HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Uncertainty Quantification in Regression Neural Networks using Evidential Likelihood-based Inference
No full textInternational audienceWe introduce a new method for quantifying prediction uncertainty in regression neural networks using evidential likelihood-based inference. The method is based on the Gaussian approximation of the likelihood function and the linearization of the network output with respect to the weights. Prediction uncertainty is described by a random fuzzy set inducing a predictive belief function. Two models are considered: a simple one with constant conditional variance and a more complex one in which the conditional variance is predicted by an auxiliary neural network. Both models are trained by regularized log-likelihood maximization using a standard optimization algorithm. The postprocessing required for uncertainty quantification only consists of one computation and inversion of the Hessian matrix after convergence. Numerical experiments show that the approximations are quite accurate and that the method allows for conservative uncertainty-aware predictions
Evidential time-to-event prediction with calibrated uncertainty quantification
No full textInternational audienceTime-to-event analysis provides insights into clinical prognosis and treatment recommendations. However, this task is more challenging than standard regression problems due to the presence of censored observations. Additionally, the lack of confidence assessment, model robustness, and prediction calibration raises concerns about the reliability of predictions. To address these challenges, we propose an evidential regression model specifically designed for time-to-event prediction. Our approach computes a degree of belief for the event time occurring within a time interval, without any strict distribution assumption. Meanwhile, the proposed model quantifies both epistemic and aleatory uncertainties using Gaussian Random Fuzzy Numbers and belief functions, providing clinicians with uncertainty-aware survival time predictions. Experimental evaluations using simulated and real-world survival datasets highlight the potential of our approach for enhancing clinical decision-making in survival analysis
La gestion du vieillissement du parc de dispositifs médicaux en milieu hospitalier
No full textThis thesis deals with the management and renewal of medical devices in healthcare institutions, highlighting the financial challenges they face. According to Decree 2001-1154, hospitals must guarantee the maintenance and quality control of their medical equipment. However, with limited budgets, hospitals are struggling to renew their equipment fleet as they would like, especially since the price of medical devices is increasing due to the rise in raw materials, transport costs and supply difficulties. This leads to increasing obsolescence of equipment that can affect its performance and therefore the quality of care. The efficient management of medical equipment is therefore a crucial issue to ensure patient safety and optimize maintenance costs. An efficient renewal of equipment would not only make it possible to comply with safety standards, but also to save money while remaining at the cutting edge of technology and improving care. This approach includes bibliographic research, interviews with professionals in the sector, as well as a questionnaire intended for French and Belgian biomedical engineers to collect feedback. The objective is to develop a decision-making tool to facilitate the choice of equipment renewal in hospitals.Ce mémoire traite de la gestion et le renouvellement des dispositifs médicaux dans les établissements de santé en France, en mettant en évidence les défis financiers auxquels ils font face. Selon le décret 2001-1154, les hôpitaux doivent garantir la maintenance et le contrôle qualité de leurs équipements médicaux. Mais, avec des budgets limités, les hôpitaux ont du mal à renouveler comme ils le souhaiteraient leur parc d'équipements, d'autant plus que le prix des dispositifs médicaux augmente en raison de la hausse des matières premières, des coûts de transport et des difficultés d’approvisionnement. Cela entraîne une vétusté croissante des équipements pouvant altérer leur performance et donc la qualité des soins. La gestion efficiente des équipements médicaux est donc un enjeu crucial pour assurer la sécurité des patients et optimiser les coûts de maintenance. Un renouvellement efficace des équipements permettrait non seulement de respecter les normes de sécurité, mais aussi de réaliser des économies tout en restant à la pointe des technologies et d'améliorer les soins. Cette approche inclut des recherches bibliographiques, des entretiens avec des professionnels du secteur, ainsi qu’un questionnaire destiné aux ingénieurs biomédicaux, français et belges, pour recueillir des retours d'expérience. L’objectif est d’élaborer un outil d’aide à la décision pour faciliter les choix de renouvellement d’équipements dans les hôpitaux
Advances in the Reliability Analysis of Coherent Systems under Limited Data with Confidence Boxes
No full textInternational audienceThis paper proposes an uncertainty quantification framework that enables the analyst to compute statistically calibrated confidence bounds of the reliability of coherent systems, even in the case in which the data are limited. More specifically, we propose to use confidence boxes to do so. Such a proposal is motivated by the fact that confidence boxes offer a guarantee on statistical performances regardless of the amount of available data through repeated use, which is especially useful when considering the fact that reliability or failure data for the reliability analysis are often limited in availability. The aim of this work is to provide tools that allow the analyst to obtain the true confidence intervals over the system failure and reliability at any desired level. This paper first reviews the basics of confidence boxes and reliability analysis before providing general computation tools. From this, a mathematical formalism is presented that relates the component configurations with the corresponding Boolean logic expressions to perform a forward propagation of the confidence boxes under varying dependencies between the components. The feasibility of the proposed framework is then demonstrated through three case studies involving complex systems under varying engineering settings in the form of the (1) pressurized tank system, (2) Training, Research, Isotopes, General Atomics (TRIGA) nuclear research reactor cooling system, and (3) bridge structure system. Through case studies, the validity of our studies is empirically shown, and an evaluation of the strengths and limitations of the proposed framework is presented. Finally, this paper provides perspectives on the future research works that can be undertaken. To provide a better understanding of the proposed framework, open R source code to reproduce the results and perform other related studies is available on GitHub
Voltage generation for Sawyer-Tower Coss loss measurement based on resonant converters
No full textInternational audienceOutput capacitance losses in modern semiconductor devices become increasingly relevantwith fast switching power converters. A new voltage generation approach is proposed to handlethe highly variable capacitive load in a SawyerTower circuit. An LLC resonant converter isapplied to generate the high voltage, high frequency sinusoidal excitation voltage, overcominglimitations of HF amplifiers. An external DCvoltage source is introduced to prevent reverseconduction of the device under test, ensuringaccurate charge-voltage (Q-V) characterization.Measurements under this voltage excitation arepresented. The calculated dissipated energy isverified by thermal measurements observing thetemperature rise of the DUT, showing less than6 % deviation. This article presents a scalable andpractical setup for power semiconductor characterization at high voltages
A Feedback-Enhanced Approach to Modeling Semiconductor Ageing in Electric Vehicle Inverters
No full textInternational audienceWith the increasing adoption of electric vehicles as a sustainable solution to curb greenhouse gas emissions, the reliability and lifetime calculation of power electronic converters, particularly semiconductor components, in traction inverters, become paramount. Utilizing a mission profile based on the worldwide harmonized lightduty test cycle, this paper introduces an advanced approach to calculate semiconductor lifetime of an EV traction inverter. Employing a functional model to calculate power losses and device temperatures. An ageing model, incorporating power cycling capability data in a LESIT model, determines consumed lifetime. The feedback loop integrates the consumed lifetime into the functional model, degrading thermal properties in proportion allowing the changing system characteristics to be taken into account during calculation. This positive feedback loop accelerates ageing, highlighting the significant influence on the lifetime assessment compared to conventional implementations
A performance comparison between different PWM techniques for the Hybrid Inverter Drive
No full textInternational audienceVariable Speed Drives (VSDs) allow energy savings for loads where operation below the grid frequency may be required. However, when operating the motor load at the grid frequency, they introduce, compared to grid-connected operation, energy losses and Electromagnetic Interference (EMI). The Hybrid Inverter Drive (HID) offers maximum efficiency at either mode of operation. With the HID the motor load is operated in inverter-driven mode for motor frequencies below grid frequency and operated in direct grid-connected mode for operation at grid frequency. This eliminates inverter losses and EMI at the grid frequency. In previous work [5], the HID and its semi-closed transitions were introduced. A performance comparison of the HID operation using pure Sinusoidal PWM (SPWM), Space Vector PWM (SVPWM) and the injection of the DC-link mid-point voltage into the SPWM reference signals during semi-closed transitions is presented in this paper. The HID is modelled in MATLAB Simulink
Allocation of Electric Vehicle Charging Stations in Urban Areas Using GIS
No full textInternational audienceThe current EV charging points can be inadequately equipped to meet the rapidly increasing battery charging demand of EVs. Therefore, in this paper, a Geographic Information System (GIS)-based Multiplecriteria decision analysis (MCDA) approach is adopted to help providing a solution of charging points location problem. Furthermore, the Analytical Hierarchy Process (AHP) and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) are used to select the optimal charging station site. A five-step approach is developed: determination of 14 criteria across four perspectives, spatial analysis using QGIS, criteria prioritization with AHP, site ranking in use of TOPSIS, and site capacity estimation. By integrating these methodologies, optimal charging point locations can be identified while ensuring effective deployment and utilization of EV charging infrastructure. Through a complete and considerate analysis, the study identifies the three most appropriate alternative locations for EVs including photovoltaic power
Multiple generator's stator windings as voltage sources for a multilevel converter
No full textInternational audienceIt is proposed to use a generator's stator windings to create multiple, isolated voltage sources for multilevel converters (MC). Under this approach the generator windings are viewed as transformer windings, with multiple secondary windings. Thus, we propose to redesign the stator windings as multiple transformers hence turning a single 3phase output multiple 3-phase outputs. After rectifying the outputs, they function as separate isolated DC voltage sources, which are applied to an MC's input. Taking advantage of the readily existing generator's magnetics, no additional magnetic components are required under the proposed method. The proposed method was validated experimentally
HVDC and Power Electronics enabling the energy transition
No full textInternational audienceTo understand the role of HVDC and Power Electronics in enabling the energy transition. It is necessary to define the meaning of the term energy transition. By definition "Energy transition is the process of shifting the world's energy production and consumption from fossil fuels to renewable energy sources to reduce greenhouse gas emissions and climate change" So it has the two components of the switch on the energy production side and the consumption side from fossil-based sources to renewable sources. In this presentation we are going to address how HVDC and power electronics are creating the right field to enable this transition on the generation side. It is interesting that on the generation side, the energy transition is also referred to as renewable energy transition. The key word here is renewable. This defines the type of generation and hence the associated transmission. HVDC and Power Electronic equipment are enabler of this transition. The presentation will develop from the stand point why HVDC, the technology and concepts, and the recent developments. It is very interesting that because of the desire to keep expanding the HVDC system, modular grid expansion is being addressed. So, for example for an offshore wind integration project the HVDC becomes part of an integrated grid