Bulletin of the American Physical Society
22nd Biennial Conference of the APS Topical Group on Shock Compression of Condensed Matter
Volume 67, Number 8
Monday–Friday, July 11–15, 2022; Anaheim, California
Session P03: Materials Properties and Design IFocus Recordings Available
|
Hide Abstracts |
Chair: Rebecca Lindsey, Lawrence Livermore Natl Lab Room: Anaheim Marriott Platinum 1 |
Wednesday, July 13, 2022 11:00AM - 11:30AM |
P03.00001: On the application of machine learning techniques to energetic materials Invited Speaker: Brian C Barnes As a data-driven approach, the application of machine learning (ML) to energetic materials problems faces many challenges. There is strong structural dissimilarity between many energetics and most molecules found in popular databases such as (but not limited to) the NIH PubChem database. Energetic material properties of interest, such as impact sensitivity or detonation pressure, can have large experimental uncertainties, with results varying by apparatus or unrecorded conditions. Energetic material datasets are typically very small compared to famous datasets used for other ML applications, such as those for image recognition. Energetic material chemistry is also unusual compared to typical pharmaceutical interests and may occur at extreme conditions. Nevertheless, the potential benefits of ML-driven models are significant. We will demonstrate their ability to create accurate, generalizable, fast-running correlations from complicated input data without prior assumption of a physical relationship. This talk will discuss results from a variety of architectures, including a physics-inspired descriptor coupled with a tree-based technique, a high-dimensional convolutional network applied to quantum mechanical data, and a graph-convolutional (message-passing) neural network. It will also discuss results from generative models and transfer learning approaches. Pros and cons of different approaches will be discussed. |
Wednesday, July 13, 2022 11:30AM - 11:45AM |
P03.00002: Establishing the structure-property-performance linkage of pressed energetic materials using physics-aware recurrent convolutional neural networks (PARC) Phong C Nguyen, Joseph Choi, Yen-Thi Nguyen, H.S. Udaykumar, Stephen Baek Establishing the structure-property-performance (SPP) linkages is a vital task for the design of energetic materials (EM). However, the current approaches for establishing SPP linkages are limited by time-consuming and costly physical and numerical experiments, rendering practical challenges in exploring the vast design space of EM microstructures. In this work, we propose a novel deep learning method called physics-aware recurrent convolutional neural network (PARC), which can assimilate the thermo-mechanics of hotspot ignition and growth in shocked heterogeneous EM microstructures. PARC is designed to predict the time evolution of temperature and pressure fields by modeling and solving the governing differential equations using convolutional neural networks (CNN). In contrast to other machine learning approaches, the unique recurrent convolutional architecture modeling the governing differential equations makes PARC highly interpretable and “physics-aware.” The validation results show that PARC can predict the thermomechanical behavior of shock-induced EM microstructures with high accuracy (within 5% error) compared to direct numerical simulation (DNS) results, despite a dramatic reduction of computation time (up to 3000 times). Furthermore, we also show that the interpretable architecture of PARC provides additional lenses for the study of SPP linkages by shedding light on identifying the morphological characteristic of microstructures that lead to energy localization and initiation. The impact of the current work is a novel capacity to estimate SPP linkages in a significantly quicker turnaround, enabling the design of EM microstructures with engineered properties. |
Wednesday, July 13, 2022 11:45AM - 12:00PM |
P03.00003: Interpretable Performance Models for Energetic Materials using Parsimonious Neural Networks Robert J Appleton, Peter Salek, Alex D Casey, Steven F Son, Brian C Barnes, Alejandro H Strachan Predictive models for the performance of explosives and propellants are important for their design, optimization, and safety. Thermochemical codes can predict certain properties from fundamental properties such as density and formation energies that can be obtained from first principles. The Kamlet-Jacobs equations provide a computationally inexpensive alternative, but still require these fundamental properties as inputs and are limited in their ability to generalize to different types of explosives. Such easy to evaluate models are desirable for the efficient screening of large numbers of candidate materials, beyond what is possible with computationally intensive methods. Therefore, we use parsimonious neural networks (PNNs) to learn interpretable models for the detonation velocity and pressure for explosives using data collected from open literature. PNNs use evolutionary optimization to create models that balance accuracy and complexity. For both detonation velocity and pressure, we establish a family of interpretable models that are pareto optimal in accuracy and simplicity space. The Kamlet-Jacobs models lie close to but not at the pareto front. We extract expressions from these models and draw conclusions based on the functional forms of the terms discovered. |
Wednesday, July 13, 2022 12:00PM - 12:15PM |
P03.00004: Chemical determinants of drop-weight impact sensitivity in high explosives Frank Marrs, Alexandra Burch, Suyana Ferreira, Jack Davis, Nicholas Lease, Marc J Cawkwell, Virginia W Manner A common experiment to evaluate handling sensitivity of high explosives is the drop-weight impact test. Although this test is known to be noisy, it is one of the most common tests used to evaluate handling sensitivity. In this paper, we compile the largest known data set of drop-weight impact sensitivity test results (mainly performed at Los Alamos National Laboratory), along with a compendium of molecular and chemical properties for the explosives under test. These data consist of over 1,000 unique explosives and over 100 properties. We use random forest methods to estimate a model of explosive handling sensitivity as a function of chemical and molecular properties of the explosives under test. Our model predicts well across a wide range of explosive types, spanning a broad range of explosive performance and sensitivity. We find that properties related to explosive performance, such as heat of explosion, oxygen balance, and functional group, are highly predictive of explosive handling sensitivity. Yet, models that omit some of these properties still perform well. Our results suggest that there is not one, or even several, factors that explain explosive handling sensitivity, but that there are many complex, interrelated effects at play. |
Wednesday, July 13, 2022 12:15PM - 12:30PM |
P03.00005: Data-driven Chemical Property Models for Energetic Materials using Transfer Learning Joshua Lansford, Brian C Barnes, Betsy M Rice, Klavs F Jensen Due to the hazardous nature of energetic materials, it is useful to have accurate estimates of physical properties related to their handling, such as impact sensitivity and vapor pressure. Unfortunately, many safety-related properties depend on multiscale interactions and cannot be directly computed with high accuracy. By themselves, physics-based property prediction models do not extrapolate well and can fail entirely. While machine learning (ML) can overcome these limitations, ML requires large datasets that are not available for energetic properties. Here, we apply two different transfer learning approaches to predict impact sensitivity and vapor pressure. In the first approach, model parameters are learned to map a chemical graph to properties that can be directly computed, and then these parameters are used to predict impact sensitivity. Specifically, we co-train a directed-message passing neural network on a diverse dataset in order to predict impact sensitivity. In the second approach, we embed a physical model into the neural network to enable extrapolation and improve out-of-sample prediction accuracy for energetic vapor pressures. Our models outperform existing models on a diverse test set and are generalizable. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700