Modeling large-scale flood inundation requires weeks of calculations using complex fluid software. The state-of-the-art in operational hydraulic modeling does not currently allow flood real-time forecasting fields. Data driven models have small computational costs and fast computation times and may be useful to overcome this problem. In this paper, we propose a new modeling approach based on a coupled of Hydrodynamics finite element model and Multi-headed Deep convolutional neural network (MH-CNN) with rain precipitations as input to forecast rapidly the water depth reached in large floodplain with few hours-ahead. For this purpose, one first builds a database containing different simulations of the physical model according to several rain precipitation scenarios (historic and synthetic). The multi-headed convolutional neural network is then trained using the constructed database to predict water depths. The pre-trained model is applied successfully to simulate the real July 2014 flood inundation in an 870 km2 area of La Nive watershed in the south west of France. Because rain precipitation forecast data is more accessible than discharge one, this approach offers great potential for real-time flood modelling for ungauged large-scale territories, which represent a large part of floodplain in the world.

The objective of this paper is to present an effective new methodology to optimize the maintenance costs of bridges stock. Optimization takes place at the network level and not in a project level (bridge by bridge). The dynamics of passage between bridges condition state (from 1 to 5) is achieved by the Markov chains probabilistic method. The Markov transition matrix is determined either by ratios of total areas and areas degraded annually, or by the resolution of an optimization problem. In the latter case, the nonlinear optimization algorithm SQP (Sequanciel Quadratic Programming) is developed. A bridge maintenance matrix is introduced in the calculation of the repair cost. The originality of our approach is to parameterize this matrix by introducing the different optimization variables of the problem. Finally, the cost function to be optimized annually is calculated and optimized by a genetic algorithm. This cost function represents the cost of maintaining the entire asset.

The prediction and reassessment of mechanical behaviour of reinforced structures affected by delayed ettringite formation (DEF) is a major challenge for structure managers. Firstly, several experimental tests were performed in laboratory to study the influence of both uni and tri-axial reinforcements on DEF expansion. Strain decreases in reinforced directions were observed, proving that DEF expansion under restraint is anisotropic. Cracks were observed parallel to the restrained directions. No strain trans-fert occurs from restrained directions to other ones. Secondly, data provided by these results are used to fit a numerical finite element model taking into account both chemical and mechanical aspects of DEF. Finally, an element of a DEF damaged structure is modelled with the whole model and compared to on site observations.

Delayed ettringite formation (DEF) is a sulfate attack affecting civil engineering structures. This chemical reaction takes place within the concrete matrix of structures and causes damage in concrete and tension in reinforcements. For managers and owners, the ability to predict and reassess the mechanical behaviour of such structures is a major challenge. The influence of both reinforcements and prestress on DEF expansion were studied in the present work. Several tests were performed in laboratory: expansion under both uni and tri-axial restraint due to reinforcements and expansion under prestress (stress level of 14.5 MPa). Uniaxial restraint led to decreased strain in the restrained direction. For prestressed concrete, the loaded direction exhibited creep strain. In both cases, expansions were not impacted in transversal free directions. Therefore, DEF expansion under uniaxial stress is anisotropic. Cracks were observed parallel to the restrained direction. The final volumetric expansion was lower than in stress-free conditions (decrease of 27% for uni-axially restrained condition). For triaxially restrained tests and prestressed specimens, the volumetric decreases were 56% and 34% respectively. Data provided by these results will be used for numerical reassessment of DEF damaged structures. (C) 2018 Elsevier Ltd. All rights reserved.

This study deals with simulation of interaction of bodies in urban flow when the bodies are assimilate to have a rectangular shape. A Galerkin type finite element discretization with velocity field richer than pressure field is employed to obtain the fluid discretized non-linear relations. The particles displacements are computed by using a rigid-body motion method and a collision strategy is developed to handle cases in which bodies touch.

This study deals with the simulation of transport and interaction between bodies considered as a rectangular shape particles, in urban flow. We used an hydrodynamic two-dimensional finite elements model coupled to the particles model based on Maxey-Riley equations, and taking into account of contact between bodies. The finite element discretization is based on the velocity field richer than pressure field, and the particles displacements are computed by using a rigid body motion method. A collision strategy is also developed to handle cases in which bodies touch.

The main limitation of the most common form of the equations of Boussinesq is that they are valid only for relatively shallow depth of water. In the present study, an innovative approach (h-s) based on the finite elements method is presented to improve the dispersion relationship. This approach, based on the so-called extended Boussinesq model, reproduces with a high degree of accuracy the propagation of waves processes on a greater range of depths than the standard Boussinesq models.

Cet article est consacré à une présentation détaillée de deux schémas de résolution non linéaires adaptés à la résolution des systèmes éléments finis des équations de Saint-Venant incompressible.Le premier schéma est la linéarisation de Newton-Raphson, le second est le schéma Newton-Asymptotique. Les relatives performances des deux schémas sont mises en évidence à travers un exemple.

We propose in this paper the premises of an Interactive Decision Support System dedicated to the unidirectional tolerancing of mechanical assemblies. The foundation of our work is an efficient representation of a mechanical assembly based on graph theory. From this niodel, we can determine all tile configurations of an assembly. An extended syntax for the functional constraints permits then to generate automatically all the tolerance chains. The final result is an autornatically generated linear system of equations and inequations expressing the tolerance chains, as well as the existence of the configurations and coherence of the dimensioning scheme. This system can then be solved by an analysis or a synthesis approach.

In this paper we present a comparative study of three non-linear schemes for solving finite element systems of Navier-Stokes incompressible flows. The first scheme is the classical Newton-Raphson linearization, the second one is the modified Newton-Raphson linearization and the last one is a new scheme called the asymptotic-Newton method. The relative efficiency of these approaches is evaluated over a large number of examples. (C) 1997 John Wiley & Sons, Ltd.No. of Figures: 14. No. of Tables: 4. No. of References: 8.

ABSTRACT This paper presents an overview of different iterative methods Conjugate gradient (CG), Biconjugate gradient (BICG), conjugate gradient squared (CGS), Biconjugate gradient stabilized (BICGSTAB), transpose-free quasi minimal residual (TFQMR), full orthogonal method (FOM) et Generalized minimal residual (GMRES) for solving finite element systems of Navier-Stokes incompressible flows. To accelerate convergence of those methods, preconditionner based on incomplete Gauss factorisation (ILU) is used. The accent is put on the necessity to renumber the unknowns to guaranty the convergence of the iterative methods. The importance of a variable stop criterion of iterative methods for each Newton-Raphson step is underlined.