NOMATEN HYBRID-SEMINAR September 16: Computationally Guided Experiments and Predictions: Accelerating Materials Research under Uncertainty

The development and characterization of new materials increasingly rely on a combination of experiments, physics-based simulations, and data-driven models. However, experiments can be costly and time-consuming, while high-fidelity computational models often require substantial computational resources. In addition, both experimental observations and model predictions are affected by uncertainty. Computational methods that efficiently combine these different sources of information can therefore help accelerate materials research while reducing experimental and computational costs.

In this talk, I will provide an overview of our work on computational approaches for prediction, model calibration, experimental design, and optimization under uncertainty. A particular focus will be on machine-learning-based surrogate models for efficient prediction and uncertainty quantification, and their integration with Bayesian calibration to infer uncertain model parameters from limited experimental data [1,2]. I will also introduce approaches for design of experiments [3] and Bayesian optimization for optimal experimental design [4-7], where information acquired from previous experiments or simulations is used to guide what should be explored next.

Drawing on examples from our work in materials modeling, characterization, processing, and experimental workflows, I will illustrate how computational methods can help connect models and experiments, quantify uncertainties, explore complex design spaces, and guide data acquisition. Depending on the application, this may involve combining information from different sources or levels of fidelity, accounting for experimental variability and noise, or iteratively selecting informative experiments.

Overall, the talk will focus on the underlying computational concepts and their potential integration into different research workflows rather than on a single materials system or application. The aim is to provide a basis for discussion and to identify opportunities where predictive modeling, uncertainty quantification, and computationally guided experimentation could complement ongoing experimental and computational research.
 

REFERENCES:

[1] Kennedy, M.; O’Hagan, A. Bayesian calibration of computer models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2001, 63, 425–464.
[2] C. Schenk, I. Romero, A framework for the Bayesian calibration of complex and data-scarce models in applied sciences, Archives of Computational Methods in Engineering, 2026. [Link]
[3] C. Schenk; M. Haranczyk, CASTRO – A novel constrained sampling method for efficient exploration in materials and chemical mixture design, Computational Materials Science, 2025, 252:1137804. [Link]
[4] B. Özdemir, M. Hernández-del-Valle, C. Schenk, D. Wang, M. Haranczyk, Bayesian Optimization Guiding the Experimental Mapping of the Pareto Front of Mechanical and Flame-Retardant Properties in Polyamide Nanocomposites, Advanced Intelligent Discovery 2(3), 2025, e202500054. [Link]
[5] M. Hernández-del-Valle, C. Schenk, L. Echevarría-Pastrana, B. Ozdemir, E. Dios-Lázaro, J. Ilarraza-Zuazo, D.-Y. Wang, and M. Haranczyk, Robotically automated 3D printing and testing of thermoplastic material specimens. Digital Discovery, 2023, 2(6):1969–1979. [Link]
[6] C. Schenk, M. Hernández-del-Valle, L. Calero-Lumbreras, M. Noack, M. Haranczyk, Noise-aware optimization in nominally identical manufacturing and measuring systems for high-throughput parallel workflows, Advanced Engineering Informatics, 76:104960, 2026. [Link]
[7] S. Zorkaltsev, M. Haranczyk, C. Schenk, Bayesian Optimization of Genetic Algorithm Hyperparameters in a Multi-Fidelity Framework for Efficient Lattice Material Design. arXiv:2607.07289, 2026. [Link]