Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Transformer model”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 55 records · Page 3
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

SYMBA: symbolic computation of squared amplitudes in high energy physics with machine learning

The cross section is one of the most important physical quantities in high-energy physics and the most time consuming to compute. While machine learning has proven to be highly successful in numerical calculations in high-energy physics, analytical calculations using machine learning are still in their infancy. In this work, we use a sequence-to-sequence model, specifically, a transformer, to compute a key element of the cross section calculation, namely, the squared amplitude of an interaction. We show that a transformer model is able to predict correctly 97.6% and 99% of squared amplitudes of quantum chromodynamics and quantum electrodynamics processes, respectively, at a speed that is up to orders of magnitude faster than current symbolic computation frameworks. We discuss the performance of the current model, its limitations and possible future directions for this work.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Thermo-Fluid Modeling Framework for Supercomputer Digital Twins: Part 1, Demonstration at Exascale

A thermo-fluid modeling framework is being developed for ExaDigiT---an open-source framework for developing comprehensive digital twins of liquid-cooled supercomputers. The work is being conducted in two parts, and discussion is divided into two companion papers. The work documented in this paper focuses on the development of a cooling system library in Dymola for the Frontier supercomputer at Oak Ridge National Laboratory. The second part, outlined in a companion paper, focuses on a templating structure called Auto-CSM for easily creating model-agnostic, physics-based thermo-fluid cooling system models for liquid-cooled supercomputers using a text-based schema. The cooling model is being developed using primarily the open-source Transient Simulation Framework of Reconfigurable Models (TRANSFORM) library. The library follows the templating architecture developed within the TRANSFORM library for modeling subsystems. A full-system validation was performed to validate a very simple model that is integrated with the system controls, and the results are presented herein.

Kumar, Vineet↗
DOE OSTI2026

ML-Based Pebble Power Reconstruction for Pebble Bed Reactor Analysis

Pebble power reconstruction has been explored to complement the conventional homogenized modeling approach in pebble bed reactor (PBR) analysis, as detailed heterogeneous geometry calculations are computationally expensive. The random distribution of pebble fuels within the core challenges the application of conventional pin power reconstruction methods. To address this, we introduce a machine learning approach based on the transformer model, composed of encoder and decoder layers, to estimate the flux and power form functions for reconstructing individual pebble neutron fluxes and powers. The homogeneous neutron flux distribution within each spectral zone (SZ) is obtained from finite element solutions of global diffusion or transport calculations. Verification tests demonstrate that the trained transformer model accurately predicts power form functions over a range of conditions, including variations in pebble enrichment, location, type, SZ size, and burnup. In particular, verification using a three-dimensional PBR benchmark with burned pebbles shows good agreement in heterogeneous pebble power distributions between Griffin and Serpent. These results highlight the potential of applying conventional pin power reconstruction approaches to PBR cores with randomly distributed pebbles.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

TX$^2$: Transformer eXplainability and eXploration

The Transformer eXplainability and eXploration (Martindale & Stewart, 2021), or TX 2 software package, is a library designed for artificial intelligence researchers to better understand the performance of transformer models (Vaswani et al., 2017) used for sequence classification. The tool is capable of integrating with a trained transformer model and a dataset split into training and testing populations to produce an ipywidget (Project Jupyter Contributors, 2021) dashboard with a number of visualizations to understand model performance with an emphasis on explainability and interpretability. The TX 2 package is primarily intended to integrate into a workflow centered around Jupyter Notebooks (Kluyver et al., 2016), and currently assumes the use of PyTorch (Paszke et al., 2019) and Hugging Face transformers library (Wolf et al., 2020). The dashboard includes visualization and data exploration features to aid researchers, including an interactive UMAP embedding graph (McInnes et al., 2018) to understand classification clusters, a word salience map that can be updated as researchers alter textual entries in near real time, a set of tools to understand word frequency and importance based on the clusters in the UMAP embedding graph, and a set of traditional confusion matrix analysis tools.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Resolving extreme jet substructure

We study the effectiveness of theoretically-motivated high-level jet observables in the extreme context of jets with a large number of hard sub-jets (up to N = 8). Previous studies indicate that high-level observables are powerful, interpretable tools to probe jet substructure for N ≤ 3 hard sub-jets, but that deep neural networks trained on low-level jet constituents match or slightly exceed their performance. We extend this work for up to N = 8 hard sub-jets, using deep particle-flow networks (PFNs) and Transformer based networks to estimate a loose upper bound on the classification performance. A fully-connected neural network operating on a standard set of high-level jet observables, 135 N-subjetiness observables and jet mass, reach classification accuracy of 86.90%, but fall short of the PFN and Transformer models, which reach classification accuracies of 89.19% and 91.27% respectively, suggesting that the constituent networks utilize information not captured by the set of high-level observables. We then identify additional high-level observables which are able to narrow this gap, and utilize LASSO regularization for feature selection to identify and rank the most relevant observables and provide further insights into the learning strategies used by the constituent-based neural networks. The final model contains only 31 high-level observables and is able to match the performance of the PFN and approximate the performance of the Transformer model to within 2%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Self-supervised learning of spatiotemporal thermal signatures in additive manufacturing using reduced order physics models and transformers

Microstructure control via additive manufacturing has enormous potential as manufacturers, materials scientists, and designers alike seek to exploit novel fabrication technologies to improve component performance. Recent works have demonstrated the feasibility of producing materials with controlled microstructures across various length scales. However, the experimental approach towards exploring the process-structure space can be laborious and costly. This is particularly true if also considering scan pattern optimization which is well suited for processes such as powder bed fusion electron beam melting. In this work we propose an approach for encoding additive manufacturing layer-wise thermal response signatures using self-supervised representation learning. Thermal simulations from a reduced order model are utilized to estimate the spatiotemporal response during printing. A machine learning framework, using video-transformers, is utilized to efficiently distill spatiotemporal patterns into a compact latent space representation. This latent state representation encodes the relevant physics which is then utilized to establish a data-driven process-structure model for an additively manufactured Ni-based superalloy. In conclusion, the proposed methodology could potentially be used towards in-situ process monitoring, scan pattern experimental design, and component qualification.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Data-driven modeling of power generation for a coal power plant under cycling

Increased penetration of renewables for power generation has negatively impacted the dynamics of conventional fossil fuel-based power plants. The power plants operating on the base load are forced to cycle, to adjust to the fluctuating power demands. This results in an inefficient operation of the coal power plants, which leads up to higher operating losses. To overcome such operational challenge associated with cycling and to develop an optimal process control, this work analyzes a set of models for predicting power generation. Moreover, the power generation is intrinsically affected by the state of the power plant components, and therefore our model development also incorporates additional power plant process variables while forecasting the power generation. We present and compare multiple state-of-the-art forecasting data-driven methods for power generation to determine the most adequate and accurate model. We also develop an interpretable attention-based transformer model to explain the importance of process variables during training and forecasting. The trained deep neural network (DNN) LSTM model has good accuracy in predicting gross power generation under various prediction horizons with/without cycling events and outperforms the other models for long-term forecasting. The DNN memory-based models show significant superiority over other state-of-the-art machine learning models for short, medium and long range predictions. The transformer-based model with attention enhances the selection of historical data for multi-horizon forecasting, and also allows to interpret the significance of internal power plant components on the power generation. This newly gained insights can be used by operation engineers to anticipate and monitor the health of power plant equipment during high cycling periods.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2023

Haplotype‐resolved genome assembly of Populus tremula × P. alba reveals aspen‐specific megabase satellite DNA

SUMMARY Populus species play a foundational role in diverse ecosystems and are important renewable feedstocks for bioenergy and bioproducts. Hybrid aspen Populus tremula × P. alba INRA 717‐1B4 is a widely used transformation model in tree functional genomics and biotechnology research. As an outcrossing interspecific hybrid, its genome is riddled with sequence polymorphisms which present a challenge for sequence‐sensitive analyses. Here we report a telomere‐to‐telomere genome for this hybrid aspen with two chromosome‐scale, haplotype‐resolved assemblies. We performed a comprehensive analysis of the repetitive landscape and identified both tandem repeat array‐based and array‐less centromeres. Unexpectedly, the most abundant satellite repeats in both haplotypes lie outside of the centromeres, consist of a 147 bp monomer PtaM147, frequently span >1 megabases, and form heterochromatic knobs. PtaM147 repeats are detected exclusively in aspens (section Populus ) but PtaM147‐like sequences occur in LTR‐retrotransposons of closely related species, suggesting their origin from the retrotransposons. The genomic resource generated for this transformation model genotype has greatly improved the design and analysis of genome editing experiments that are highly sensitive to sequence polymorphisms. The work should motivate future hypothesis‐driven research to probe into the function of the abundant and aspen‐specific PtaM147 satellite DNA.

Zhou, Ran↗
NASA NTRSDate not supplied

Harnessing Artificial Intelligence for Medical Diagnosis and Treatment During Space Exploration Missions

BACKGROUND The medical capabilities necessary for long-duration exploration missions (LDEMs) will differ tremendously from those currently available to crew medical officers (CMOs) on the International Space Station (ISS). Ground support will be more challenging due to distance-related communication delays and data transmission, and resource utilization must be optimized given limited ability for resupply. Clinical decision support systems (CDSSs) can help mitigate these limitations. The recent launch of generative artificial intelligence (AI) tools based upon large language models (LLM) support the creation of a smart assistant for onboard triage, diagnosis, and guided treatment of medical conditions during these missions. The Informing Mission Planning via Analysis of Complex Tradespaces (IMPACT) tool can help predict which clinical problems and outcomes are likely to occur for a design reference mission (DRM) and assist Medical Operations and systems engineering teams in creating a medical system that may optimally mitigate the predicted risks. The purpose of this study was to identify AI tools currently available or in development for the assistive diagnosis and care of medical conditions predicted for an extended duration Lunar mission. METHODS The 119 medical conditions currently built into the IMPACT suite were categorized into systems, and these diagnoses were used as keywords for our literature search. Using PubMed and Google Scholar, we performed a literature survey of AI tools applicable to these conditions. Article inclusion criteria included publication between the years 2017-2023, as the sentinel paper discussing the “selective attention” driving ChatGPT and other generative transformer models was published in June 2017. Where applicable, we reviewed only the top 1000 research articles (based on relevance) for each of the keywords/phrases. AI tools whose training sets were exclusive to a pediatric patient population were excluded. We also excluded any medical diagnostic tools (such as CT, MRI, mass spectrometry) or procedures (such as endoscopy, surgery) that are unlikely to be available during LDEMs due to mass and volume constraints, CMO knowledge, skills, and abilities, and/or inherent procedural risks. RESULTS Our survey highlighted several AI-driven tools for the triage, diagnosis, and management of those medical conditions highlighted by IMPACT. Selected publications for each medical condition were then screened for inclusion within ten systems-based categories including: general diagnostic tools (25), tools to diagnose and manage respiratory (40), dermatologic (34), neurologic (28), auditory and vestibular (30), ophthalmic (34), musculoskeletal (104), infection-associated (92), and gynecologic (19) conditions, as well as tools that could be deployed in the setting of trauma and emergency (34). CONCLUSIONS Numerous AI-driven tools were highlighted within this literature survey, ranging from chatbot assistants that triage knee pain to vision transformer models for diagnosis of ophthalmic conditions using ocular surface images captured with a mobile phone. Remaining challenges include optimizing connectivity and integration of existing and developing systems into the vehicles or habitats. Notably, findings from this survey could help guide the initial design of an all-encompassing, onboard medical AI assistant for use during future LDEMs.

R A Lacinski↗
DOE OSTI2025

Modeling phase transformations in Mn-rich disordered rocksalt cathodes with machine-learning interatomic potentials

Mn-rich disordered rocksalt (DRX) cathode materials exhibit a phase transformation from a disordered to a partially disordered spinel-like structure (δ-phase) during electrochemical cycling. Here, in this computational study, we use charge-informed molecular dynamics with a fine-tuned CHGNet foundation potential to investigate the phase transformation in LixMn 0.8 Ti 0.1 O 1.9 F 0.1 . Our results indicate that transition metal migration occurs and reorders to form the spinel-like ordering in an FCC anion framework. The transformed structure contains a higher concentration of nontransition metal (0-TM) face-sharing channels, which are known to improve Li transport kinetics. Analysis of the Mn valence distribution suggests that the appearance of tetrahedral Mn 2+ is a consequence of spinel-like ordering, rather than the trigger for cation migration as previously suggested. Calculated equilibrium intercalation voltage profiles demonstrate that the δ-phase, unlike the ordered spinel, exhibits solid-solution signatures at low voltage. A higher Li capacity is obtained than in the DRX phase. This study provides atomic insights into solid-state phase transformation and its relation to experimental electrochemistry, highlighting the potential of machine-learning interatomic potentials for understanding complex oxide materials.

Zhong, Peichen [University of California, Berkeley↗
DOE OSTI2024

Inverse-Velocity Transformation Wall Model for Reacting Turbulent Hypersonic Boundary Layers

The present study builds on prior work by taking advantage of the novel framework proposed by Griffin et al. (hereafter referred to as the GFM) as a baseline. The model is progressively extended to multicomponent reacting mixtures, accounting for differential diffusion and finite-rate chemistry in a similar fashion to Di Renzo & Urzay and Di Renzo et al. The accuracy of the present approach, as well the prior model of Di Renzo & Urzay, is assessed in an a priori sense for the first time in a turbulent reacting boundary layer, using the boundary-layer data of Williams et al. Five species are included in the present analysis, i.e., N s = 5, namely N 2 , O 2 , NO, N and O, a neutral mixture most representative of dissociation/recombination phenomena for temperatures below 6000 K. A schematic of a flow over a wedge representative of the described configuration is presented in Figure 1. The brief is organized as follows: In Section 2, the wall-model equations and the computational framework are presented. In Section 3, the a priori results of the proposed model are described and compared to the extended EWM. Finally, in Section 4, some conclusions are offered.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Modelica-json: Transforming energy models to digitize the control delivery process

Building simulation models are typically not used to generate the documentation required for bidding and project delivery of commercial building systems, or for their semantic modeling and commissioning. This paper presents a software tool that aids in digitizing the control delivery process, spanning simulation during design to implementation and formal verification during commissioning. The tool can generate from Modelica models digital documentation of control sequences. This digital documentation, along with other project drawings and specifications can be used for project bidding. It can also be used for implementation of control sequences through machineto-machine translation to commercial legacy control products, for which we are currently developing the proposed ASHRAE Standard 231P based on the presented work. Moreover, as-installed control sequences can be formally verified against the design specification, and a semantic model in Brick can be exported to aid in configuration of building analytics and fault detection. The paper presents what we believe is the first translation of a Modelica-implemented control sequence to a native implementation on a commercial control platform, using the webCTRL product line from Automated Logic. The paper also shows how a webCTRL implementation can be formally verified against its Modelica specification. These use cases have all been demonstrated with a prototype implementation that is now being further developed.

Wetter, Michael↗