An official website of the United States government
Abstract Over the last decades, remote observation tools and models have been developed to improve the forecasting of ash‐rich volcanic plumes. One challenge in these forecasts is knowing the properties at the vent, including the mass eruption rate and grain size distribution (GSD). Volcanic lightning is a common feature of explosive eruptions with high mass eruption rates of fine particles. The GSD is expected to play a major role in generating lightning in the gas thrust region via triboelectrification. Here, we experimentally investigate the electrical discharges of volcanic ash as a function of varying GSD. We employ two natural materials, a phonolitic pumice and a tholeiitic basalt (TB), and one synthetic material (soda‐lime glass beads [GB]). For each of the three materials, coarse and fine grain size fractions with known GSDs are mixed, and the particle mixture is subjected to rapid decompression. The experiments are observed using a high‐speed camera to track particle‐gas dispersion dynamics during the experiments. A Faraday cage is used to count the number and measure the magnitude of electrical discharge events. Although quite different in chemical composition, TB and GB show similar vent dynamics and lightning properties. The phonolitic pumice displays significantly different ejection dynamics and a significant reduction in lightning generation. We conclude that particle‐gas coupling during an eruption, which in turn depends on the GSD and bulk density, plays a major role in defining the generation of lightning. The presence of fines, a broad GSD, and dense particles all promote lightning.
more »
« less
Constructing and Compressing Global Moment Descriptors from Local Atomic Environments
Local atomic environment descriptors (LAEDs) are used in the materials science and chemistry communities, for example, for the development of machine learning interatomic potentials. Despite the fact that LAEDs have been extensively studied and benchmarked for various applications, global structure descriptors (GSDs), i.e., descriptors for entire molecules or crystal structures, have been mostly developed independently based on other approaches. Here, we propose a systematically improvable methodology for constructing a space of representations of GSDs from LAEDs by incorporating statistical information and information about chemical elements. We apply the method to construct GSDs of varying complexity for lithium thiophosphate structures that are of interest as solid electrolytes and use an information-theoretic approach to obtain an optimally compressed GSD. Finally, we report the performance of the compressed GSD for energy prediction tasks.
more »
« less
- Award ID(s):
- 1940290
- PAR ID:
- 10477154
- Publisher / Repository:
- OpenReview.net
- Date Published:
- Journal Name:
- Workshop on "Machine Learning for Materials" ICLR 2023
- Format(s):
- Medium: X
- Location:
- https://openreview.net/forum?id=4Hl8bjobpl9
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The dust grain size distribution (GSD) likely varies significantly across star-forming environments in the Universe, but its impact on star formation remains unclear. This ambiguity arises because the GSD interacts nonlinearly with processes like heating, cooling, radiation, and chemistry, which have competing effects and varying environmental dependencies. Processes such as grain coagulation, expected to be efficient in dense star-forming regions, reduce the abundance of small grains and increase that of larger grains. Motivated by this, we investigate the effects of similar GSD variations on the thermochemistry and evolution of giant molecular clouds (GMCs) using magnetohydrodynamic simulations spanning a range of cloud masses and grain sizes, which explicitly incorporate the dynamics of dust grains within the full-physics framework of the STARFORGE project. We find that grain size variations significantly alter GMC thermochemistry: the leading-order effect is that larger grains, under fixed dust mass, GSD dynamic range, and dust-to-gas ratio, result in lower dust opacities. This reduced opacity permits interstellar radiation field and internal radiation photons to penetrate more deeply. This leads to rapid gas heating and inhibited star formation. Star formation efficiency is highly sensitive to grain size, with an order-of-magnitude reduction when grain size dynamic range increases from 10−3–0.1μm to 0.1–10μm. Additionally, warmer gas suppresses low-mass star formation, and decreased opacities result in a greater proportion of gas in diffuse ionized structures.more » « less
-
Hypergraph neural networks (HyperGNNs) are a family of deep neural networks designed to perform inference on hypergraphs. HyperGNNs follow either a spectral or a spatial approach, in which a convolution or message-passing operation is conducted based on a hypergraph algebraic descriptor. While many HyperGNNs have been proposed and achieved state-of-the-art performance on broad applications, there have been limited attempts at exploring high-dimensional hypergraph descriptors (tensors) and joint node interactions carried by hyperedges. In this article, we depart from hypergraph matrix representations and present a new tensor-HyperGNN (T-HyperGNN) framework with cross-node interactions (CNIs). The T-HyperGNN framework consists of T-spectral convolution, T-spatial convolution, and T-message-passing HyperGNNs (T-MPHN). The T-spectral convolution HyperGNN is defined under the t-product algebra that closely connects to the spectral space. To improve computational efficiency for large hypergraphs, we localize the T-spectral convolution approach to formulate the T-spatial convolution and further devise a novel tensor-message-passing algorithm for practical implementation by studying a compressed adjacency tensor representation. Compared to the state-of-the-art approaches, our T-HyperGNNs preserve intrinsic high-order network structures without any hypergraph reduction and model the joint effects of nodes through a CNI layer. These advantages of our T-HyperGNNs are demonstrated in a wide range of real-world hypergraph datasets. The implementation code is available at https://github.com/wangfuli/T-HyperGNNs.git.more » « less
-
Abstract Streamflow prediction is a long‐standing hydrologic problem. Development of models for streamflow prediction often requires incorporation of catchment physical descriptors to characterize the associated complex hydrological processes. Across different scales of catchments, these physical descriptors also allow models to extrapolate hydrologic information from one catchment to others, a process referred to as “regionalization”. Recently, in gauged basin scenarios, deep learning models have been shown to achieve state of the art regionalization performance by building a global hydrologic model. These models predict streamflow given catchment physical descriptors and weather forcing data. However, these physical descriptors are by their nature uncertain, sometimes incomplete, or even unavailable in certain cases, which limits the applicability of this approach. In this paper, we show that by assigning a vector of random values as a surrogate for catchment physical descriptors, we can achieve robust regionalization performance under a gauged prediction scenario. Our results show that the deep learning model using our proposed random vector approach achieves a predictive performance comparable to that of the model using actual physical descriptors. The random vector approach yields robust performance under different data sparsity scenarios and deep learning model selections. Furthermore, based on the use of random vectors, high‐dimensional characterization improves regionalization performance in gauged basin scenario when physical descriptors are uncertain, or insufficient.more » « less
-
Kokossis, A.; Georgiadis, M.C.; Pistikopoulos, E.N. (Ed.)A quantitative model can play an essential role in controlling critical quality attributes of products and in designing the associated processes. One of the challenges in designing a dry granulation process is to find the optimal balance between improving powder flowability and sacrificing powder tabletability, both of which are highly affected by ribbon solid fraction and granule size distribution (GSD). This study is focused on developing a hybrid machine learning (ML)-assisted mechanistic model to predict ribbon solid fraction, GSD, and throughput for the purpose of implementing model predictive control of an integrated continuous dry granulation tableting process. It is found that the predictability of ribbon solid fraction and throughput are improved when modification is made to Johanson’s model by incorporating relationships between roll compaction parameters and ribbon elastic recovery. Such relationships typically are either not considered or assumed to be a constant in the models reported in the literature. To describe the nature of the bimodal size distribution of roller compactor granules instead of only using traditional 𝐷𝐷10, 𝐷𝐷50 and 𝐷𝐷90 values, the GSD is represented by a bimodal Weibull distribution with five fitting parameters. Furthermore, these five GSD parameters are predicted by ML models. The results indicate the ribbon solid fraction and screen size are the two most significant factors affecting GSD.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
