Want to make creations as awesome as this one?

Transcript

Machine Learning as a Bridge: Enhanced High-Throughput Materials Screening Considering Application Scale

All-FACET Meeting

Associate Professor in Process Systems EngineeringDepartment of Chemical Engineeringidelfonso.b.d.r.nogueira@ntnu.no +47 73592399 Kjemi 4, 232, Gløshaugen

A short tal by Idelfonso Nogueira

15

L4

L3

L2

L1

Concise Integration: Developing a new standpoint of Chemical Engineering

Bridging two scales L1+L2

X High computational cost

High predictive capacity for several systems

X Limited to the system in analysis

X Time demanding

Reliable

Modeling

Estimation

Ab initio molecular simulations

Experimental approach

Bridging two scales L1+L2

14

Encoding molecular and materials structures.

Creating a data base.

Challenges

Reverse Engineering is easier, i.e., generative models.

Advantages

Molecular strucutures can be encoded as inputs

If ML is used, only computational resources are used to build the data base

Approach Ongoing Research

Bridging two scales L1+L2

Bridging two scales L1+L2

General database

Bridging two scales L1+L2

Bridging two scales L1+L2

Bridging two scales L1+L2

Bridging two scales L1+L2

Generative Neural Networks + Reinforcement learning + Process Scale

Bridging three scales

Generative Neural Networks + Reinforcement learning + Process Scale

Bridging three scales

Generative Neural Networks + Reinforcement learning + Process Scale

Bridging three scales

L2

Process Design & Modeling

L1

FundamentalScale

15

Idelfonso NogueiraAssociate ProfessorDepartment of Chemical Engineering Process Systems Engineering Research Groupidelfonso.b.d.r.nogueira@ntnu.no +47 73592399 Kjemi 4, 232, Gløshaugen

Let's discuss these topics Please feel free to ask or to get in touch!