Predicting the mechanical properties of disordered materials by exploiting Bayesian Machine Learning

by Prof. Mikko Alava (Aalto University, Finland)

Monday 12 Oct 2026, 15:30 → 16:30 Europe/Rome

P2B (Dipartimento di Fisica e Astronomia - Edificio Ricci-Curbastro (ex-Paolotti))

Predicting the mechanical response of disordered materials is challenging because structure-property relations are noisy, datasets are often small, and yielding or failure may emerge abruptly. I discuss how Bayesian machine learning can turn these limitations into uncertainty-aware predictions. First, rheological measurements provide a low-dimensional route to predict and optimize the mechanical properties of bio-based foams. I then present unpublished work extending Bayesian inference to complete foam stress-strain behavior. For metallic glasses, Bayesian optimization explores composition space for target mechanical properties, while physics-informed Bayesian inference predicts plastic-strain growth and yielding from the early mechanical response. Finally, evolving activation barrier landscapes inferred from acoustic-emission data allow creep lifetime to be predicted well before failure.

Selected papers:

1. I. Y. Miranda-Valdez et al., "Accelerated design of solid bio-based foams for plastics substitutes," Materials Horizons 12, 1855-1862 (2025).
2. T. Makinen, A. D. S. Parmar, S. Bonfanti & M. J. Alava, "Bayesian exploration of the composition space of CuZrAl metallic glasses for mechanical properties," npj Computational Materials 11, 96 (2025).
3. T. Makinen, A. D. S. Parmar, S. Bonfanti & M. J. Alava, "Growth and prediction of plastic strain in metallic glasses," Physical Review Materials 10, 025601 (2026).
4. J. C. Verano-Espitia, T. Makinen, M. J. Alava & J. Weiss, "Early Prediction of Creep Failure via Bayesian Inference of Evolving Barriers," arXiv:2603.16419 (2026).

In collaboration at Aalto with: Isaac Y. Miranda-Valdez, Tero Makinen, Juha Koivisto,  Juan Carlos Verano-Espitia, Jerome Weiss