Topic: “Dinky, Dirty, Dynamic & Deceptive Data (1)”: An overview of hybrid machine learning and equation-based modelling
Speaker: Dr. Eleni Koronaki, University of Luxembourg, Department of Computational Science
Time: Wednesday 2020.10.21, 10:00 CET
In the era of “Big Data”, machine learning frameworks are attractive candidates for leveraging abundant data and transforming it into meaningful information. Despite the success of methods such as Deep Neural Networks in diverse sectors, ranging from finance to healthcare and language recognition, to name just a few, their implementation in traditional engineering fields is not universal. The reason is twofold: Firstly, first principles-based models, albeit computationally expensive remain consistent decision-making tools, more so now with the evolution of computational algorithms and infrastructure. Secondly, in many applications, the available data is not “big” enough to ensure accuracy and reliability of machine learning workflows. This dichotomy has not been unnoticed in the engineering community and various efforts to address this have been published, surprisingly as early as in the early 90s.
Nowadays the advent of Physics-Informed Neural Networks (PINNs) revisits older concepts with remarkable results. In this presentation, an illuminating overview is attempted of the “hybrid physics-informed machine learning” paradigm.
Speaker: Saurabh Deshpande, University of Luxembourg, Department of Computational Science
Time: Wednesday 2020.10.14, 10:00 CET
Since a decade, machine learning has started to revolutionize several fields due to the development of new algorithms and the availability of more data every day. Deep learning, a class of machine learning methods based on learning data representations, has demonstrated strong abilities at extracting high-level representations of complex processes. In this work, we will implement a particular class of machine learning architecture called Convolutional Neural Network (CNN) to replace the finite element solver for 3D hyper-elastic simulations. Also, we will briefly touch upon the dropout technique and its possible use in predicting uncertainties of the neural network predictions.
Topic: Bayesian Neural Networks for uncertainty estimation on regression problems
Speaker: Vasilis Krokos, University of Cardiff & Synopsys-Simpleware
Time: Wednesday 2020.10.07, 10:00 CET
Although nowadays Neural Networks (NN) are vastly used in numerous applications, traditionally they lack a very important characteristic. They fail to incorporate uncertainty into their predictions. NNs are notoriously known to extrapolate really badly, something that can lead to catastrophic consequences. It is obvious that it is essential to incorporate uncertainty into any model used for critical tasks. A common example of this type of model is a Gaussian Process (GP). The GP is a stochastic regression model that outputs a mean prediction along with credible intervals, thus quantifying the uncertainty. Unfortunately, GPs are known to scale very badly to large datasets and can’t be used in tasks like classification, segmentation, etc where the inputs are images while Convolutional Neural Networks clearly dominate this field. In this work we will discuss how we can transform any deterministic NN into a probabilistic one, using the Bayes by Backpropagation method [Blundell et al., 2015], and we will test this NN into a simple 1D regression case.
The computational mechanic’s group of Prof. Stéphane Bordas (www.legato‐team.eu, www.uni.lu ) is searching for two Ph.D. students to work on one of the following projects. Both projects focus on
phase-field damage modeling, rubbery polymers, and experimental validation. All work is carried out in close cooperation with the industrial partner SISTO Armaturen S.A. (www.sisto-aseptic.com,
The project aims to analyze the formation and growth of micro-cracks in rubber due to fatigue loading. Recently published articles on fatigue failure of rubber show that micro-cracks nucleate at
initially present natural flaws. The initial distribution of flaws and the micro-crack growth is investigated with CT-scans and interrupted fatigue tests. These results are used to refine a fatigue
phase-field damage model for this scale. Key points are:
Measure the initial flaw size with CT-scans for various process parameter and compounds
Interrupted fatigue tests with CT-scans to study the micro-crack growth
Fatigue tests with samples with artificially introduced flaws
Refine a fatigue phase-field damage model for micro-scale simulations
Implementation of a fatigue phase-field fatigue model
Parameter calibration and validation of the numerical model with experimental data
This project focuses on the thermal aging of rubber. An Arrhenius law is often applied to the test data to predict behavior at lower temperatures. However, this implies that the degradation mechanism at
the highest temperature is the same as at the lower temperatures. A new thermo-chemical model should be developed to study the degradation of material properties over the entire temperature
range. This model is combined with a recently published fatigue phase-field damage model to study the influence of aging on the failure of rubber parts. Key points are:
Experimental study of thermal aging taking into account the influence of temperature,
media, and sample thickness
Development of a numerical model to predict the influence of aging on the material
properties of rubber
Couple the new model to a fatigue phase-field damage model
Parameter calibration and validation of the numerical model with experimental data
Four years of funding for the Ph.D. student is provided with a competitive salary. Funds to attend conferences and summer schools are available. Each student will be employed by the university, but
she/he will also spend a significant amount of time in the company.
Aspiring researchers interested in a Ph.D. topic with strong industrial relevance are encouraged to apply. Only those who hold an MSc degree or will hold one in the near future will be considered. We
are specifically looking for those who hold an MSc degree in Engineering. A background in and affinity with some of the following fields is required:
Material/constitutive modeling and structural mechanics
have obtained a PhD degree during the last 8 years (prior to the application year deadline) or be in the last year of the thesis work with an outstanding academic record. Nevertheless, before starting the grant, a proof of the PhD degree will be requested.
complete, submit the application form and send via the online system:
We would like to bring to your attention our upcoming IWR School which will focus on machine
learning with applications from Natural Sciences and Life Sciences. The school will take place
from September 23 to 27, 2019, in Heidelberg, and addresses young researchers (PhD,
PostDoc, Master) from Natural and Life Sciences who want to learn more about machine
A background in Machine Learning is not required. Besides introducing the basic concepts we
teach selected topics in more depth, such as deep learning, metric learning, transfer learning,
Bayesian inverse problems, and causality. Experts from machine learning, Natural Science and
Life Science explain how these machine learning approaches are utilized to solve problems in
their respective fields of research.
Further information on the IWR School 2019 are available at:
It was a pleasure to welcome Michael Ortiz from the California Institute of Technology to give a seminar on the topic of “Model-Free Data-Driven Computing” at the University of Luxembourg. You can watch the entire seminar below.
We develop a new computing paradigm, which we refer to as Data-Driven Computing, according to which calculations are carried out directly from experimental material data and pertinent kinematic constraints and conservation laws, such as compatibility and equilibrium, thus bypassing the empirical material modeling step of conventional computing altogether. Data-driven solvers seek to assign to each material point the state from a prespecified data set that is closest to satisfying the conservation laws. Equivalently, data-driven solvers aim to find the state satisfying the conservation laws that is closest to the data set. The resulting data-driven problem thus consists of the minimization of a distance function to the data set in phase space subject to constraints introduced by the conservation laws. We demonstrate the data-driven paradigm and investigate the performance of data-driven solvers by means of several examples of application, including statics and dynamics of nonlinear three-dimensional trusses, and linear and nonlinear elasticity. In these tests, the data-driven solvers exhibit good convergence properties both with respect to the number of data points and with regard to local data assignment, including noisy material data sets containing outliers. The variational structure of the data-driven problem also renders it amenable to analysis. We find that the classical solutions are recovered in the case of linear elasticity. We identify conditions for convergence of Data-Driven solutions corresponding to sequences of approximating material data sets. Specialization to constant material data set sequences in turn establishes an appropriate notion of relaxation. We find that relaxation within the Data-Driven framework is fundamentally different from the classical relaxation of energy functions. For instance, we show that in the Data-Driven framework the relaxation of a bistable material leads to effective material data sets that are not graphs. I will finish my presentation with highlights on work in progress, including closed-loop Data-Driven analysis and experiments, Data-Driven molecular dynamics, Data-Driven inelasticity and publicly-editable material data repositories and data management from a Data-Driven perspective.
Privacy & Cookies Policy
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.