HAL-INSA Toulouse
Not a member yet
34325 research outputs found
Sort by
Towards optimal reconfigurable constant multipliers
No full textInternational audienceThis paper introduces a novel algorithm for generating run-time reconfigurable single constant multipliers (RSCMs) which are optimal within their model in terms of hardware cost. Optimality is ensured by an exhaustive exploration of the design space mixing constraint programming, depth-first search, and branch-and-prune techniques. The cost model of previous works is also refined. Compared to the state of the art, this approach enables much larger constant sets, and also significantly improves the performance of the resulting architectures. Applications to neural network inference and small floating point multiplication units are evaluated
Average current mode control for GaN-based single-phase 2-level DAB converter
No full textInternational audienceDual Active Bridge (DAB) converters are widely employed in electric vehicles charging equipment, both in Onboard Chargers (OBCs) and offboard charging infrastructure, either in Grid-to-Vehicle (G2V) or Vehicle-to-Grid (V2G) power flow. However, due to the unavoidable mismatches in device parameters, a DC bias current can bring the transformer core to saturation, compromising the converter performance. Developing a means with a dedicated current control to eliminate the parasite DC current part is therefore essential for the structure to avoid operating points in magnetic saturated regions. This work aims, through MATLAB/Simulink simulations, to design a current mode control law with its low steady-state error reducing the DC parasite current. Operating in high switching frequency, GaN or SiC devices could represent today the best option available on the market for enhancing power density, efficiency and dynamic performance of the converter
Stability analysis and design of an event-triggered control scheme for a coupled ODE-heat PDE system. ⋆
No full textInternational audienceThis paper tackles the stability analysis of a system driven by linear ordinary differential equation coupled to a one-dimensional heat equation. The particularity of this study concerns the nature of the coupling, which is performed using an event-triggered scheme, through a boundary condition of the partial differential equation. After proving the existence and regularity of solutions of the system, the idea is to introduce an enriched energy functional as a candidate Lyapunov functional to prove the exponential stability. Actually, we will obtain a sufficient stability condition expressed as a linear matrix inequality to satisfy. This condition is suitable for solving an emulation problem corresponding to the appropriate tuning of the event-triggered control. Our result is finally illustrated with a numerical example
Mesures µ-OBIC sur diode bipolaire Diamant : une première !
No full textInternational audienceCette communication décrit pour la première fois la réalisation d'une diode bipolaire verticale en diamant qui présente des caractéristiques I(V) en direct et en inverse, pouvant donner lieu à des mesures micro-OBIC. L'OBIC est une technique de caractérisation spécifique permettant d'obtenir une image du champ électrique au sein d'un composant et ainsi d'analyser l'efficacité de ses protections périphériques
Direct Numerical Simulation of an aluminum particle combustion in oxidizing flow, with gas and condensed-phase reactions
No full textInternational audienceAluminum (Al) particles are promising fuels for propulsion and energy conversion, sparking decades of research into their combustion. When micron-sized aluminum burns in an oxidizing environment, the process unfolds in two stages: an initial steady, symmetric vapor-phase combustion, followed by unsteady, asymmetric combustion as alumina decomposes, releasing gaseous sub-oxides. These sub-oxides condense and dissolve in the liquid droplet, forming liquid aluminum oxide. However, the multiphase mechanisms and reactions driving this transition remain poorly understood, which undermines the accuracy of the combustion model. To address this, a 3-dimensional Direct Numerical Simulation (DNS) approach based on Navier-Stokes equations is used to model single-particle Al combustion in air. Finally, this model underscores three critical aspects: (1) the development of an accurate and robust numerical framework to elucidate the complex physiochemical processes governing aluminum combustion, (2) detailed quantification of the role of sub-oxides condensation on the burning droplet and its subsequent influence on heterogeneous surface reactions and aluminum vaporization dynamics, and (3) an evaluation of existing vaporization laws (burn rates), contributing valuable insights for macroscopic combustion model refinement
Euronanolab Lithography Expert Group
No full textNational audienceThis presentation by the lithography expert group of the EuroNanoLab network details the actions carried out in 2024-2025, including the organization of online workshops on laser lithography, e-beam and mask fabrication, as well as a face-to-face meeting on ion lithography. It offers a comparison of resist processes with common masks and patterns, to characterize resolution, profile and ease of use. Staff exchanges are facilitated between platforms thanks to Erasmus+ and NFFA. A collaborative database on resits, tools and processes is under construction. The initiative encourages knowledge sharing and the collective resolution of technical problems. Finally, actions that can be transposed to Renatech can be proposed, such as cross-training and inter-platform comparison projects.Cette présentation du groupe d’experts lithographie du réseau EuroNanoLab détaille les actions menées en 2024-2025, notamment sur l'organisation de workshops en ligne sur la lithographie laser, e-beam et la fabrication de masques, ainsi qu’une réunion présentielle sur la lithographie ionique. Elle propose une comparaison des procédés de résines avec des masques et motifs communs, pour caractériser la résolution, le profil et la facilité d’utilisation. Des échanges de personnels sont facilités entre plateformes grâce à Erasmus+ et NFFA. Une base de données collaborative sur les résines, outils et procédés est en construction. L’initiative encourage le partage des connaissances et la résolution collective de problèmes techniques. Enfin, des actions transposables à Renatech peuvent être proposées, comme des formations croisées et des projets de comparaison inter-plateformes
The global convergence of stochastic gradient descent in non-convex landscapes: Sharp estimates via large deviations
No full textInternational audienceIn this paper, we examine the time it takes for stochastic gradient descent (SGD) to reach the global minimum of a general, non-convex loss function. We approach this question through the lens of randomly perturbed dynamical systems and large deviations theory, and we provide a tight characterization of the global convergence time of SGD via matching upper and lower bounds. These bounds are dominated by the most "costly" set of obstacles that the algorithm may need to overcome in order to reach a global minimizer from a given initialization, coupling in this way the global geometry of the underlying loss landscape with the statistics of the noise entering the process. Finally, motivated by applications to the training of deep neural networks, we also provide a series of refinements and extensions of our analysis for loss functions with shallow local minima
Neural Network-Based Estimation of Optical Feedback Factor in Self-Mixing Interferometry Systems across various feedback regimes
No full textInternational audienceSelf-mixing interferometry (SMI) system is a promising and low-cost approach for displacement measurements. The resolution and precision of displacement estimation depend on key parameters of the SMI system and, in particular, on the optical feedback factor (C). This article presents a new method for estimating this C factor based on an artificial neural network (ANN) that analyzes certain specific fringe shape features of SMI signals. Specifically, only six features of the SMI signal are taken into account as input to a compact neural network with one hidden neuron layer (14 neurons) by the estimation process, which reduces the computational cost and simplifies its deployment. This study evaluates the C-estimation capability of the method in various simulated and experimental scenarios, including variations in displacement amplitude and frequency, random displacements, noise levels, and speckle effects. These analyses show that the method works correctly for a moderate feedback regime ( 14.6 ), this error is less than 2% (for C<8 ). Comparisons with other methods show similar or better accuracy but with the advantage of being able to handle all displacement waveforms (sinusoidal or arbitrary) at low computational cost, which facilitates integration into a real-time embedded system
Form factor of prismatic particles for Small Angle Scattering analysis
No full textInternational audienceSince the morphology of nanoparticles directly influences many of their properties, accurately determining their shape is crucial for targeted applications. In this work, we focus on nanoprisms due to their widespread use and the limitations of direct imaging techniques in accurately describing their polygonal cross section. Specifically, we introduce a new tool for small-angle scattering (SAS) analysis of nanoprisms that requires minimal computation time compared with all-atom simulations and other form factor analyses. A key innovation in this work is the implementation of the Lebedev quadrature for isotropic averaging, which allows for accurate form factor calculations using few sampling points. This form factor model is developed for any n-sided prism and is compared with small-angle X-ray scattering and transmission electron microscopy experimental data for gold and/or silver nanoprisms (n = 3, 4, 5). For small sizes, the nanoprism form factor model is compared with the result obtained with the Debye equation from atomic coordinates, showing a very good agreement. We explore the effects of the aspect ratio and cross-sectional shape of the nanoprisms on the form factor curves and discuss the limitations of our approach. Overall, our method combines precise shape determination with rapid computation time, paving the way for detailed characterization of nanoprisms using SAS techniques, potentially even during their growth
Uncertainty quantification and metamodeling of multi-fidelity CFD computation of a heated fuel assembly
No full textThis study proposes to quantify the uncertainty in a CPU-time costlyComputational Fluid Dynamics (CFD) model used to evaluate local temperaturefield in the situation of blocked fuel assembly in a PWR transfertube. Several uncertain parameters are identified and a first uncertaintypropagation study is conducted on a low-fidelity (poorly refined) mesh forCPU cost issues. Then, using the concept of “support points”, an algorithmis employed to reduce the size of the initial design of experiments.A high-fidelity model (finer mesh, more CPU-time expensive) is then runon this small-size design of experiments. A metamodel was finally builton those high-fidelity results to propagate uncertainties and finely analyzethe results. The successful results that are obtained show that metamodelinghas the potential to overcome the issue of costly CPU-time CFDmodels in the near future. Despite good quantitative results, the mainpurpose of the present study remains the novel methodology that was setup for uncertainty propagation in CFD