Prof. Dr.-Ing. Bojana Rosic
Stochastic multiscale analysis of nonlinear materials indlucing uncertainties
Many man made and naturally occurring materials exhibit spatial heterogeneity and randomness with regards to their features on fine-scale. This gives rise to many interesting research problems to be investigated particularly in the field of computational mechanics, e.g. characterization of fine-scale randomness and investigation of its effect on the structural behaviour through numerical simulations that take scale-coupling into account, inferring material parameters of low-fidelity phenomenological models using measurements given high-fidelity fine-scale model responses, quantification of uncertainty of irreversible phenomena etc. With the recent advancement of probabilistic numerics and machine/deep learning, there is an increased interest in incorporating stochastic considerations into above mentioned research problems.
Figure 1: Probabilistic multiscale approach (left), upscaled bulk modulus w.r.t. correlation length of fine scale model
The main focus of this research project is to develop efficient numerical methods for the computational prediction of such materials by taking into account ideas from stochastic spectral methods, low-rank tensor approximations and proper generalised decomposition, non-intrusive computations, Bayesian identification etc. By building computational replica of physical assests, the so-called digital twins, this project will allow virtual testing of existing or design of new engineering materials.
Figure 2: 3D representative volume element and its discretisation
The project is done in cooperation with Prof. Adnan Ibrahimbegovic, Chair for Computational Mechanics (Mechanics), Sorbonne Universités/ Univ. Technology Compiègne: Compiègne, Ile-de-France/ Picardie, and doctoral students of Institute of Scientific Computing: Ehsan Adeli , Simona Dobrilla, Muhammad Sarfaraz Sadiq and Sharana Kummar Shivanand .
Publikationen im Rahmen des GRK:
Veröffentlichungen in wissenschaftlichen Zeitschriften mit review:
B. Rosić, J. Sỳkora, O. Pajonk, A. Kučerová and H.G. Matthies. Comparison of Numerical Approaches to Bayesian Updating. Computational Methods for Solids and Fluids. Multiscale Analysis, Probability Aspects and Model Reduction, Adnan Ibrahimbegović (Eds.), 427-462, Springer, Berlin, 2016.
E. Adeli, B. Rosić, H.G. Matthies and S. Reinstädler. Effect of Load Path on Parameter Identification for Plasticity Models using Bayesian Methods. QUIET, 2018.
V. Dunić, N. Busarac, V. Slavković, B. Rosić, R. Niekamp, H.G. Matthies, R. Slavković and Živković. A Thermo-Mechanically Coupled Finite Strain Model Considering Inelastic Heat Generation. Continuum Mechanics and Thermodynamics, 28: 993-1007, 2016.
H.G. Matthies, E. Zander, B. Rosić and A. Litvinenko. Parameter Estimation via Conditional Expectation - A Bayesian Inversion. Advanced Modeling and Simulation in Engineering Sciences, 3:24, 2016.[ DOI ]
H.G. Matthies, E. Zander, B. Rosić, A. Litvinenko and O. Pajonk. Inverse Problems in a Bayesian Setting. Computational Methods for Solids and Fluids. Multiscale Analysis, Probability Aspects and Model Reduction, Adnan Ibrahimbegović (Eds.), 245-286, Springer, Berlin, 2016.
M.S. Sarfaraz, B. Rosić, H.G. Matthies and A. Ibrahimbegović. Stochastic upscaling via linear Bayesian updating. Coupled systems mechanics, 7(2): 211-231, 2018.
M.S. Sarfaraz, B. Rosić, H.G. Matthies and A. Ibrahimbegović. Stochastic Upscaling via Linear Bayesian Updating, In book: Multiscale Modeling of Heterogeneous Structures. J. Sorić, Jurica, P. Wriggers, and O. Allix (Eds.), 163-181, Springer, 2018.
J. Waeytens, B. Rosić, P.-E. Charbonnel, E. Merliot, D. Siegert, X. Chapeleau, R. Vidal, V. le Corvec and L.-M. Cottineau. Model updating techniques for damage detection in concrete beam using optical fiber strain measurement device. Engineering Structures, 129:2-10, 2016.
Konferenzbeiträge mit Veröffentlichung:
B. Rosić, S.K. Shivanand, T.V. Hoang and H.G. Matthies. Iterative multilevel spectral identification of bone macroscopic properties described by probability box. PAMM 2018(1), pp. 1-4, 2018.
B. Rosić, M.S. Sarfaraz and H.G. Matthies. Stochastic upscaling of random microstructures. PAMM, 17: 869-870, 2017.
B. Rosić and H.G. Matthies. Sparse Bayesian polynomial chaos approximations of elasto-plastic material models. COMPLAS 2017 Proceedings, E. Oñate, D.R.J. Owen, D. Peric and M. Chiumenti (Eds.), 256-267, CIMNE Barcelona, Spain, 2017.
E. Adeli, B. Rosić, H.G. Matthies and S. Reinstädler. Bayesian Parameter Identification in Plasticity. XIV International Conference on Computational Plasticity, COMPLAS 2017.
T.V. Hoang, B. Rosić and H.G. Matthies. Characterisation and propagation of uncertainties accociated with limited data using a hierarchical parameteric probability box. PAMM 2018(1), pp. 1-2, 2018.
Konferenzteilnahmen mit eigenem Beitrag:
B. Rosić and H.G. Matthies. Spectral nonlinear Kalman filtering, GAMM Annual Meeting, Weimar, Germany, March 2017.
B. Rosić and H.G. Matthies. Conditional expectation based spectral Bayesian filter for time dependent and nonlinear systems, WCSMO 2017, Braunschweig, Germany, June 2017.
B. Rosić and H.G. Matthies. Nonlinear Kalman filtering, UNCECOMP 2017, Rhodes Island, Greece, June 2017.
B. Rosić, M. Sadiq Sarfaraz, H.G. Matthies and A. Ibrahimbegović. A computational perspective on stochastic upscaling in nonlinear multiscale problems, ECCOMAS MSF, Ljubljana, Slovenia, September 2017.
B. Rosić and H.G. Matthies. Sparse Bayesian polynomial chaos approximations of elasto-plastic material models, COMPLAS 2017, Barcelona, Spain, September 2017.
B. Rosić. Hierarchical uncertainty quantification of elasto-plastic material models, GAMM Annual Meeting, Münich, Germany, March 2018.
B. Rosić. Sparse spectral Bayesian learning of nonlinear models, SIAM Conference on Uncertainty Quantification, Garden Grove, California, USA, April 2018.