Abbas Jafari

Short Bio

Abbas obtained B.Sc. and M.Sc. degrees both in Mechanical Engineering from Isfahan University of Technology (in 2006) and Amirkabir University of Technology (in 2009), respectively. He also received an M.Sc. degree in Computational Engineering from Ruhr University of Bochum (in 2019), where he worked on dynamics of flexible multibody systems in his thesis in cooperation with the BOSCH research center at Renningen, Germany. Thereafter he joined the German Federal Institute for Materials Research and Testing (BAM) in Berlin as a research assistant. In September 2020 he became an extern PhD candidate at IBK with a cooperation with BAM. His PhD topic is about Bayesian approaches in calibrating nonlinear material models with a particular focus on concrete material under damage behaviour.

Research

Although elaborate models for different materials have been developed, those models usually suffer from being too sensitive to their parameters. One often challenge is, therefore, to identify such model parameters in accordance with appropriate measured data collected from experiments. This is a starting point towards achieving a Digital-Twin representing the reality within a range of applicability.

The core of Abbas’ research is Bayesian Model Updating methods, which is supposed to cope with various uncertainties for example in model selection and noisy measurements. Furthermore, the procedure must be able to account for heterogeneity of brittle and quasi-brittle materials like concrete. One way to achieve this is trying to identify the displacement field (already too many random parameters) in relation to the heterogenous data (measured displacements). To this end, variational Bayesian methods are particularly considered in view of their high computational efficiency when it comes to the identification of too many parameters. Another important aspect of Abbas’ research is benefiting from spatially-dense data such as in-situ X-ray Computed Tomography (CT) or Digital Image Correlation (DIC) data. The former (CT-data) is in the form of volumetric images and first generated by running in-situ experiments under precisely controlled loading and boundary conditions. The raw images are then analysed by using the Digital Volume Correlation (DVC) technique, resulting in the calculation of field quantities like the displacement field over the domain, which is employed in the model calibration process. A major advantage of such dense data compared to other standard experimental data (like tension/compression tests) is that, it is much more informative, removing any necessity for doing a lot of standard experiments accounting for different loading scenarios.