DEEP REINFORCEMENT LEARNING FOR DE NOVO THERMOSETTING POLYMER DESIGN
Abstract
Development of thermosetting polymers such as epoxies has traditionally have been experimentally driven, trial-and-error process, guided by experience and intuition. This approach, however, is limited to small-scale studies, which may easily miss promising compounds. In addition, automation of organic molecules and materials design is considerably under-developed due to challenges associated with searching the vast design space, defined by the almost infinite combinations of molecular constituent, microstructures and synthesis conditions. We aim to formulate a novel, deep reinforcement leaning approach to computationally design next-generation thermoset polymers with unprecedented yet predictable combinations of properties. Specifically, we will use experimentally validated molecular dynamics simulations to perform high-throughput virtual screening on thermosetting polymers, thereby building a database for different epoxies. Next, we will establish the synthesis-structure-property relationships for thermosetting polymers through advanced microstructure characterization and machine learning techniques. Then, we will develop an ‘on-demand’ Bayesian optimization approach integrated with deep reinforcement learning that can effectively explore a design space that guide adaptive data collection towards achieving specific properties targets of thermosetting polymers. Finally, we will experimentally synthesize these thermosetting polymers with thermomechanical testing, which can be used to validate the suite of data, models and tools. If successful, the proposed work can address a wide range of scientific questions in computational materials design and synthesis-structure-property relationship. This work will also benefit the broader technical community that are interested in developing new thermoset polymers with tailored thermomechanical properties, which are important for many DoD applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2021
- Source ID
- FA95502010183
Entities
People
- Ying Li
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Connecticut