Towards rationalised computational expense for simulating fracture over multiple scales

Lead Research Organisation: Cardiff University
Department Name: Sch of Engineering

Abstract

The project focuses on the numerical simulation of the failure of complex, heterogeneous structures. The simulation of such physical phenomena is of particular interest to practitioners as it enables to limit the number of destructive tests required to design and assess structures, and, ultimately, to decrease the safety factors used in design.

In such heterogeneous media, the description of crack or damage initiation and propagation must be done at the scale of the inhomogeneities (e.g. aggregates in a concrete structure) in order for the results to be predictive. If one uses such a fine-scale material model to simulate structures at an engineering scale (e.g. an aircraft composite panel or a concrete beam), very large numerical problems need to be solved. In addition, there is a strong need for engineers to run their models numerous times, for different sets of the design parameters (e.g. loading conditions, geometry or material properties).

Tackling such parametric multiscale problems is prohibitively expensive when using brute force parallel computing. However, one can use the fact that solutions to parametric problems usually evolve in a relatively coarse space: solutions to nearby parameter sets are usually close in a certain sense. This idea is classically used in Model Order Reduction, which proposes to reduce the size of the initial problem by several order of magnitude by simply reusing the information generated when solving the initial problem for several different sets of parameters.

However, in the case of fracture, the information provided by the initial problem is most of the time insufficient to describe the behaviour of the system for arbitrary parameters. Crack paths, defects, and subsequent ultimate strengths are strongly influenced by an even slight variation in the parameter set. Fortunately, we showed in our previous research that this characteristic only affects a local region surrounding the structural defects, whilst the behaviour far from these regions is remains relatively unchanged for a wide range of parameter values.

The proposed project will make use of this observation in a generic way, by coupling Reduced Order Modeling and Domain Decomposition. The structure will be divided in smaller subcomponents, on which Reduced Order Modeling will be applied separately. The consequence will be that the computational efforts will be greatly decreased in the regions that are far away from the damaged zone. Within the process zone itself, the substructuring framework will allow us to automatically switch to classical direct solvers.

In this sense, the research aims at rationalising the computational costs associated to the simulation of parametrised multiscale fracture simulations, by concentrating the numerical effort where it is most required and with minimal intervention of the user.

Planned Impact

The project has the potential to have impact on engineering researchers working in various area of expertise where modelling over multiple scale is involved. The proposed method is mainly based on Linear Algebra, and can be used without significant modifications to solve problems in fluid dynamics, electromagnetics and nano-mechanics (see letters of support).

The fundamental part of the work addresses a problem where there is an intrinsic lack of separation of scales. This is one of the most challenging research areas in computational engineering and physics and this project will open new doors toward a new family of solvers for cheap and reliable numerical simulations of such problems.

Developers, designers, manufacturers will, in the medium to long term, benefit from cheaper numerical methods for multiscale problems, and in particular for fracture mechanics. It will enable them to incorporate more details in their models thereby allowing more optimised products. Fracture is important in most areas of engineering and science and this project will impact various domains of industry such as manufacturing, aerospace, mechanical, environmental, civil and electronics engineering (see letters of support).

In order to ensure that the beneficiaries, both industrial and academic, do benefit from the research, a workshop will be organised at the end of the 13th month of the programme. This will increase the visibility of the research at the academic level but also involve key stakeholders (stated below) to ensure that the next phases of the research are relevant to their needs. At this workshop, we will invite the relevant members of the UK Association for Computational Mechanics in Engineering as well as industrialists and stakeholders with whom we have ongoing and meaningful relationships (BAE Systems, Defense systems and technology laboratory Dstl, Rolls-Royce, EADS, European Space Agency, Cenaero, inuTech, ZenCrack). This workshop will be inscribed within the Multiscale Fracture (MultiFRAC FP7 IRSES (International Research Staff Exchange Scheme) exchange programme linking Weimar (Prof. Rabczuk, Germany), Cardiff (UK), Northwestern University (Prof. Ted Belytschko and colleagues, USA), University of Witwaterstrand (Dr. Muthu, South Africa), Indian Institute of Science Bangalore (Dr. Mahapatra, India), which will ensure dissemination beyond the UK.

The numerical developments will be carried out in MATLAB so as to allow a ``race to prototype'' for fast publication and module release as an Open Source toolbox (BSD(new) license). When a sufficient level of maturity of the MATLAB code is achieved, production codes will be developed by inuTech GmbH to ensure wide uptake of the work by industry and academia. Negotiations between the legal teams at Cardiff University and inuTech on licensing, intellectual property and non-disclosure are ongoing.

Soitec SA state in their letter of support that they will support the project financially by contributing to a President PhD Scholarship at Cardiff. The negotiations between the two parties' legal teams are still ongoing. This will ensure that the technology is directly applicable to the needs of high-end manufacturing industry, which is one of the focus directions to redress the UK Economy.

Publications

10 25 50

publication icon
Akbari Rahimabadi A (2015) Scale selection in nonlinear fracture mechanics of heterogeneous materials in Philosophical Magazine

publication icon
Claus S (2018) A stable and optimally convergent LaTIn-CutFEM algorithm for multiple unilateral contact problems in International Journal for Numerical Methods in Engineering

publication icon
Kerfriden P (2014) Certification of projection-based reduced order modelling in computational homogenisation by the constitutive relation error in International Journal for Numerical Methods in Engineering

 
Description Multiscale phenomena appear in most fields of engineering and physics. Their study is particularaly important in engineering applications related to the understanding and optimisation of a new generation of materials that are being designed over several scales (composite materials, active multi-level materials, self-healing materials, multi-functional materials). Nonetheless, the application of predictive simulation methods to the quantification of multiscale phenomena remains severely restricted by practical and theoretical challenges, notably the prohibitive numerical cost associated with nonlinear, multiscale simulations, and the lack of theoretical and practical approaches when the scales lose their separability (fracture in composites, transient dynamics in multi-level materials, ...). This award helped Cardiff's group of computational mechanics make significant academic contributions to these fundamental and cross-disciplinary challenges, as detailed below.
Feasible nonlinear numerical predictions over several scales. We have pioneered an approach that combines machine learning and high-fidelity multiscale simulations to allow for prohibitively expensive small-scale modelling features to be brought back to engineering scales at which computations are feasible, whilst controlling the loss of accuracy of the upscaling process. In the context of composite materials, we have shown gains of several orders of magnitude in terms of CPU time required to calculate the microscale damage accumulation within an engineering scale component. The novel inclusion of a machine learning component was key to our findings, as it allowed us to extract the relevant features of the non-intuitive microscale physics of the composite without prior bias, in a fully automatised and weakly problem-dependent manner.
Equation-free numerical upscaling. We have developed fundamentally new mathematical approaches to perform simulations over several scales. Instead of assuming the shape of engineering-scale equations, knowing the fine-scale physics, we have proposed to solve full microscale problems using a small number of "solution guesses" that contain relevant microscale information. The key of the proposed way of thinking is a probabilistic "blurring" of the equation-free multiscale techniques that had been traditionally reserved to computations over small, deterministically defined macroscale domains.
Multiscale computations with guarantees of accuracy. The accuracy of multiscale methods is poorly understood, and largely uncontrolled. We made significant progresses in the direction of control, by proposing an algorithm that guarantees the macroscale accuracy of engineering quantities of interest, approximated using well-known multiscale schemes. These advanced mathematical tools are limited to linear multiscale phenomena, but the resulting understanding may open doors for the control of accuracy of multiscale scheme applied to nonlinear, possibly unstable phenomena
Exploitation Route Although we have applied many of the methodologies that we have developed to the reliable and efficient multiscale prediction of damage in composite materials, our main findings are generic in nature. As such, research groups may use our contributions as stepping stones towards affordable, fully automatized and reliable simulation strategies for multiscale simulations. This may have important impact in various fields of engineering. For instance, our group is currently involved in computer-aided healthcare, where adequate multiscale strategies combined with machine learning techniques may bring to light a new generation of surgical simulators and computer-aided tools for surgical planning. As another example, we have started a collaboration with researchers and practitioners in the field of thermal ablation micro-manufacturing. Efficient multiscale predictions could be an an important contributing factor for the success of industry 4.0, through the active control and increased flexibility of current and future manufacturing technology,
Sectors Aerospace, Defence and Marine,Construction,Digital/Communication/Information Technologies (including Software),Energy,Manufacturing, including Industrial Biotechology,Transport

URL https://computengincardiff.blogspot.co.uk/2015/07/bayesian-optimisation-for-selection-of.html
 
Description H2020-MSCA-ITN-2016
Amount £250,000 (GBP)
Organisation European Commission H2020 
Sector Public
Country Belgium, Kingdom of
Start 09/2018 
 
Description National Research Network in Advanced Engineering and Materials - 2nd Call
Amount £59,349 (GBP)
Organisation National Research Network in Advanced Engineering and Materials 
Sector Academic/University
Country United Kingdom of Great Britain & Northern Ireland (UK)
Start 03/2015 
End 03/2018
 
Description National Research Network in Advanced Engineering and Materials - 2nd Call
Amount £81,952 (GBP)
Organisation National Research Network in Advanced Engineering and Materials 
Sector Academic/University
Country United Kingdom of Great Britain & Northern Ireland (UK)
Start 01/2014 
End 01/2016
 
Description National Research Network in Advanced Engineering and Materials - Fellowships Call
Amount £185,304 (GBP)
Organisation National Research Network in Advanced Engineering and Materials 
Sector Academic/University
Country United Kingdom of Great Britain & Northern Ireland (UK)
Start 03/2015 
End 03/2019
 
Description European Model Reduction Network (EU-MORNET) 
Organisation European Research Council (ERC)
Country European Union (EU) 
Sector Public 
PI Contribution This is a European network on model order reduction. I have been invited to join this network following the recognition of the work done at Cardiff on reduced order modelling for advanced materials. This consortium is in charge of coordinating future research in the area at an international level, and explore the possibilities for economic impact. My group contributes to this aim by providing an expertise in the field of computational material engineering. I have personally been invited to give a keynote lecture to the kick-off meeting, introducing an engineer view and expertise to this highly mathematical area.
Collaborator Contribution The consortium provides funding for research meetings and covered my travel expenses for the kick-off meeting.
Impact NA
Start Year 2011
 
Description Stanford Charbel Farhat research group 
Organisation Stanford University
Country United States of America 
Sector Academic/University 
PI Contribution I have spent 3 weeks in Stanford to develop new reduced order modelling algorithm for statistical inference. I have brought my expertise in error estimation for this type of problems.
Collaborator Contribution My collaborators in Stanford invited me for a research visit, during which I have worked with them.
Impact 3 papers are currently in preparation.
Start Year 2014