Université de technologie de Troyes open archive
Not a member yet
10722 research outputs found
Sort by
Sustainable recyclable waste collection routing for waste pickers using antcolony algorithm
No full textFORENSICS ANALYSIS OF RESIDUAL NOISE TEXTURE IN DIGITAL IMAGES FOR DETECTION OF DEEPFAKE
No full textInternational audienceThis paper proposes an original approach for the automatic detection of AI-generated images, using features derived from noise residuals artefacts. Contrary to most current research that leverages sophisticated deep learning models to further improve performance, this study highlights the distinct noise residual characteristics in deepfakes, facilitating the identification of AI-generative images. Our findings highlight some limitations of image models, which can be used for forensic analysis and for future AI-based text-to-image generative models. Broad numerical results on a large and diverse dataset show the interest of the identified features as well as the relevance of the present method.</div
Effect of sputtering geometry and processing parameters on microstructure and texture developments in reactive magnetron co-sputtered (Ti,Al)N films
No full textInternational audienceIn view of creating improved hydrogen permeation barriers, co-sputtering of titanium aluminium nitride (Ti,Al)N was carried out - using two PVD magnetron sputtering machines with different geometries - to gain more insights concerning the effect of processing conditions on the microstructure and texture developments in the deposited film. Three types of substrates: amorphous glass SiO2, monocrystalline 〈100〉 Si wafers and ferritic API 5 L X65 grade steel were used. After, a deposition of a Ti adhesion layer was performed. The nature of the initial substrate had an effect on both the texture and the columnar grain size of the (TiAl)N film. Such effect is suggested to be inherited from differences in this adhesion Ti layer. When deposited under the collinear geometry condition, the (Ti,Al)N films were characterized by columnar grains having a rather flat surface and creating rather weak 〈111〉//ND fibre textures. For the orthogonal geometry conditions, the columnar oriented (Ti,Al)N grains had a triangular surface morphology, truncated by {100} planes. These columnar grains initially displayed randomly oriented {100} facets in the (RD, TD) plane, i.e. //ND fibre texture at low thicknesses. Geometrically driven competitive growth led to the selection of “well-aligned” columnar grains, resulting in a texture characterized by a narrow distribution around a single orientation and much higher texture index J. The columnar (Ti,Al)N grains were tilted away from ND by an angle β when varying the sample height in the OC chamber, causing a δ angle. Such dependence is described by the material related generalized relationship for E = 0.26
Chebyshev polynomials involved in the Householder's method for square roots
No full textInternational audienceThe Householder's method is a root-find algorithm which is a natural extension of the methods of Newton and Halley. The current paper mostly focuses on approximating the square root of a positive real number based on these methods. The resulting algorithms can be expressed using Chebyshev polynomials. An extension to the nth root is also proposed
Multi-objective disassembly line optimization with collaborative robots
No full textInternational audienceDisassembly is a crucial phase in remanufacturing and recycling, gaining increasing attention over the past decades due to its importance in sustainable resource management. While various decision-making challenges within the disassembly process have been studied, we focus on two closely related problems: the disassembly sequencing problem and the disassembly line balancing problem, specifically in the context of human-robot collaboration. We consider a bi-objective problem of total time-dependent cost minimization and workload balancing between activated workstations. The goal is to identify the best disassembly task sequence and the optimal assignment of tasks to operators and workstations in order to meet the problem’s objectives. To overcome these challenges, we formulate the problem as a mixed integer program and compare the ϵ-constraint and weighted sum methods to generate Pareto fronts. Computational experiments conducted on real-world end-of-life product structures validate the proposed model’s efficiency on small and medium-scale structures
Dual-Wavelength Surface Plasmon Resonance Microscopy Combined with Laser-Induced Bubble-Cell Perforation: A Novel Single-Cell Manipulation and Real-Time Monitoring Platform
No full textInternational audiencePrecise single-cell membrane perforation holds great promising in membrane biology, yet reported approaches face two major challenges: 1) lacking of both spatiotemporal precision and efficiency at a single-cell resolution; 2) current methods can not non-invasively monitor the realtime perforation dynamics. Herein, we proposed a novel methodology for controlling the cellular membrane perforation and real-time monitoring the mechano-biological imaging with a single-cell resolution. The proposed methodology was developed by an advanced dual-wavelength surface plasmon resonance microscopy (SPRM) platform combined with a femtosecond laser-induced microbubble generation, enabling to achieve precise spatial profile in membrane perturbation through laser focal positioning. Moreover, a typical dual-wavelength fitting algorithm was employed to determine the resonance wavelength (RW), providing a sub-second temporal resolution of 0.1 s/frame. More importantly, the label-free imaging method can provide three key advantages: 1) Non-invasive monitoring of membrane dynamics via an adhesion-dependent RW mapping; 2) Accurate and controllable perforation at a single-cell level; 3) low cost configuration. The integrated platform can establish an important framework for non-invasive investigating the dynamic process of cell membrane under controllable external stimulation in real-time. It can be expected that, this advancement in live-cell imaging field can offer a versatile analytical platform for performing fundamental membrane biophysics study
Correlative Near-Field Characterizations with KPFM, sMIM and SCM to Characterize the Geometry of a N-Channel Failed SiC JFET
No full textInternational audienceAbstract The growing demand for high-performance power electronic devices in a variety of applications has led to a need for efficient, compact designs with increasingly complex architectures. To achieve this, precise and accurate analysis of the local electrical properties at the nanoscale is required to confirm the doping layer geometries at the end of the process. This study employs non-destructive nanoscale characterization techniques to investigate the local properties of designed and fabricated Silicon Carbide (SiC) power devices based on lateral Junction Field-Effect Transistors (JFETs). The SiC sample under study comprises failed n-type lateral JFET channels fabricated with multiple SiC layers in a mesa structure, featuring varying doping levels and several SiC homojunctions. Various electrical modes based on Atomic Force Microscopy (AFM), including the well-known Kelvin Probe Force Microscopy (KPFM), as well as advanced Scanning Capacitance Microscopy (SCM) and scanning Microwave Impedance Microscopy (sMIM) modes, were employed. These three AFM electrical modes enable the mapping of electrical properties, the identification of junctions and the local doping geometries, which are all crucial for device operation and performance. These approaches are highly effective in resolving small-scale variations within multilayers and in providing clear information on doping concentrations, types and work function differences. Mappings from the three modes are compared in terms of sensitivity and signal-to-noise ratio. The impact of the applied VDC during SCM mode on the characterization of SiC junctions is also highlighted
A Novel Approach for Electric Load Prediction Using Convolutional Lstms Networks with Sorted Wavelet Transform Coefficient
No full textInternational audienceTo ensure a constant availability of electrical energy, power companies focus on maintaining a balance between supply and demand. However, the electrical load is influenced by non-linear factors, necessitating the development of robust prediction models. This study aims to enhance electrical load prediction by proposing a model based on wavelet decomposition and LSTM networks. Various models, including stacked LSTM, BiLSTM, CNN-LSTM, and ConvLSTM, were compared and optimized. The results indicate that the ConvLSTM model outperforms the others in terms of accuracy, as assessed by the MAPE, RMSE, and R2 criteria. Subsequently, the ConvLSTM model is refined using wavelet decomposition with coefficient sorting. This approach significantly reduces the discrepancy between actual and predicted load, resulting in a low maximum absolute error of 3 MW. In comparison, the proposed model (WT+ConvLSTM) surpasses ARIMA and MLP-type ANNs in terms of MAPE, with only 0.485%. While the other models demonstrate acceptable performance, WT+ConvLSTM exhibits an RMSE of 0.61 MW and an R2 of 0.99, outperforming alternative approaches. In conclusion, the proposed model, based on the combination of sorted wavelet decomposition with ConvLSTM, demonstrates significant accuracy in electric load forecasting compared to competing approaches in the power sector
"They've given us so many tools". Movements in French civil security's collaborative Ecologies of Artifacts
No full textInternational audienceIn this paper, we investigate the evolutions of digital tools and collaborative practices in a French regional civil security service (SDIS) and its partners (firefighters, gendarmes, emergency health services, and local government). Through the lens of Ecologies of Artifacts (EoA), we analyze such evolutions as "movements" within complex socio-technical collaborative environment. Drawing on empirical material collected through the observation of an inter-service training, three on-site visits, and 40 interviews with 33 different professionals, we characterise a dense, redundant, interconnected, and misunderstood EoA. We identify several movements in its evolutive EoA: personal-to-organizational, outside-introduction, top-down, bottom-up, horizontal, suppression, realignment, replacement, as well as their associated ripple effects, and accelerating factors. From our findings and previous research, we discuss the adaptability of the EoA, the movements' emerging features at the different levels of EoA, and generalizability to other contexts. We advocate for a movement-sensitive design approach, which means supporting tools integration within existing EoA rather than replacing them outright