is proposed which ensures minimum switching while eliminating the fifth and seventh order harmonics and suppressing the eleventh and thirteenth harmonics, eliminating the need for bulky filters. The motor phase voltage is a 24-stepped waveform for the entire modulation range thereby reducing the number of switchings of the individual inverter modules. Experimental results for steady state operation, transient operation including start-up have been presented and the results of Fast Fourier Transform (FFT) analysis is also presented for validating the proposed concept. (iii) A method to obtain PWM timings for a general N-level dodecagonal voltage space vector structure using only sampled reference values is proposed. Typical methods that are used to find PWM timings for dodecagonal SV structures use modulation index and the reference vector angle, to get the timings T1 and T2 using trigonometric calculations. This method requires look-up tables and is difficult to implement in closed loop systems. The proposed method requires only two additions to compute these timings. For multilevel case, typical iterative methods need timing calculations (matrix multiplications) to be performed for each triangle. The proposed method is generic and can be extended to any number of levels with symmetric structures and does not require any iterative searching for locating the triangle in which the tip of the reference vector lies. The algorithm outputs the triangle number and the PWM timing values of T0, T1 and T2 which can be set as the compare values for any carrier based PWM module to obtain space vector PWM like switching sequences. Simulation and experimental results for steady state and transient conditions have been presented to validate the proposed method. A 3.7 kW, 415 V, 50 Hz, 4-pole open-end winding induction motor was used for the experimental studies. The semiconductor switches that were used to realize the power circuit for the experiment were 75 A, 1200 V insulated-gate bipolar transistor (IGBT) half-bridge modules (SKM75GB12T4). Opto-isolated gate drivers with desaturation protection (M57962L) were used to drive the IGBTs. For the speed control and PWM timing computation a digital signal processor (DSP-TMS320F28335) with a clock frequency of 150 MHz was used. For modulation frequencies 10 Hz and below, a constant sampling frequency of 1 kHz was used as the frequency modulation ratio is high. For modulation frequencies above 10 Hz, synchronous PWM strategy was used. The time duration Ts is the sampling interval for which the timings T1 , T2 and T0 are calculated. As in the case of any synchronous PWM method, the duration of sampling time (Ts) is a function of the fundamental frequency of the modulating signal. In this case, Ts = 1_.fm • 12n) sec. where fm is fundamental frequency in Hertz and ‘n’ is the number of samples per 30ý dodecagonal sector. The PWM timings and the triangle data (from the DSP) is fed to field programmable gate array (FPGA) (SPARTAN XC3S200) clocked at 50 MHz where the actual gating pulses are generated. The capacitor balancing algorithm and the dead-time modules were implemented within FPGA. No external hardware was used for generation of dead-time. The dead-time block generates a constant dead-time of 2 s for all the switches. Extensive testing was done for steady state operations and transient operations including quick acceleration and start-up to validate the proposed concepts. With the advantages like extension of linear modulation range, elimination of fifth and seventh harmonics in phase voltages and currents for the full modulation range, suppression of eleventh and thirteenth harmonics in phase voltages and currents, reduced device voltage ratings, lesser dv_dt stresses on devices and motor phase windings, lower switching frequency, inherent cascaded H-bridge (CHB) capacitor balancing, the proposed space vector structures, the inverter power circuit topologies, the switching techniques and the PWM timing calculation methods can be considered as viable schemes for medium voltage, high power motor drive applications.