Comptes Rendus Mecanique最新文献

The hip fracture is one of the most common diseases for elder people and also, one of the most worrying one since it usually is the starting point of further complications for both, the health of the patient and their daily life. Additionally, reports shown that there exist differences between people living in different regions, thus limiting the use of global models. In this work we propose a hip fracture prediction tool for a local region, using clinical data of the population of that region. The data is processed with a dimensionality reduction tool in combination with and hyper-parametrization process and the corresponding hyper-parameter optimization process for obtaining good predictions in the diagnoses, as the results shown.

The aim of this work is to derive an accurate model of two-dimensional switched control heating system from data generated by a Finite Element solver. The nonintrusive approach should be able to capture both temperature fields, dynamics and the underlying switching control rule. To achieve this goal, the algorithm proposed in this paper will make use of three main ingredients: proper orthogonal decomposition (POD), dynamic mode decomposition (DMD) and artificial neural networks (ANN). Some numerical results will be presented and compared to the high-fidelity numerical solutions to demonstrate the capability of the method to reproduce the dynamics.

We present here the statistical models that are most in use in survival data analysis. The parametric ones are based on explicit distributions, depending only on real unknown parameters, while the preferred models are semi-parametric, like Cox model, which imply unknown functions to be estimated. Now, as big data sets are available, two types of methods are needed to deal with the resulting curse of dimensionality including non informative factors which spoil the informative part relative to the target: on one hand, methods that reduce the dimension while maximizing the information left in the reduced data, and then applying classical stochastic models; on the other hand algorithms that apply directly to big data, i.e. artificial intelligence (AI or machine learning). Actually, those algorithms have a probabilistic interpretation. We present here several of the former methods. As for the latter methods, which comprise neural networks, support vector machines, random forests and more (see second edition, January 2017 of Hastie, Tibshirani et al. (2005) [1]), we present the neural networks approach. Neural networks are known to be efficient for prediction on big data. As we analyzed, using a classical stochastic model, risk factors for Alzheimer on a data set of around 5000 patients and $p=17$ factors, we were interested in comparing its prediction performance with the one of a neural network on this relatively small sample size data.

In the past, data in which science and engineering is based, was scarce and frequently obtained by experiments proposed to verify a given hypothesis. Each experiment was able to yield only very limited data. Today, data is abundant and abundantly collected in each single experiment at a very small cost. Data-driven modeling and scientific discovery is a change of paradigm on how many problems, both in science and engineering, are addressed. Some scientific fields have been using artificial intelligence for some time due to the inherent difficulty in obtaining laws and equations to describe some phenomena. However, today data-driven approaches are also flooding fields like mechanics and materials science, where the traditional approach seemed to be highly satisfactory. In this paper we review the application of data-driven modeling and model learning procedures to different fields in science and engineering.

In this paper, an algorithm for identifying equations representing a continuous nonlinear dynamical system from a noise-free state and time-derivative state measurements is proposed. It is based on a variant of the extended dynamic mode decomposition. A particular attention is paid to guarantee that the physical invariant quantities stay constant along the integral curves. The numerical methodology is validated on a two-dimensional Lotka–Volterra system. For this case, the differential equations are perfectly retrieved from data measurements. Perspectives of extension to more complex systems are discussed.

The present work aims at proposing a new methodology for learning reduced models from a small amount of data. It is based on the fact that discrete models, or their transfer function counterparts, have a low rank and then they can be expressed very efficiently using few terms of a tensor decomposition. An efficient procedure is proposed as well as a way for extending it to nonlinear settings while keeping limited the impact of data noise. The proposed methodology is then validated by considering a nonlinear elastic problem and constructing the model relating tractions and displacements at the observation points.

This paper introduces a new vision of data-driven structure computation taking advantage of Material Science, especially for highly nonlinear and time-dependent material behaviours. Technical solutions are also derived, in order to build internal hidden variables defining the so-called “Experimental Constitutive Manifold”.

Reduced order modeling (ROM) provides an efficient framework to compute solutions of parametric problems. Basically, it exploits a set of precomputed high-fidelity solutions—computed for properly chosen parameters, using a full-order model—in order to find the low dimensional space that contains the solution manifold. Using this space, an approximation of the numerical solution for new parameters can be computed in real-time response scenario, thanks to the reduced dimensionality of the problem. In a ROM framework, the most expensive part from the computational viewpoint is the calculation of the numerical solutions using the full-order model. Of course, the number of collected solutions is strictly related to the accuracy of the reduced order model. In this work, we aim at increasing the precision of the model also for few input solutions by coupling the proper orthogonal decomposition with interpolation (PODI)—a data-driven reduced order method—with the active subspace (AS) property, an emerging tool for reduction in parameter space. The enhanced ROM results in a reduced number of input solutions to reach the desired accuracy. In this contribution, we present the numerical results obtained by applying this method to a structural problem and in a fluid dynamics one.

This paper deals with a pseudo-parabolic equation involving variable exponents under Dirichlet boundary value condition. The author proves that the solution is not global in time when the initial energy is positive. This result extends and improves a recent result obtained by Di et al. (2017) [1].

In this work, we have investigated numerically the disappearance of wrinkles from a tended membrane by the Asymptotic Numerical Method (ANM) using the finite-element DKT18. The ANM is a path-following technique that has been used to solve bifurcation problems. We show numerically the influence of the terms corresponding to the membrane displacement gradient in the Föppl–von Kármán (FvK) theory on the bifurcation curves in the case of a stretched elastic membrane. We will also study numerically, by using the ANM algorithm, the influence of the thickness and of the aspect ratio on the re-stabilization of a rectangular elastic membrane during stretching. The results obtained by our model are compared with those obtained using the industrial code ABAQUS.