Jose Vicente Aguado

Real-time monitoring of thermal processes by reduced-order modeling

1 Institut de Recherche en Génie Civil et Mécanique (GeM UMR CNRS 6183), Ecole Centrale de Nantes. 1 rue de la Noë, BP 92101, F-44321 Nantes cedex 3, France. e-mail: {jose.aguado-lopez,francisco.chinesta}@ec-nantes.fr, web http://rom.ec-nantes.fr 2Laboratori de Calcul Numeric (LaCaN). Departament de Matematica Aplicada III E.T.S. de Ingenieros de Caminos, Canales y Puertos, Universitat Politecnica de Catalunya, BarcelonaTech, 08034 Barcelona, Spain. e-mail: antonio.huerta@upc.edu, web http://www.lacan.upc.edu 3Zienkiewicz Centre for Computational Engineering, College of Engineering, Swansea University, Swansea SA2 8PP, UK. 4Aragon Institute of Engineering Research (I3A), Universidad de Zaragoza, Maria de Luna 3, E-50018 Zaragoza, Spain. e-mail: ecueto@unizar.es, web http://amb.unizar.es

Reduced order modelling for efficient numerical optimisation of a hot-wall chemical vapour deposition reactor

Purpose The purpose of this paper is to present a reduced order computational strategy for a multi-physics simulation involving a fluid flow, electromagnetism and heat transfer in a hot-wall chemical vapour deposition reactor. The main goal is to produce a multi-parametric solution for fast exploration of the design space to perform numerical prototyping and process optimisation. Design/methodology/approach Different reduced order techniques are applied. In particular, proper generalized decomposition is used to solve the parameterised heat transfer equation in a five-dimensional space. Findings The solution of the state problem is provided in a compact separated-variable format allowing a fast evaluation of the process-specific quantities of interest that are involved in the optimisation algorithm. This is completely decoupled from the solution of the underlying state problem. Therefore, once the whole parameterised solution is known, the evaluation of the objective function is done on-the-fly. Originality/value Reduced order modelling is applied to solve a multi-parametric multi-physics problem and generate a fast estimator needed for preliminary process optimisation. Different order reduction techniques are combined to treat the flow, heat transfer and electromagnetism problems in the framework of separated-variable representations.

Separated Representations of Incremental Elastoplastic Simulations

Forming processes usually involve irreversible plastic transformations. The calculation in that case becomes cumbersome when large parts and processes are considered. Recently Model Order Reduction techniques opened new perspectives for an accurate and fast simulation of mechanical systems, however nonlinear history-dependent behaviors remain still today challenging scenarios for the application of these techniques. In this work we are proposing a quite simple non intrusive strategy able to address such behaviors by coupling a separated representation with a POD-based reduced basis within an incremental elastoplastic formulation.

Fractional modelling of functionalized CNT suspensions

Experimental findings and rheological modelling of chemically treated single-wall carbon nanotubes suspended in an epoxy resin were addressed in a recent publication (Ma et al., J Rheol 53:547–573, 2009). The shear-thinning behaviour was successfully modelled by a Fokker-Planck-based orientation model. However, the proposed model failed to describe linear viscoelasticity using a single mode as well as the relaxation after applying a finite step strain. Both experiments revealed a power-law behaviour for the storage and relaxation moduli. In this paper, we show that a single-mode fractional diffusion model is able to predict these experimental observations.

Reshaping of large aeronautical structural parts: A simplified simulation approach

Large aeronautical structural parts present important distortions after machining. This problem is caused by the presence of residual stresses, which are developed during previous manufacturing steps (quenching). Before being put into service, the nominal geometry is restored by means of mechanical methods. This operation is called reshaping and exclusively depends on the skills of a well-trained and experienced operator. Moreover, this procedure is time consuming and nowadays, it is only based on a trial and error approach. Therefore, there is a need at industrial level to solve this problem with the support of numerical simulation tools. By using a simplification hypothesis, it was found that the springback phenomenon behaves linearly and it allows developing a strategy to implement reshaping at an industrial level.

A manifold learning approach to data-driven computational materials and processes

Standard simulation in classical mechanics is based on the use of two very different types of equations. The first one, of axiomatic character, is related to balance laws (momentum, mass, energy, …), whereas the second one consists of models that scientists have extracted from collected, natural or synthetic data. In this work we propose a new method, able to directly link data to computers in order to perform numerical simulations. These simulations will employ universal laws while minimizing the need of explicit, often phenomenological, models. They are based on manifold learning methodologies.

Model reduction & manifold learning — Based parametric computational electromagnetism: Fundamentals & applications

Techniques based on the use of separated representations, are at the heart of the so-called Proper Generalized Decomposition methods. Such separated representations were employed for solving multidimensional models suffering the so-called curse of dimensionality, transient models within a non-incremental integration schema and in the context of uncertainty propagation. Then, its use was extended for separating space coordinates making possible the solution of models defined in degenerated domains, e.g. plate, shells or laminates, and finally for addressing parametric models where model parameters were considered as problem extra-coordinates.

Real-time monitoring of thermal processes by reduced-order modeling: REAL-TIME MONITORING OF THERMAL PROCESSES BY REDUCED-ORDER MODELING

1 Institut de Recherche en Génie Civil et Mécanique (GeM UMR CNRS 6183), Ecole Centrale de Nantes. 1 rue de la Noë, BP 92101, F-44321 Nantes cedex 3, France. e-mail: {jose.aguado-lopez,francisco.chinesta}@ec-nantes.fr, web http://rom.ec-nantes.fr 2Laboratori de Calcul Numeric (LaCaN). Departament de Matematica Aplicada III E.T.S. de Ingenieros de Caminos, Canales y Puertos, Universitat Politecnica de Catalunya, BarcelonaTech, 08034 Barcelona, Spain. e-mail: antonio.huerta@upc.edu, web http://www.lacan.upc.edu 3Zienkiewicz Centre for Computational Engineering, College of Engineering, Swansea University, Swansea SA2 8PP, UK. 4Aragon Institute of Engineering Research (I3A), Universidad de Zaragoza, Maria de Luna 3, E-50018 Zaragoza, Spain. e-mail: ecueto@unizar.es, web http://amb.unizar.es

Elastic-Plastic Reduced Order Modelling of Sheet and Profiles Bending-under-Tension

The aircrafts fuselage structure is usually composed of an assembly of stringers and frames made of cold-worked aluminium profiles. In particular, frames need of a forming process that shapes the profile into the frame’s curved shape. To do this, both profile ends are clamped, and then the profile is simultaneously stretched and pressed against the mould so that the material is plastically deformed. Industrial experience shows that most of times the resultant frame does not fulfil neither curvature nor planarity tolerances. These defects are mainly due to spring-back, residual stresses, and some technologic restrictions related to the machinery. The lack of understanding has led industry to reduce the automation level, and thus the forming process is frequently interrupted to perform verifications and adjustments that make the process to be time-consuming and very much dependent on the know-how of the machine operator. Aiming to improve the frame’s industrialisation, this work first analyses the influence of several parameters in the final shape. Then, we propose a computer-aided forming process based on the concept of Computational Vademecum (CV), which is also introduced in this work. It allows reducing the dependence on the operator know-how, while reliability and repeatability of the process can be improved.

Towards Online Control of Forming Processes Involving Residual Stresses: Defining Multi-Parametric Computational vademecums

Manufacturing processes usually involve several physical or chemical treatments aiming to confer improved properties to the material. As a result of these processes, a certain residual stress field is developed inside the material. When the material is plastically deformed during a forming process, residual stresses lead to spring-back. In practice, the residual stress field cannot be calculated directly due to the lack of information and the high computational cost that would be required in order to simulate the material history. Instead, some authors have suggested using Artificial Neural Networks (ANN) in order to be able to predict the final shape of the piece without being required to compute the residual stress field. However, the network needs experimental data to be trained and any configuration outside the experimental range makes the prediction to be not reliable. Aiming to overcome these limitations, first steps of a new approach is presented here. This work proposes to compute a general solution by means of the PGD method which depends on a series of parameters defining the residual stress field. This computation can be done offline. Then, the parameters could be identified by comparing online the measured behaviour with the predicted one. Introduction Predicting the final shape of a certain piece which is plastically deformed during a forming process is a goal of major interest for the industry. Frequently we observe non negligible deviations from the predicted shape to the real one. This fact occurs even if we are capable of performing strong simulations of the forming process because of the uncertainty associated to the residual stress field. These is an extense variety of sources for the residual stresses, being the thermal and chemical treatments the classical ones. In any case, the lack of knowledge about the residual stress field constitutes the main difficulty for accurately predicting the shape after the forming process. Several approaches exist for adressing this problem. The first one could be based on simulating not only the forming process but the whole history of the piece of interest, or even the entire material’s history. In practice, this is computationally unaffordable but even in the case that the computation was possible, there exist another constraint which is the necessity of going back into manufacturing chain in order to collect a huge quantity of information that very often is not available. For these reasons, some authors have suggested other possibilities. In particular, Artificial Neural Networks (ANN) have been explored [1, 2]. This approach requires to choose a set of parameters and to perform a series of experiments for some combinations of them. In fact, it can be seen as a sort of sampling of the parametric space. From the experimental results, the ANN is trained and in some cases it has been demonstrated to provide acceptable results. The principal advantage is that we circumvent all mechanical calculations and therefore we do not need to know the residual stress field anymore. However, there exist some major inconvinients that should be pointed out. Firstly, the number of parameters should not be too big in order to keep the quantity of experiments reasonably low. With the same argument, it is obvious that Key Engineering Materials Online: 2013-06-13 ISSN: 1662-9795, Vols. 554-557, pp 699-705 doi:10.4028/www.scientific.net/KEM.554-557.699