Abstract We propose to improve the efficiency of computation reduced‐state variables related a given reduced basis. This basis is supposed be built by using snapshot proper orthogonal decomposition (POD) model reduction method. In framework non‐linear mechanical problems involving internal variables, local integration constitutive laws can dramatically limit computational savings provided order model. drawback due fact that, Galerkin formulation, size has no effect on complexity equations....

Abstract In this paper, we propose a general framework for projection-based model order reduction assisted by deep neural networks. The proposed methodology, called ROM-net , consists in using learning techniques to adapt the reduced-order stochastic input tensor whose nonparametrized variabilities strongly influence quantities of interest given physics problem. particular, introduce concept dictionary-based ROM-nets where networks recommend suitable local from dictionary. dictionary models...

ABSTRACT Reduced order modeling (ROM) is applied to the finite element thermo‐mechanical simulation of metal additive manufacturing at part scale. This a significant challenge because continuously evolving computational domain, on which local reduced basis required apply projection‐based ROM. In this paper, ROM mechanical resolution, much more time‐consuming than thermal one. Considering DED processes (directed energy deposition), it proposed organize training set snapshots according an...

Abstract In this paper a new extension of the mesh‐free natural element method (NEM) is presented. approach, coined as constrained (C‐NEM), visibility criterion introduced to select neighbours in computation shape functions. The these functions based on modified, Voronoi diagram. With technique, some difficulties inherent non‐convex domains are avoided and analysis problems involving cracks or discontinuities now easily performed. As NEM satisfies Kronecker delta property, imposition...

In this work, we propose a framework that constructs reduced order models for nonlinear structural mechanics in nonintrusive fashion, and can handle large scale simulations. We identify three steps are carried out separately time, possibly on different devices: (i) the production of high-fidelity solutions by commercial software, (ii) offline stage model reduction (iii) online where is exploited. The nonintrusivity assumes only displacement field solution known, relies operations simulation...

Summary The model reduction of mechanical problems involving contact remains an important issue in computational solid mechanics. In this article, we propose extension the hyper‐reduction method based on a reduced integration domain to frictionless written by mixed formulation. As potential zone is naturally through mesh involved hyper‐reduced equations, dual basis chosen as restriction full‐order basis. We then obtain hybrid combining empirical modes for primal variables with finite element...

A novel algorithmic discussion of the methodological and numerical differences competing parametric model reduction techniques for nonlinear problems is presented. First, Galerkin reduced basis (RB) formulation presented, which fails at providing significant gains with respect to computational efficiency problems. Renowned methods computing time order models are Hyper-Reduction (Discrete) Empirical Interpolation Method (EIM, DEIM). An description a comparison both provided. The accuracy...

X-Ray Computed Tomography (XCT) techniques have evolved to a point that high-resolution data can be acquired so fast classic segmentation methods are prohibitively cumbersome, demanding automated pipelines capable of dealing with non-trivial 3D images. Meanwhile, deep learning has demonstrated success in many image processing tasks, including materials science applications, showing promising alternative for human-free pipeline. However, the rapidly increasing number available architectures...

We consider the dictionary-based ROM-net (Reduced Order Model) framework [Daniel et al., Adv. Model. Simul. Eng. Sci. 7 (2020) https://doi.org/10.1186/s40323-020-00153-6 ] and summarize underlying methodologies their recent improvements. The object of interest is a real-life industrial model an elastoviscoplastic high-pressure turbine blade subjected to thermal, centrifugal pressure loadings. main contribution this work application complete workflow quantification uncertainty dual quantities...

We propose an a posteriori estimator of the error hyper-reduced predictions for elastoviscoplastic problems. For given fixed mesh, this aims to forecast validity domain in parameter space, hyper-reduction approximations. This evaluates if simulation outputs generated by model represent convenient approximation that finite element would have predicted. do not account related upon which is introduced. restrict our attention generalized standard materials. Upon use incremental variational...

The availability of accurate and efficient numerical simulation tools has become utmost importance for the design optimization phases existing industrial processes. latter requires computation multiple physical fields governed by coupled systems partial differential equations tends to require large computational resources. Recently, coupling machine learning techniques with allowed lifting part this burden, replacing parts resolution process trained neural networks, whose execution cost is...

This paper presents a new nonlinear projection based model reduction using convolutional Variational AutoEncoders (VAEs). framework is applied on transient incompressible flows. The accuracy obtained thanks to the expression of velocity and pressure fields in manifold maximising likelihood pre-computed data offline stage. A confidence interval for each time instant definition reduced dynamic coefficients as independent random variables which posterior probability given known. parameters are...