HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Asymptotic Analysis of a Kernel-Type Estimator for Parabolic Stochastic Partial Differential Equations Driven by Cylindrical Sub-Fractional Brownian Motion
No full textInternational audienceThe main purpose of the present paper is to investigate the problem of estimating the time-varying coefficient in a stochastic parabolic equation driven by a sub-fractional Brownian motion. More precisely, we introduce a kernel-type estimator for the time-varying coefficient θ(t) in the following evolution equation:du(t,x)=(A0+θ(t)A1)u(t,x)dt+dξH(t,x),x∈[0,1],t∈(0,T],u(0,x)=u0(x), where ξH(t,x) is a cylindrical sub-fractional Brownian motion in L2[0,T]×[0,1], and A0+θ(t)A1 is a strongly elliptic differential operator. We obtain the asymptotic mean square error and the limiting distribution of the proposed estimator. These results are proved under some standard conditions on the kernel and some mild conditions on the model. Finally, we give an application for the confidence interval construction
Investigating turbulence and turbulent flame propagation in dust clouds
No full textInternational audienc
Computing Improved Explanations for Random Forests: k-Majoritary Reasons
No full textInternational audienceThis work focuses on improving explanations for random forests, which, although efficient and providing reliable predictions through the combination of multiple decision trees, are less interpretable than individual decision trees. To improve their interpretability, we introduce k-majoritary reasons, which are minimal implicants for inclusion supporting the decisions of at least k trees, where k is greater than or equal to the majority of the trees in the forest. These reasons are robust and provide a better explanation of the forest’s decision. However, due to their large size and our cognitive limitations, they may be too hard to interpret. To overcome this obstacle, we propose probabilistic majoritary explanations, which provide a more concise interpretation while maintaining a strict majority of trees. We identify the computational complexity of these explanations and propose algorithms to generate them. Our experiments demonstrate the effectiveness of these algorithms and the im provement in interpretability in terms of size provided by probabilistic majoritary explanations (δprobable majoritary reasons)
Adaptive Integral-Gain Controller for Robust Quadrotor Navigation with Fourier Neural Network Compensation
No full textInternational audienceIn this paper, an adaptive integral-gain controller with Fourier Neural Network (NN) compensation is proposed to handle changing operational conditions, particularly aerodynamic disturbances in quadrotor navigation. The proposed scheme compensates for external disturbances by approximating the residual error dynamics, ensuring that the parameters are bounded over time without prior knowledge. Theoretical guarantees are provided in the Lyapunov sense, considering the closed-loop system with adaptive and NN mechanisms. The experimental results, using a Parrot AR Drone 2.0 in torque mode, validate the control proposal under different operational conditions, showing superior tracking accuracy and faster recovery from sustained and intermittent disturbances compared to its non-adaptive counterpart.</div
Energy Optimization in IoT Adaptive Security: a Performance Comparison of Deep Reinforcement Learning Approaches
No full textInternational audienc
Novel Approach to the Correlated Storage Location Assignment Problem (CSLAP) in Automated Warehouse Environments.
No full textInternational audienc
Modeling Size Effects in the Elastoplastic Behavior of Nanocomposites
No full textInternational audienc
Phosphorylated chitosan as a bio-based modifier for urea formaldehyde adhesives in wood composite applications
No full textInternational audienceThis study employs chitosan (CS) derived from chitin (CT) isolated from shrimp shell waste (SSW) and its phosphorylated form as sustainable bio-based modifiers for urea-formaldehyde (UF) adhesives, with the aim of improving the mechanical and physical performances of particleboard (PB) panels and reducing formaldehyde emissions (FE). CS and phosphorylated chitosan (P-CS) were prepared and characterized using SEM, XRD, ATR-FTIR, and TGA-DTG analyses. PBs were manufactured using varying proportions of CS and P-CS (4–10 wt%) and compared to a commercial UF adhesive (0:100). Mechanical and physical properties were evaluated to assess performance. The results revealed that the optimal addition of 8 wt% P-CS significantly improved mechanical properties. The dry internal bond (0.65 ± 0.03 MPa), modulus of elasticity (3659 ± 14 MPa), modulus of rupture (28 ± 0.70 MPa), and surface soundness (1.88 ± 0.04 MPa) increased by 44.44 %, 43.26 %, 75 %, and 19.74 %, respectively, compared to the control UF resin. Additionally, FE values decreased by approximately 34.63 % with the UF:P-CS (92:8; w/w) formulation. In terms of physical performance, the thickness swelling (TS) and water absorption (WA) values of the optimal adhesives, UF:CS /P-CS (92:8; w/w), were significantly lower than those of the control UF:CS /P-CS (100/0; w/w). TS values were 8.06 ± 0.09 % and 10.99 ± 0.27 %, while WA values were 39.33 ± 1.84 % and 90.01 ± 2.02 % after 2 and 24 h, respectively. These findings demonstrate the potential of P-CS as a sustainable additive for improving UF adhesives in PB manufacturing
Enhanced cellulose extraction from delignified oil palm empty fruit bunches using sequential ultrasound-microwave processing
No full textInternational audienceOil palm empty fruit bunches (OPEFBs) are among the most abundant agricultural residues in palm oil producing countries, yet they are often underutilized and discarded as waste. Rich in cellulose, a valuable biopolymer used in bio-based materials, OPEFBs present a sustainable opportunity to convert agricultural waste into high-value products through cellulose isolation. This study presents a new approach to isolating cellulose from OPEFBs. The method offers several operational advantages that promote sustainable biomass conversion. Conventional cellulose extraction methods use significant amounts of chemicals and energy, which considerably hinders their industrial application. Instead, the present work introduces an alternative to the conventional method of treating organic biomass by developing a two-step process. The first step involves ultrasound-assisted alkaline pretreatment to initiate delignification and disrupt the fiber structure. This is followed by microwave-assisted delignification, which delivers uniform thermal energy to enhance lignin removal and enable efficient cellulose extraction. Compared to the conventional hot-plate method, which achieved a cellulose recovery of 61.11 % with chemical additives, the combined ultrasound–microwave process yielded a significantly higher recovery of 97.48 % under the same NaOH concentration (2 M) and reaction time (30 min). Even without additives, the combined method achieved 81.05 % recovery, outperforming standalone ultrasound (72.80 %) and microwave (76.78 %) treatments. The process was further optimized using CCD and RS), yielding an optimal condition at 1.619 M NaOH and 49.381 min, with a predicted cellulose recovery of 98.90 %, closely validated by experimental results (98.9 ± 0.01 %). Compositional analysis revealed significant removal of lignin and hemicellulose, while thermal analysis demonstrated remarkable physicochemical properties in the treated pulp samples. This integrated method offers an economical and eco-friendly approach for biomass valorization. While the process achieved high efficiency under experimental conditions, further investigation are necessary to evaluate its scalability and feasibility for industrial applications, particularly regarding energy consumption and cost
Control for (highly) flexible geometrically exact 2D Reissner beam by energy shaping of distributed port-Hamiltonian system
No full textInternational audienceThe main focus of this work is control of large overall motion of (highly) flexible Reissner beam that can represent in geometrically exact manner large displacements, large rotations and large strains. This nonlinear control problem is cast in distributed port-Hamiltonian framework, extending the previous works (Duindam et al. in Modeling and Control of Complex Physical Systems: The Port-Hamiltonian Approach, Springer, Berlin, 2009) dealing with finite dimensional systems (such as rigid components of multibody system interconnected with flexible joints) to infinite dimensional system with Hamiltonian density (Ljukovac et al. in Int. J. Numer. Methods Eng. 126, 2025). The nonlinear control is performed by defining the desired state of the flexible system through energy-shaping of distributed port-Hamiltonian, previously introduced for lumped-parameter system (Ortega et al. in IEEE Control Syst. Mag. 21:18-33, 2001; Brogliato et al. in Dissipative Systems Analysis and Control: Theory and Aplications, Springer, Cham, 2020). We first show how to perform the energy shaping for internal energy density by bringing the flexible beam with large overall motion into deformed configuration that is in static equilibrium, which is also the closest to the desired configuration of Reissner beam after large overall motion. We then show how to perform the energy shaping for kinetic energy, with the illustration provided for the choice of uniform rotational motion, which also requires the corresponding choice of internal energy defined by Casimir functional (Marsden et al. in Hamiltonian Reduction by Stages, Springer, Berlin, 2007). We finally discuss different procedures for damping injection that will stabilize the system to configurations with desired Hamiltonian, including viscous damping, frictional damping and energy decaying of high frequency modes for Reissner nonlinear beam. The results of illustrative numerical simulations of large overall motion of (very) flexible geometrically exact beam confirm the good performance of the proposed approach.</div