21549 research outputs found
Sort by
Network flow and flood routing model for water resources optimization
Full text linkReal-time management of hydraulic systems composed of multi-reservoir involves conflicting objectives. Its representation requires complex variables to consider all the systems dynamics. Interfacing simulation model with optimization algorithm permits to integrate flow routing into reservoir operation decisions and consists in solving separately hydraulic and operational constraints, but it requires that the water resource management model is based on an evolutionary algorithm. Considering channel routing in optimization algorithm can be done using conceptual models such as the Muskingum model. However, the structure of algorithms based on a network flow approach, inhibits the integration of the Muskingum model in the approach formulation. In this work, a flood routing model, corresponding to a singular form of the Muskingum model, constructed as a network flow is proposed and integrated into the water management optimization. A genetic algorithm is involved for the calibration of the model. The proposed flood routing model was applied on the standard Wilson test and on a 40 km reach of the Arrats river (southwest of France). The results were compared with the results of the Muskingum model. Finally, operational results for a water resource management system including this model are illustrated on a rainfall event
Expert-guided Symmetry Detection in Markov Decision Processes
No full textLearning a Markov Decision Process (MDP) from a fixed batch of trajectories is a non-trivial task whose outcome’s quality depends on both the amount and the diversity of the sampled regions of the state-action space. Yet, many MDPs are endowed with invariant reward and transition functions with respect to some transformations of the current state and action. Being able to detect and exploit these structures could benefit not only the learning of the MDP but also the computation of its subsequent optimal control policy. In this work we propose a paradigm, based on Density Estimation methods, that aims to detect the presence of some
already supposed transformations of the state-action space for which the MDP dynamics is invariant. We tested the proposed approach in a discrete toroidal grid environment and in two notorious environments of OpenAI’s Gym Learning Suite. The results demonstrate that the model distributional shift is reduced when the dataset is augmented with the data obtained by using the detected symmetries, allowing for a more thorough and data-efficient learning of the transition functions
Coaxial-Injector Surrogate Modeling Based on Reynolds-Averaged Navier–Stokes Simulations Using Deep Learning
Full text linkFacing the need to increase the accuracy of rocket engine design tools, the present work introduces an innovative methodology for the design and optimization of rocket engine combustion chambers using numerical simulations and deep learning. An experimental test case of a single coaxial injector is taken as a reference point, and a design of experiments is generated by varying nine parameters (geometrical and operative conditions). Reynolds-averaged Navier–Stokes simulations are carried out to generate the data set. The data are used to train surrogate models of different fidelities, from low-dimensional outputs (zero-dimensional and one-dimensional) toward the two-dimensional temperature field. Attention is given on the selection of the proper machine learning technique. For low-dimensional outputs, results show that deep neural networks outperform other standard machine learning tools, namely, radial basis function and kriging. Regarding high-dimensional outputs, convolutional neural networks with gradient-based loss functions are found effective to capture the large and smooth temperature variations, as well as the thin and sharp temperature gradients at the flame front. Eventually, the models are used in the framework of an optimization problem. Results highlight the benefits of new design and optimization tools based on deep learning, capable of real-time predictions of complex flowfields
Formal Monotony Analysis of Neural Networks with Mixed Inputs: An asset for certification
Full text linkThe use of ML technology to design safety-critical systems requires a complete understanding of the neural network's properties.
Among the relevant properties in an industrial context, the verification of partial monotony may become mandatory. This paper proposes a method to evaluate the monotony property using a Mixed Integer Linear Programming (MILP) solver. Contrary to the existing literature, this monotony analysis provides a lower and upper bound of the space volume where the property does not hold, that we denote ``Non-Monotonic Space Coverage''. This work has several advantages: (i) our formulation of the monotony property works on discrete inputs, (ii) the iterative nature of our algorithm allows for refining the analysis as needed, and (iii) from an industrial point of view, the results of this evaluation are valuable to the aeronautical domain where it can support the certification demonstration. We applied this method to an avionic case study (braking distance estimation using a neural network) where the verification of the monotony property is of paramount interest from a safety perspective
Effect of anodization and loading on fatigue life of 2618- T851 aluminum alloy
Full text linkAnodizing process is largely performed on
aluminum alloys to enhance corrosion and wear
resistances. However, this process can significantly
reduce the fatigue resistance of aluminum alloys
depending on the alloy and the process parameters.
This paper deals with the influence of anodizing
process on fatigue resistance of the aeronautical
aluminum alloy 2618-T851. The anodic film has been
characterized after each step using scanning electron
microscopy (SEM). This analysis showed that the
anodic layer was crazed just after impregnation
operation with micro-cracks extending from the
surface up to half the anodic layer. Simultaneously,
uniaxial fatigue tests under tension or torsion loadings
have been conducted in order to evaluate the effect of
each process step in relation with loading nature. The
results showed that the anodizing step has a detrimental
effect on fatigue life of 2618 alloy when specimens
are subjected to tensile loading. However, anodizing
process has no effect on fatigue life under torsion
loading. Fractographic observations have been conducted
in order to get an insight on the mechanisms
involved in each case
Simulations of wind turbine wakes in realistic atmospheric conditions: from large eddy simulations to analytical models.
Full text linkWind turbines are often clustered in wind farms where they are subject to the wakes emitted by the upstream rotors. Wind turbine wakes are regions of decreased wind velocity and increased turbulence (velocity variations), that respectively lead to a loss of production and a decreased life expectancy, in the end increasing the cost of wind power. The wind farm layouts must thus be optimised by taking wakes into account to make the wind power more efficient. Since optimisation studies require a lot of simulations, steady analytical models have been developed to quickly provide an estimation of the velocity and turbulence in the wake of a wind turbine, and compute the effect on the downstream rotor. Moreover, wind turbine wakes interact with the atmospheric boundary layer (ABL). Depending on the atmospheric stability, large-scale turbulent motions are created in the ABL that induce lowfrequency displacements of the wake. This phenomenon, called meandering, modi�es the mean wake properties but is rarely taken into account explicitly in steady analytical models. The present work aims at better understanding and modelling the interactions between wind turbine wakes and the ABL, based on high-fidelity numerical simulations. The first part of the manuscript is dedicated to the state-of-the-art. It starts by describing the general fluid dynamic and meteorological concepts on which is based the whole work. Then, we propose a literature review of the wakes and their interactions with the ABL. We also describe the existing steady analytical models and the models for wake meandering. This part ends with a description of our high-fidelity code, Meso-NH, as well as the chosen parametrisation for the turbine. In the second part, we focus on the high-fidelity simulations. A validation of Meso-NH against measurements and other equivalent codes of the community is performed, for three cases of atmospheric stability: neutral, stable and unstable. After a comparative study of different post-processing methods, an in-depth physical analysis of the wakes under these three conditions is proposed. We concluded that atmospheric stability mainly affects the wake meandering, and not the wake expansion which is attributed to the operating conditions. The third part is based on this result. We separate the turbulence and velocity fields into dfferent terms that can be associated to meandering, wake expansion, or both. These terms are compared for the different cases and a model is proposed for the preponderant ones. It results in a new steady analytical model for velocity and turbulence that independently takes into account wake expansion and meandering. This thesis thus presents and analyses new results that enhance our understanding of the behaviour of wakes in the ABL, and proposes a new steady model that allows taking into account wake meandering
Projections régionales haute-résolution spatiale du niveau de la mer sur les côtes d'Europe de l'Ouest sur le 21ème siècle
Full text linkL'élévation du niveau de la mer induite par le changement climatique augmente la fréquence des niveaux marins extrêmes. Les projections du niveau de la mer sont donc d’intérêt mais sont généralement produites avec des modèles de climat globaux qui ont une résolution spatiale grossière. De plus, ils ne résolvent pas certains processus clés responsables des variations du niveau de la mer à la côte comme les vagues, les marées et les surcotes de tempête, ni leurs interactions. L’objectif de cette thèse est de discuter de pratiques méthodologiques à suivre pour régionaliser des projections de niveau de la mer, notamment des évènements extrêmes, et de quantifier les incertitudes associées à ces choix méthodologiques. Dans cette thèse, des projections régionales de niveau de la mer sont réalisées sur les côtes d’Europe de l’Ouest sur la période 1950 à 2100 pour deux scénarios de changement climatique (SSP1-2.6 et SSP5-8.5). Pour cela, une descente d’échelle dynamique d’un modèle de climat global CMIP6, le modèle CNRM-CM6-1-HR, est réalisée avec la mise en place d’un modèle océanique régional au 1/12° (IBI-CCS) résolvant explicitement les marées et les surcotes de tempête en plus de la circulation générale océanique. Des corrections de biais sont appliquées aux forçages globaux et la hausse du niveau de la mer moyen du modèle de climat est imposée comme forçage. De plus, pour inclure la surcote des vagues aux projections du niveau d’eau à la côte, une descente d’échelle dynamique d'un modèle global de vagues est effectuée. Les interactions océan-vagues sont prises en compte en forçant le modèle régional de vagues par des sorties horaires de courants de surface et de niveau de la mer du modèle océanique régional IBI-CCS. Cette méthodologie permet donc d’inclure, en plus de tous ces processus côtiers, une grande partie de leurs interactions. A partir des sorties horaires des simulations régionales océan et vagues produites, des EVA (Extreme Value Analyses) sont effectuées pour analyser les extrêmes sur toute la période. Dans un premier temps, nous montrons que l’impact de la descente d’échelle dynamique et des corrections sur les projections de niveau de la mer est faible sur les côtes, dû à un modèle forceur déjà relativement haute résolution (1/4° pour l’océan) par rapport aux résolutions typiques des modèles de climat (environ 1°). Ensuite, l’importance d’inclure la surcote des vagues dans le niveau d’eau à la côte est mise en évidence par le fait que cela rallonge jusqu’à 30 ans l'année où l'événement centennal historique devient annuel sur les côtes atlantiques européennes et méditerranéennes. Ces résultats dépendent toutefois de la chaîne de modélisation mise en place, de la méthode d’analyse appliquée, et de la région d’étude. Par ailleurs, nous montrons que la prise en compte des interactions non-linéaires vagues-niveau de la mer influence les extrêmes de hauteurs significatives des vagues dans les zones macro-tidales. Plus particulièrement, la prise en compte de ces interactions non-linéaires peut causer des différences sur le flux d’énergie des vagues (utilisé comme proxy pour l’érosion côtière événementielle) allant jusqu’à +80% en 2100. Cette thèse a permis de développer une approche méthodologique qui servira par la suite à la production d’ensembles de projections de niveau de la mer, afin de caractériser les incertitudes associées à ces projections
Towards an efficient cost function equation for DDR SDRAM interference analysis on heterogeneous MPSoCs
Full text linkReal-time applications must finish their execution within an imposed deadline to function correctly. DDR memory interference on multicore platforms can make tasks overpass
their respective deadline, leading to critical errors. Bandwidth regulators and SDRAM bank partitioning are examples of techniques used to mitigate or avoid this interference type. Another possibility is to optimally place tasks and memory on the platform, i.e., task/memory mapping optimization. The algorithms used for finding optimal mapping solutions work using a cost function that indicates the fitness of the found solution. In this work, we propose a DDR SDRAM cost function that estimates the worst-case execution time for a giving map, and hence, implementable in an optimization algorithm. Our cost function considers the DDR memory device operation, the SoC manufacturer memory controller, the heterogeneity of the platform and the characteristics of the tasks to map. The cost function is evaluated by measuring directly the interference from the heterogeneous MPSoCs Keystone II and Sitara AM5728 by Texas Instruments
Assessment of Large Eddy Simulation in the Conjugate Heat Transfer context for engine operability : application to Hydrogen enrichment and Spinning Combustion Technology
Full text linkAs society evolves towards a green economy to face climate change, the combustion community is expected to develop new technologies and design low emission combustors for the aviation and energy sectors. In that respect, hydrogen is today a promising technical solution since it offers no direct CO2 production and even when it is mixed with classical fossil fuels it helps the stabilization of leaner and greener flames. However, the development of H2 combustion chamber is a technological challenge raising multiple questions in terms of reliability, efficiency and safety, especially for airplanes. When it comes to helicopter engines, there exists no specific pollutant emission regulation as of now and, due to their low power, helicopters are ideal testbed for new technologies. More specifically and to illustrate this ideal development context, Safran Helicopter Engines (SHE) has recently developed the Spinning Combustion Technology (SCT) gaining in engine operability and lean blow-out (LBO) capabilities. Due to its large potential in predicting complex reactive flows, Large Eddy Simulation (LES) has proven useful to support this design challenge, whether it is oriented toward a change in fuel (H2) or a change in combustor geometry (SCT). However, since engine operability is a very fine phenomenon given its multi-physics nature, large efforts and attention should be paid on the proper modeling of the different physics coexisting in these systems. In this work, a full assessment of Conjugate Heat Transfer based high-fidelity LES models is proposed and organized in three parts. First, main modeling challenges are addressed. As H2-enrichment and real engine conditions yield reduced flame thickness and more stringent requirements in terms of domain discretization, a physics-based Static Mesh Refinement (SMR) approach is derived and validated on different configurations. In parallel and since real flow prediction will depend on the applied thermal boundaries, Conjugate Heat Transfer (CHT) based LES simulations are validated and assessed compared to simpler strategies for a partially premixed swirled flame, the right dynamics being correctly predicted only with a correct estimation of the heat transfer at the walls. Finally, the effect of variable transport properties, typical of H2 mixture flows, on a swirled premixed flame is analyzed, confirming that a proper description of the chemistry and transport properties are needed when dealing with notconventional fuel mixtures.Second, the effects of H2-enrichment and elevated pressure (up to 5 bar) are investigated for a swirled CH4 flame. Both drastic changes on the flame shape and its dynamics are observed, eventually triggering thermoacoustic oscillations. Third, the flame stabilization and the LBO dynamics in the SCT are specifically studied. CHT-LES is able to retrieve the experimentally observed dynamics when decreasing the equivalence ratio and provides better results than typical adiabatic simulations. To finish, LES is used as an industrial tool to design a new burner closer to real SCT engines. By addressing these challenges, this work demonstrates the assessment of LES, in a CHT context, for predicting engine operability when dealing with innovative technologies and therefore highlights the central role of High Power Computing (HPC) and high-fidelity LES in the transition towards a decarbonized future
Vers une recherche reproductible en optimisation topologique des aérostructures
No full textCet article cherche à démontrer notre contribution en recherche reproductible pour l’op- timisation topologique par les méthodes d’optimisation topologique intitulées “Generalized Geometry Projection” (GGP) et “efficient multi-scale topology optimization” (EMTO). La première méthode uni- fie les méthodes existantes en optimisation topologique explicite; la seconde offre un environnement d’optimisation topologique multi-échelle. Leur point commun est d’être basées sur deux articles livrés avec le code associé top99 et top88 (initiative du DTU). La comparaison de ces méthodes est faite sur le cas test d’une nervure d’aile d’avion (2D). Les résultats mettent en évidence leur complémentarité : l’une donnant un concept simple d’assemblage d’éléments structuraux, l’autre une conception encore plus performante via des cellules micro-architecturée dont la fabrication doit être automatisé