(3m) Data-Driven Modeling in Chemical Engineering

A convergent stream of big data, modeling and analytics enabled by high fidelity experimentation and computation creates a unique opportunity for the discovery of novel chemical physics. I will present my work on learning from microscopic images, diffraction and molecular simulations using the techniques of inversion and Bayesian statistics.

Examples include 1) extracting unknown functional forms of reaction kinetics and free energy, surface kinetic map and chemo-mechanical coupling from images of lithium iron phosphate (LFP) particles, 2) identifying the role of autocatalytic reaction kinetics in the phase behavior of a transition metal oxide (NMC) and its quantification using a combination of X-ray diffraction, microscopy and electrochemistry data, and 3) learning the concentration-dependent free energy and diffusivity from molecular simulations of phase separation via the Cahn-Hilliard equation. All examples above achieved quantitative matching between data and models and trained models are predictive on test data. The methods can potentially be used to discover quantitative models for complex systems such as soft matter, energy materials, biology that are central to chemical engineering applications.

Research Interests

Applied mathematics, modeling and computation, data analytics, energy

Teaching Interests

Numerical methods, electrochemistry, transport phenomena

References

[1] Zhao, H. & Bazant, M. Z. Population dynamics of driven autocatalytic reactive mixtures. Physical Review E 100, 012144 (2019).

[2] Zhao, H., Storey, B. D., Braatz, R. D. & Bazant, M. Z. Learning the Physics of Pattern Formation from Images. Physical Review Letters 124, 060201 (2020).

[3] Park, J., Zhao, H., Kang, S. D., Lim, K., Chen, C., Yu, Y., Braatz, R. D., Shapiro, D. A., Hong, J., Toney, M. F., Bazant, M. Z., Chueh, W. C. Fictitious Phase Separation in Li Layered Oxides Driven by Electro-Autocatalysis, submitted