“策之不以其道，食之不能尽其材，鸣之而不能通其意，... 其真不知马也。” 韩愈《马说》

Algorithms: physics- & data-driven for materials science and manufacturing

big data, machine learning, artificial intelligence <--> novel materials and/or structures <--> energy, environment

In the past, my research was mostly physics-driven. I created and implemented various effective, scalable numerical schemes for nontrivial transport phenomena. While physics will always be a beautiful driver...

now I'm extending to ride the big-data wave, adding data-driven flavor to my research. My interest in this aspect has been to architect, develop and automate data and analytics solutions, which manipulate and analyze large data sets to distill insights from data to compensate analytical thinking. Meanwhile, we're concentrating ourselves on materials informatics, using artificial intelligence to discover new materials. For instance, my recent work developed a framework to automatically search the polyaromatic hydrocarbon for solar thermal fuels and carbon fiber processing. With this, currently I'm completing the framework with the capability to model the manufacturing processes of carbon fibers, which is a multi-institute project supported by DOE.

2.1 Development of computational schemes/packages

The methods we've developed and implemented ranges from finite difference/volume method, to molecular dynamics and discrete element method, and to stochastic simulations.

Theoretical and numerical studies of noncontinuum gas-phase heat conduction in micro/nano devices (Numerical Heat Transfer, Part B: Fundamentals, 2010)
Theoretical Two-Dimensional Modeling of Gas Conduction Between Finite Parallel Plates in High Vacuum (Journal of Heat Transfer, 2012)
Multiple temperature kinetic model and its applications to micro-scale gas flows (Computers & fluids, 2012)
DEM codes acknowledged in Mean-field theory of random close packings of axisymmetric particles (Nature communications, 2013)
A unified Monte Carlo framework for simulations of transport phenomena of particles, phonons, electrons, photons, and beyond (in preparation)

2.2 Material informatics

Novel machine learning and big data techniques for materials discovery

The science of learning plays a key role in the fields of statistics, data mining and artificial intelligence, intersecting with areas of engineering and many other disciplines. We take advantage of the fast-developing learning techniques to discover new materials. Complementing the physical principles that have been developed for hundreds of years, we're attempting to learn from voluminous data.

2.3 Manufacturing modeling

Currently working on the processing of carbon fibers.

2.4 Uncertainty quantification

Uncertainty quantification of various computational approaches, such as MD and DFT

We concern both the forward propagation of uncertainty (where the various sources of uncertainty are propagated through the model to predict the overall uncertainty in the system response) and the inverse assessment of model uncertainty and parameter uncertainty (where the model parameters are calibrated simultaneously using test data). There has been a proliferation of research on the former problem and a majority of uncertainty analysis techniques were developed for it. On the other hand, the latter problem is drawing increasing attention in the engineering design community, since uncertainty quantification of a model and the subsequent predictions of the true system response(s) are of great interest in designing systems or building models. We combine a series of methods, such as Monte Carlo, polynomial chaos expansion, and Bayesian approach, to assess existing models and build more accurate models.