The Flightpath 2050 European Union stringent regulations for aviation noise reduction, along with the new generation of Ultra-High-Bypass-Ratio turbofans to reduce fuel consumption, significantly challenge the scientific community to find unprecedented acoustic liner designs. The SALUTE H2020 project has taken up this challenge, by designing and testing a programmable metasurface made up of electroacoustic resonators. Each electroacoustic resonator is composed by a loudspeaker and four microphones in a compact design, allowing to synthesize tunable local impedance behaviours thanks to a current-driven control strategy. A steel wiremesh mounted onto a perforated plate allows to protect the elctromechanical devices from the aerodynamic disturbances. For the first time, such advanced liner concept has been tested in a scaled turbofan rig: the ECL-B3 PHARE-2 in the Laboratory of Fluid Mechanics and Acoustics of the Ecole Centrale of Lyon. The performances of the electroacoustic liner reported in this paper, correspond to three different regimes: 30%, 40% and 100% of the nominal engine speed. The electroacoustic technology demonstrated robustness faced with a realistic reproduction of actual turbofan conditions, as well as its tunability to target different frequency bandwidth, attaining good radiated noise reduction. The results reported in this experimental campaign open the doors for unprecedented liner designs, by exploiting the huge potentialities of programmable surfaces.

L’ensemble documentaire concerne le Centre international de Conférences de Genève (CICG), construit en 1973 par André et François Gaillard et Alberto Camenzind (architectes), dans le but d'accueillir les conférences des Organisations Internationales et des Organisations Non Gouvernementales. Le dossier contient principalement des plans d'ingénieurs, complétés par quelques plans d'architectes, et de la documentation technique produite par d’autres intervenants. Il s’agit essentiellement de copies héliographiques, parfois en plusieurs exemplaires. Ils fournissent un aperçu essentiel des aspects techniques et architecturaux du projet (génie civil, construction et infrastructure).

[Fonds] 0142 - STEINMANN Georges A., 195?-2000 [Série] 0142.01 - Projets et réalisations, 195?-2000 [Dossier] 0142.01.0001 - Centre international de Conférences, Genève (CICG), 1968-1972 [SDossier] 0142.01.0001/01 - CICG : Dalles, murs, armatures, coffrages, 1968-1972 [SDossier] 0142.01.0001/02 - CICG : Garage, route de Varembé, AELE, 1968-1972 [SDossier] 0142.01.0001/03 - CICG : Cuisine, bar, self-service, système sanitaire, 1968-1972 [SDossier] 0142.01.0001/04 - CICG : Banque, poste, cabines téléphoniques, 1968-1972 [SDossier] 0142.01.0001/05 - CICG : Grande salle, terrasses, bacs à fleurs, 1968-1972 [SDossier] 0142.01.0001/06 - CICG : Étage presse, cabines d'interprètes, 1968-1972 [SDossier] 0142.01.0001/07 - CICG : Soumissions, 1968-1972 [SDossier] 0142.01.0001/08 - CICG : Administration, correspondances, 1968-1972 [SDossier] 0142.01.0001/09 - CICG : Autres intervenants, 1968-1972

Abstract A recurrent neural network fitted to large electrophysiological datasets may help us understand the chain of cortical information transmission. In particular, successful network reconstruction methods should enable a model to predict the response to optogenetic perturbations. We test recurrent neural networks (RNNs) fitted to electrophysiological datasets on unseen optogenetic interventions, and measure that generic RNNs used predominantly in the field generalize poorly on these perturbations. Our alternative RNN model adds biologically informed inductive biases like structured connectivity of excitatory and inhibitory neurons and spiking neuron dynamics. We measure that some of the biological inductive biases can improve the model prediction under perturbation in a simulated dataset and a dataset recorded in mice in vivo. Furthermore, we show in simulations that gradients of the fitted RNN can predict the effect of micro-perturbations in the recorded circuits, and discuss potentials for measuring brain gradients or using gradient-targeted stimulation to bias an animal behavior.

Collaborative Machine Learning (CML) allows participants to jointly train a machine learning model while keeping their training data private. In many scenarios where CML is seen as the solution to privacy issues, such as health-related applications, safety is also a primary concern. To ensure that CML processes produce models that output correct and reliable decisions even in the presence of potentially untrusted participants, researchers propose to use robust aggregators to filter out malicious contributions that negatively influence the training process. In this paper, we prove that the two prevalent forms of robust aggregators in the literature cannot eliminate the risk of compromise without preventing learning: in order to learn from collaboration, participants must always accept the risk of being the subject of harmful adversarial manipulation. Therefore, these robust aggregators are unsuitable for high-stake applications such as health-related or autonomous driving in which errors can result in physical harm. We empirically demonstrate the correctness of our theoretical findings on a selection of existing robust aggregators and relevant applications, including end-to-end results where we show that using existing robust aggregators can lead to an adversary can cause incorrect medical diagnosis or can cause self-driving cars to miss turns.

Humanitarian organizations distribute aid to people affected by armed conflicts or natural disasters. Digitalization has the potential to increase the efficiency and fairness of aid-distribution systems, and recent work by Wang et al. has shown that these benefits are possible without creating privacy harms for aid recipients. However, their work only provides a solution for one particular aid-distribution scenario in which aid recipients receive a predefined set of goods. Yet, in many situations it is desirable to enable recipients to decide which items they need at each moment to satisfy their specific needs. We formalize these needs into functional, deployment, security, and privacy requirements, and design a privacy-preserving digital wallet for aid distribution. Our smart-card-based solution enables aid recipients to spend a predefined budget at different vendors to obtain the items that they need. We prove our solution's security and privacy properties, and show it is practical at scale.

Performance tuning is a pervasive challenge in science and engineering, involving optimization of expensive black-box functions to achieve desired outcomes. This thesis focuses on Bayesian optimization (BO) as a promising solution for addressing the complexities of performance tuning. It proposes a series of principled Bayesian optimization algorithms in scenarios with black-box constraints, prior structural knowledge, and human-in-the-loop feedback. Key contributions include: (1) a constrained BO algorithm with theoretical performance guarantees, applied to tune the performance of a building controller and reduce building electricity costs by up to 27% while maintaining comfort constraints; (2) a scalable distributed BO algorithm leveraging additive structures; (3) human-in-the-loop BO algorithms that optimizes human preferences; and (4) human-AI collaborative BO that exploits expert supervisory feedback to enhance the convergence speed. These advancements address practical challenges in performance tuning, expanding the applicability of Bayesian optimization to real-world problems.

In the modern world, soft robots are gaining importance with technological advances. Compared to rigid robots, their elasticity enables safer and more adaptable integration into various systems, making them suitable for wearable devices, biomedical engineering, and robotics. As demand increases for lightweight, energy efficient, and elastic materials, both academia and industry are focusing more on soft actuators. Among these, dielectric elastomer actuators (DEAs) stand out for their large deformation, low power consumption, silent operation, and fast response. However, most DEA materials require high voltages for actuation which limits their application areas.

This thesis aims to develop high dielectric permittivity elastomers with good mechanical performance for DEA applications, focusing on polysiloxanes and polyphosphazenes. Polysiloxanes are well-established, known for their flexibility, processability, and environmental stability. Polyphosphazenes, though newer to DEA research, are elastic, biocompatible, and chemically resistant. Both, however, have relatively low dielectric permittivity. To address this, chemical modifications with highly polar functional groups were carried out to enhance their dielectric properties.

In the first study, a single-layer DEA was developed using polyphosphazene. While polyphosphazenes are well known, their actuator applications is underexplored. Polydichlorophosphazene (PDCP) was synthesized via living cationic polymerization and modified to create a stable polymer. Trifluoroethoxy groups were introduced as polar substituents to increase permittivity, leveraging fluorine's electron withdrawing nature. The resulting polymer showed a dielectric permittivity of 5.65 at 1 MHz. Varying cross-linker concentrations allowed tuning of mechanical properties, and the best per-forming material achieved 5.8% actuation at 80 V µm-1 with excellent stability across frequencies.

In the second study, both single-layer and stack actuators were made using a novel polysiloxane. The backbone was functionalized with ethyl sulfone groups, while butane thiol groups were varied to tune the glass transition temperature. Several polymers were synthesized, and the best material combined high dielectric permittivity with low mechanical losses. An increase in ethyl sulfone content led to higher permittivity. The best performing sample showed a dielectric permittivity of 16.2 at 10 kHz and lateral actuation strain of 13% at 8.2 V µm-1. A stack actuator made from this material showed 3% strain at 14.5 V µm-1 and maintained stable performance over 2000 cycles at 1 Hz.

The third study focused on another modified polysiloxane, incorporating 3-mercaptosulfolane as a polar group. Due to its bulky structure, the resulting polymer was thermoplastic. To adjust the Tg, butane thiol groups were introduced in different ratios, and the optimal material was selected through electromechanical testing. The DEA achieved 5% strain at 1.00 kV and 7.34% at 1.20 kV. A stack actuator based on this material showed a thickness change of 49 µm, corresponding to 2.7% strain at 13.8 V µm-1.

Overall, this research demonstrates the potential of polysiloxane and polyphosphazene based elastomers as high performance materials for DEA applications. By addressing the limitations of traditional elastomers, this work contributes to the development of next generation soft actuators which offers improved efficiency, stability, and low voltage operation.

Non-invasive beam diagnostics are essential for operating modern high-intensity accelerators such as the Large Hadron Collider (LHC). Among these diagnostics, Schottky monitors stand out for their ability to measure tune, chromaticity, momentum spread, and emittance based on the intrinsic noise (Schottky noise) of the circulating beam. Yet, the accuracy of Schottky analysis can be compromised by collective effects - particularly beam coupling impedance - and by non-linear fields, such as those introduced by Landau octupoles.

This thesis presents a comprehensive theoretical and computational framework for Schottky diagnostics, addressing the effects of beam coupling impedance and octupolar non-linearities. It provides a rigorous analysis of the statistical properties of Schottky spectra, demonstrating that individual "instantaneous" spectra represent only a single realization of an underlying random process. We develop and validate a novel simulation method that computes the beam's spectral content efficiently for the LHC, overcoming the limitations of standard discrete Fourier transform approaches.

Building on this foundation, the influence of Landau octupoles is explored in detail. We derive new expressions that reveal how octupole-induced amplitude detuning modifies the betatron sidebands of the Schottky spectrum, and we confirm these findings through macro-particle simulations and experiments with LHC ion beams. Despite introducing distortions that complicate classical parameter-extraction formulas, the presence of octupoles can also mitigate coherent components, sometimes improving overall diagnostic capabilities.

The study concludes with an investigation into beam coupling impedance, demonstrated to significantly influence synchrotron and betatron tunes in high-intensity proton beams. Broad-band resonator models are integrated into macro-particle simulations and analytical formulas, highlighting alterations in the Schottky spectrum and enabling comparisons with measured spectra across varying bunch intensities at the LHC. Preliminary evidence suggests that parts of the LHC's longitudinal impedance model may be underestimated.

Despite widespread helmet usage in high-speed sports, the incidence of brain injuries continues to rise, indicating some deficiencies in current helmet technologies. Existing helmet safety standards only address linear impacts, failing to account adequately for rotational accelerations, a contributor to severe brain injuries such as traumatic axonal injury (TAI). These injuries induce healthcare challenges and highlight the need for enhanced helmet design capable of mitigating rotational forces to improve safety outcomes. This thesis aims to develop a solution that reduces the risk of brain injury upon impact and improves helmet capabilities to lower the risk of TAI. To reach this goal, this thesis was divided into four studies.

First, a biomechanical analysis was performed using a finite element model to simulate head impacts in an AI-driven approach. Deep learning models were trained to predict peak head kinematics based on material properties and impact energy. The optimization framework enabled the identification of ideal compressive stress-strain profiles tailored to specific impact energies, showing a potential TAI risk reduction of up to 65% compared to conventional expanded polystyrene foams (EPS).

In parallel with the biomechanical analysis, a database was created integrating current protective technologies such as EPS. To complete this database, novel materials were developed and characterized for their impact absorbing potential. A formulation of a hybrid iono-organogel was identified with low shear modulus and high dissipation capabilities. The gel was obtained by the co-polymerization of acrylic acid and DMAPS inside a hybrid solvent composed of an ionic liquid and an oligomer. Its network structure, featuring reversible hydrogen bonding and ion-dipole interactions, efficiently absorbs and dissipates energy under shear and compression during successive impacts. Loss factors exceeded 0.5 under various dynamic conditions. Additionally, six different bioinspired 2D lattice structures were produced via 3D printing and characterized under quasi-static and impact conditions. Surrogate modeling and multi-objective optimization were employed to identify designs and parameters that maximize energy absorption while minimizing buckling stress.

Finally, optimized stress-strain profiles for three target impact energies, derived from the biomechanical analysis, guided material selection for ski helmet liners. Experimental stress-strain curves were compared to ideal profiles using a scoring algorithm evaluating peak stress, specific energy absorption, and curve similarity. FEM simulations demonstrated that hexagonal and beetle-inspired lattices outperformed EPS by reducing rotational and linear acceleration by up to 36% and 63%, respectively, thereby lowering the risk of TAI for users. These materials were then combined into layered structures tailored for different impact severities and tested under compressive and 45° impacts, confirming the improved energy absorptions.

Overall, this thesis presents a transferable and integrative approach to materials design and selection, combining artificial intelligence, FEM, experimental mechanics and advanced manufacturing. The resulting methodology provides a generalizable framework for improving the performance of protective equipment. Though focused on ski helmets, the findings hold relevance for broader applications in sports, automotive safety and biomedical devices.

Urinary tract infections (UTIs) are common bacterial infection primarilly caused by Uropathogenic E. coli (UPEC). A major concern with UTIs their elevated rate of recurrence. The principal mechanisms for recurrence are believed to be reinfection from an endogenous reservoir or a persistent infection. Uropathogenic bacteria's capacity to invade the bladder tissue as intracellular bacterial communities are believed to be a major contributor in persistent infection. Recently, L-forms were found in urine samples from patients with recurrent UTIs suggesting that cell wall deficient bacteria may play a role in UTI recurrence. L-forms are proliferating bacteria with a deficient cell wall. Normally rod shapped bacteria can be induced into the L-form state by cell wall targetting bacteria. While the L-form state is fragile and requires osmotic support, L-forms are also refractory to killing by the inducing cell wall targetting antibiotic. Once the cell wall synthesis inhibitor is removed, L-forms can revert to their original rod-shape morphology and grow rapidly. Therefore, there is a clear meachnism outlining how L-forms could represent a form of persistent infection. Here, we investigated how L-forms could cause recurrence in the UTI context. We used synthetic human urine and UPEC as an axenic model for UTIs. We exposed UPEC to fosfomycin, a cell wall synthesis inhibitor that is recommended for treating UTIs. We found that the vast majority of cells which survived fosfomycin exposure were converted to L-forms before they regrew as rod-shapped bacteria. We identified that colanic acid synthesis is highly upregulated by UPEC during fosfomycin exposure. We found that the ability to produce colanic acid was not highly protective for fosfomycin treated UPEC. However, UPEC mutants lacking the colanic acid exporter Wza showed significantly decreased survival to fosfomycin. We demonstrate that the Wza is present in approximately 90% of UPEC genomes and represents a general target for reducing L-from survival in fosfomycin treat UTIs. Timelapse microscopy was essential to the findings detailed above. L-forms lack easily identifiable marker and are usually identified by visual analysis. Therefore, we developped an image analysis pipeline to segment UPEC L-forms. Furthermore, to identify if growth rate correlated with UPEC's ability to survive L-form induction we created TrackEcoli which created lineages for cells pre-antibiotic exposure. In this thesis, we provide insight on the expected frequency of L-form survivors in fosfomycin treated UTIs. Furthermore, we determined that inhibiting colanic acid exporter Wza is a generalized target which can help reduce UPEC L-form survival.