Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “model training strategies”

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

Evaluation of pre-training large language models on leadership-class supercomputers

Large language models (LLMs) have arisen rapidly to the center stage of artificial intelligence as the foundation models applicable to many downstream learning tasks. However, how to effectively build, train, and serve such models for many high-stake and first-principle-based scientific use cases are both of great interests and of great challenges. Moreover, pre-training LLMs with billions or even trillions of parameters can be prohibitively expensive not just for academic institutions, but also for well-funded industrial and government labs. Furthermore, the energy cost and the environmental impact of developing LLMs must be kept in mind. Here, in this work, we conduct a first-of-its-kind performance analysis to understand the time and energy cost of pre-training LLMs on the Department of Energy (DOE)’s leadership-class supercomputers. Employing state-of-the-art distributed training techniques, we evaluate the computational performance of various parallelization approaches at scale for a range of model sizes, and establish a projection model for the cost of full training. Our findings provide baseline results, best practices, and heuristics for pre-training such large models that should be valuable to HPC community at large. We also offer insights and optimization strategies for using the first exascale computing system, Frontier, to train models of the size of GPT-3 and beyond.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

A machine learning estimator trained on synthetic data for real-time earthquake ground-shaking predictions in Southern California

Abstract After large-magnitude earthquakes, a crucial task for impact assessment is to rapidly and accurately estimate the ground shaking in the affected region. To satisfy real-time constraints, intensity measures are traditionally evaluated with empirical Ground Motion Models that can drastically limit the accuracy of the estimated values. As an alternative, here we present Machine Learning strategies trained on physics-based simulations that require similar evaluation times. We trained and validated the proposed Machine Learning-based Estimator for ground shaking maps with one of the largest existing datasets (<100M simulated seismograms) from CyberShake developed by the Southern California Earthquake Center covering the Los Angeles basin. For a well-tailored synthetic database, our predictions outperform empirical Ground Motion Models provided that the events considered are compatible with the training data. Using the proposed strategy we show significant error reductions not only for synthetic, but also for five real historical earthquakes, relative to empirical Ground Motion Models.

Environmental Sciences & Ecology↗
NASA NTRS2015

Evaluating the Utility of Satellite Soil Moisture Retrievals over Irrigated Areas and the Ability of Land Data Assimilation Methods to Correct for Unmodeled Processes

Earth's land surface is characterized by tremendous natural heterogeneity and human-engineered modifications, both of which are challenging to represent in land surface models. Satellite remote sensing is often the most practical and effective method to observe the land surface over large geographical areas. Agricultural irrigation is an important human-induced modification to natural land surface processes, as it is pervasive across the world and because of its significant influence on the regional and global water budgets. In this article, irrigation is used as an example of a human-engineered, often unmodeled land surface process, and the utility of satellite soil moisture retrievals over irrigated areas in the continental US is examined. Such retrievals are based on passive or active microwave observations from the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), the Advanced Microwave Scanning Radiometer 2 (AMSR2), the Soil Moisture Ocean Salinity (SMOS) mission, WindSat and the Advanced Scatterometer (ASCAT). The analysis suggests that the skill of these retrievals for representing irrigation effects is mixed, with ASCAT-based products somewhat more skillful than SMOS and AMSR2 products. The article then examines the suitability of typical bias correction strategies in current land data assimilation systems when unmodeled processes dominate the bias between the model and the observations. Using a suite of synthetic experiments that includes bias correction strategies such as quantile mapping and trained forward modeling, it is demonstrated that the bias correction practices lead to the exclusion of the signals from unmodeled processes, if these processes are the major source of the biases. It is further shown that new methods are needed to preserve the observational information about unmodeled processes during data assimilation.

land data↗
DOE OSTI2022

A graph neural network (GNN) approach to basin-scale river network learning: the role of physics-based connectivity and data fusion

Abstract. Rivers and river habitats around the world are under sustained pressure from human activities and the changing global environment. Our ability to quantify and manage the river states in a timely manner is critical for protecting the public safety and natural resources. In recent years, vector-based river network models have enabled modeling of large river basins at increasingly fine resolutions, but are computationally demanding. This work presents a multistage, physics-guided, graph neural network (GNN) approach for basin-scale river network learning and streamflow forecasting. During training, we train a GNN model to approximate outputs of a high-resolution vector-based river network model; we then fine-tune the pretrained GNN model with streamflow observations. We further apply a graph-based, data-fusion step to correct prediction biases. The GNN-based framework is first demonstrated over a snow-dominated watershed in the western United States. A series of experiments are performed to test different training and imputation strategies. Results show that the trained GNN model can effectively serve as a surrogate of the process-based model with high accuracy, with median Kling–Gupta efficiency (KGE) greater than 0.97. Application of the graph-based data fusion further reduces mismatch between the GNN model and observations, with as much as 50 % KGE improvement over some cross-validation gages. To improve scalability, a graph-coarsening procedure is introduced and is demonstrated over a much larger basin. Results show that graph coarsening achieves comparable prediction skills at only a fraction of training cost, thus providing important insights into the degree of physical realism needed for developing large-scale GNN-based river network models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

The Rise of Neural Networks for Materials and Chemical Dynamics

Machine learning (ML) is quickly becoming a premier tool for modeling chemical processes and materials. ML-based force fields, trained on large data sets of high-quality electron structure calculations, are particularly attractive due their unique combination of computational efficiency and physical accuracy. This Perspective summarizes some recent advances in the development of neural network-based interatomic potentials. Designing high-quality training data sets is crucial to overall model accuracy. One strategy is active learning, in which new data are automatically collected for atomic configurations that produce large ML uncertainties. Another strategy is to use the highest levels of quantum theory possible. Transfer learning allows training to a data set of mixed fidelity. A model initially trained to a large data set of density functional theory calculations can be significantly improved by retraining to a relatively small data set of expensive coupled cluster theory calculations. These advances are exemplified by applications to molecules and materials.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

HPC Analytics of Fused Thermal Plants Data to Optimize Operating Envelope

In this project, ORNL extensively reviewed the ORAP RAM data, and it guided us to develop machine learning models that can predict time to next failures and forecast failure trends, which will be useful for optimizing power plant operation strategies. More specifically, we trained multiple random forest models and evaluated the model accuracy to validate with 10+ years of historical data. In addition, we implemented a web-based graphical user interface system for the models to show how our models can be used in more intuitive ways. This proof of concept allowed exploration of model use with power plant operators in mind. Developed machine learning models will be helpful for managing risks, planning maintenance and operation, ultimately reducing the down time and increasing the service hours. For future work, there are several interesting research topics including but not limited to model enhancement, creating synergy with traditional failure modeling approaches, and data-driven actionable recommendation and suggestions.

20 FOSSIL-FUELED POWER PLANTS↗
DOE OSTI2025

SLAB: simultaneous labeling and binding affinity prediction for protein–ligand structures

Machine learning models are often used as scoring functions to predict the binding affinity of a protein–ligand complex. These models are trained with limited amounts of data with experimentally measured binding affinity values. A large number of compounds are labeled inactive through single-concentration screens without measuring binding affinities. These inactive compounds, along with the active ones, can be used to train binary classification models, while regression models are trained using compounds with binding affinities only. However, the classification and regression tasks are often handled separately, without sharing the learned feature representations. In this paper, we propose a novel model architecture that jointly performs regression and classification objectives, aiming to maximize data utilization and improve predictive performance by leveraging two complementary tasks. In our setup, the regression yields the binding affinity, whereas the classification task yields the label as active or inactive. We demonstrate our method using PDBbind, the standard 3D structure database, as well as a dataset of flavivirus protease compounds with binding affinity data. Our experiments show that the new joint training strategy improves the accuracy of the model, increasing applicability in various practical drug screening scenarios.

Biological and medical sciences↗
DOE OSTI2022

A Qualitative Strategy for Fusion of Physics into Empirical Models for Process Anomaly Detection

To facilitate the automated online monitoring of power plants, a systematic and qualitative strategy for anomaly detection is presented. This strategy is essential to provide credible reasoning on why and when an empirical versus hybrid (i.e., physics-supported) approach should be used and to determine the ideal mix of these two approaches for a defined anomaly detection scope. Empirical methods are usually based on pattern, statistical, and causal inference. Hybrid methods include the use of physics models to train and test data methods, reduce data dimensionality, reduce data-model complexity, augment data, and reduce empirical uncertainty; hybrid methods also include the use of data to tune physics models. The presented strategy is driven by key decision points related to data relevance, simple modeling feasibility, data inference, physics-modeling value, data dimensionality, physics knowledge, method of validation, performance, data availability, and suitability for training and testing, cause-effect, entropy inference, and model fitting. The strategy is demonstrated through a pilot use case for the application of anomaly detection to capture a valve packing leak at the high-pressure coolant injection system of a nuclear power plant.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2020

Toward Large-Scale Image Segmentation on Summit

Semantic segmentation of images is an important computer vision task that emerges in a variety of application domains such as medical imaging, robotic vision and autonomous vehicles to name a few. While these domain-specific image analysis tasks involve relatively small image sizes (~ 10 2 × 10 2 ), there are many applications that need to train machine learning models on image data with extents that are orders of magnitude larger (~10 4 × 10 4 ). Training deep neural network (DNN) models on large extent images is extremely memory-intensive and often exceeds the memory limitations of a single graphical processing unit, a hardware accelerator of choice for computer vision workloads. Here, an efficient, sample parallel approach to train U-Net models on large extent image data sets is presented. Its advantages and limitations are analyzed and near-linear strong-scaling speedup demonstrated on 256 nodes (1536 GPUs) of the Summit supercomputer. Using a single node of the Summit supercomputer, an early evaluation of a recently released model parallel framework called GPipe is demonstrated to deliver ~ 2X speedup in executing a U-Net model with an order of magnitude larger number of trainable parameters than reported before. Performance bottlenecks for pipelined training of U-Net models are identified and mitigation strategies to improve the speedups are discussed. Together, these results open up the possibility of combining both approaches into a unified scalable pipelined and data parallel algorithm to efficiently train U-Net models with very large receptive fields on data sets of ultra-large extent images.

Seal, Sudip↗
DOE OSTI2024

Leveraging Calibration Transfer Techniques for Remote Monitoring of Samarium and Europium in LiCl Using Laser-Induced Florescence Spectroscopy for Radioisotope Production Applications

Radioisotope production relies on complex chemical processes that must be performed in radiological hot cells or glove boxes because of the radioactive and otherwise hazardous materials being used. In these situations, optical sensors can provide real-time monitoring to users, which is unobtainable by more traditional methods. This study explores the use of calibration transfer methods to train a model on one instrument and date and then transfer it to another instrument of the same or different configuration on a different date. By performing laser-induced fluorescence measurements of Eu(III) and Sm(III) in 10 M LiCl over the course of 6 months using two disparate spectrometers and two different training sets, a strategy for calibrating and deploying models for online monitoring was established. Three transfer techniques were compared: direct standardization (DS), piecewise direct standardization (PDS), and external parameter orthogonalization (EPO). DS and PDS outperformed EPO for day-to-day transfers, and EPO was not effective for instrument-to-instrument transfers. Transferring the initial date’s full factorial model provided better prediction performance compared with retraining models the day of measurements using a D-optimal designed calibration set. For both day-to-day and instrument-to-instrument transfers, five Kennard–Stone selected samples were sufficient. Based on this choice, the initial-date, high-resolution spectrometer model was transferred to a lower-resolution, compact spectrometer 6 months later to monitor a simulated, real-time demonstration. Here, the combined predictions of the DS and PDS transferred models were able to accurately track the anticipated concentration profiles, maintaining root-mean-square error of prediction values below 10 ppm.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Spatiotemporal forecasting of the edge localized modes in tokamak plasmas using neural networks

Artificial intelligence techniques have been increasingly adopted by the plasma and fusion science to address problems like plasma reconstruction, surrogate modeling, and tokamak/stellarator optimization. A key focus in sustained fusion research is the prediction and mitigation of edge-localized-modes (ELMs), instabilities that occur in short, periodic bursts and can cause erosion to the tokamak vessel wall. Recent research has demonstrated the power of neural networks in approximating continuous functions. In this work, we build spatiotemporal forecasting models that can predict the onset of ELMs and their evolution at early stages. We leverage recent advances in generative modeling, sequence-to-sequence modeling, and Fourier neural operators to propose architectures and training strategies that can learn to forecast short to long term dynamics of the noisy signals due to ELMs. We benchmark the developed model against a state-of-the-art foundation model using the beam emission spectroscopy (BES) data that captures the plasma fluctuations due to ELMs over a 8 x 8 spatial grid. Our models demonstrate high accuracy, outperforming the baselines, in predicting the evolution of BES signals during ELM events. Furthermore, the developed models exhibit high accuracy in predicting the rapid rise and relaxation of the signals due to ELMs within 30–80 µs.

edge localized modes↗
DOE OSTI2020

Compressing deep neural networks on FPGAs to binary and ternary precision with hls4ml

We present the implementation of binary and ternary neural networks in the hls4ml library, designed to automatically convert deep neural network models to digital circuits with FPGA firmware. Starting from benchmark models trained with floating point precision, we investigate different strategies to reduce the network's resource consumption by reducing the numerical precision of the network parameters to binary or ternary. We discuss the trade-off between model accuracy and resource consumption. In addition, we show how to balance between latency and accuracy by retaining full precision on a selected subset of network components. As an example, we consider two multiclass classification tasks: handwritten digit recognition with the MNIST data set and jet identification with simulated proton-proton collisions at the CERN Large Hadron Collider. The binary and ternary implementation has similar performance to the higher precision implementation while using drastically fewer FPGA resources.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Unsupervised Clustering and Supervised Regression Learning to Select High Temperature Oxidation-Resistant Materials

High temperature oxidation and corrosion degradation mechanisms dictate the lifetime of materials critical to energy production. The combination of modeling and experimental approaches such as machine learning (ML) and data analytics, with sufficient experimental data, can accelerate the development of new materials while limiting its cost. In the present work, ML will be applied to two high temperature oxidation data libraries (Oak Ridge National Laboratory and National Air and Space Administration) that comprised of about 5000 mass change sample datasheets for a variety of materials and temperatures in dry air and air + 10 % H2O. A python code was developed to prepare the data for machine learning by collecting and formatting oxidation rate constants, alloy compositions and environment of exposure into a single data frame. Scikit-learn library and Statistics and Machine Learning Toolbox within MathWorks were then used to perform unsupervised clustering and supervised regression learning. The impact of dataset distribution on the performance of the developed ML models was evaluated. Potential strategies to improve the predictions and enhance extrapolative capability of the previously trained model were investigated.

Romedenne, Marie [ORNL] (ORCID:0000000317936561)↗
DOE OSTI2021

Lessons Learned in Employing Data Analytics to Predict Oxidation Kinetics and Spallation Behavior of High-Temperature NiCr-Based Alloys

Machine learning (ML) can offer many advantages in predicting material properties over traditional materials development methods based solely on limited experimental investigations or physical-based simulations with the capability to reduce development cost, risk, and time. However, so far, limited efforts have been made to predict alloy oxidation kinetics and spallation behavior via ML due to the lack of consistently measured and sufficient experimental data and the inherent complexity in oxidation behavior of multicomponent high-temperature alloys. A previous study reported the ability of ML to predict oxidation kinetics of NiCr-based alloys as a function of alloy composition and operating conditions. Here, the performance of a ML model in predicting rate constants and spallation probability was evaluated in light of the roles of the data distribution of the experimental dataset (data analytics), the alloy composition, the exposure environment and the chosen oxidation approach to extracting kinetic values from the measured mass changes (but using either a simple parabolic law or a statistical cyclic oxidation model). Potential strategies to improve the predictions and enhance the extrapolative capability of the previously trained model will be discussed.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Coupled social and infrastructure approaches for enhancing solar energy adoption. Final Report

The goal of this project was to work with rural electric cooperatives to facilitate the diffusion of solar energy adoption in households located in the rural and semi-urban areas of Virginia by identifying social and behavioral factors that might be unique to rural regions; and develop a model to calculate the solar adoption propensity score for household based on their demographics, social and behavioral characteristics which would provide an objective metric to cooperatives that can be further used to do targeted marketing of rooftop solar panels. This was achieved through the following tasks: (1) Conducted a survey of the members of Virginia electric cooperatives to identify demographic, social, financial and behavioral attributes of individuals who are likely to adopt rooftop solar panels. (2) Developed a highly detailed, data-driven, agent-based model of the population of Virginia, focusing on the rural regions. (3) Developed diffusion models that use social, behavioral, and demographic factors, and peer effects to study their impact on solar adoption in rural areas. (4) Built a prototype tool based on the diffusion model to help study market segmentation in rural areas and made it available to National Rural Electric Cooperative Association (NRECA). (5) Results and recommendations derived from the model were provided to NRECA to be shared with participating cooperatives. (6) Results were published in peer reviewed journals, conference proceedings and book chapters, and ideas disseminated through presentations and newsletters. There were several important methodological contributions made under this project which are detailed in the published papers, including: (1) Built a decision-adjusted model for predicting adoptors with imbalanced training data; (2) designed seeding strategies to maximize adoption given a fixed budget; (3) built a methodology to compare different agent based models; (4) created models to identify important factors that influence decision to adopt solar panels; and (5) built a methodology for building household profiles of solar generation to study the duck curve phenomenon. The team included members from the University of Virginia (lead), National Rural Electric Cooperative Association (NRECA), Arizona State University, Virginia Tech and Sandia National Laboratory. Note that no individual entity or stakeholder has incentive to promote solar in rural regions. Most of the research and work focuses around urban regions where the potential for growth in solar adoption is higher due to higher population density. This puts rural areas at a disadvantage. By improving the diffusion of solar adoption in rural parts of the country, we can not only provide clean energy to rural areas but also promote job growth and improves energy independence.

14 SOLAR ENERGY↗
DOE OSTI2020

Toward fusion plasma scenario planning for NSTX-U using machine-learning-accelerated models

One of the most promising devices for realizing power production through nuclear fusion is the tokamak. To maximize performance, it is preferable that tokamak reactors achieve advanced operating scenarios characterized by good plasma confinement, improved magnetohydrodynamic (MHD) stability, and a largely non-inductively driven plasma current. Such scenarios could enable steady-state reactor operation with high \emph{fusion gain} --- the ratio of produced fusion power to the external power provided through the plasma boundary. Precise and robust control of the evolution of the plasma boundary shape as well as the spatial distribution of the plasma current, density, temperature, and rotation will be essential to achieving and maintaining such scenarios. The complexity of the evolution of tokamak plasmas, arising due to nonlinearities and coupling between various parameters, motivates the use of model-based control algorithms that can account for the system dynamics. In this work, a learning-based accelerated model trained on data from the National Spherical Torus Experiment Upgrade (NSTX-U) is employed to develop planning and control strategies for regulating the density and temperature profile evolution around desired trajectories. The proposed model combines empirical scaling laws developed across multiple devices with neural networks trained on empirical data from NSTX-U and a database of first-principles-based computationally intensive simulations. The reduced execution time of the accelerated model will enable practical application of optimization algorithms and reinforcement learning approaches for scenario planning and control development. An initial demonstration of applying optimization approaches to the learning-based model is presented, including a strategy for mitigating the effect of leaving the finite validity range of the accelerated model. The approach shows promise for actuator planning between experiments and in real-time.

machine learning↗
DOE OSTI2022

Recurrent neural networks for short-term and long-term prediction of geothermal reservoirs

Accurate prediction of geothermal reservoir responses to alternative energy production scenarios is critical for optimizing the development of the underlying resources. While the conventional physics-based models offer a comprehensive prediction tool, data-driven models provide an efficient alternative to build fit-for-purpose predictive models by extracting and using the statistical patterns in the collected data to make predictions. The recurrent neural network (RNN) is a data-driven model that is commonly applied to predict time series sequences. This paper presents a variant of RNN that also utilizes the efficiency of convolutional neural networks (CNN) for the prediction of energy production from geothermal reservoirs. Specifically, a CNN–RNN architecture is developed that takes historical well controls as input (features) and their corresponding production response data as output (labels) to learn an input-output mapping that can predict the future well production responses/performance for any given future well control inputs. The model is paired with a labeling scheme to handle real field disturbances that create data gaps. In addition to the model structure, we introduce a thorough workflow for applying the model, which includes data pre-processing, feature selection, as well as different training strategies for short-term and long-term prediction. Finally, the performance and accuracy of the model are evaluated by applying it to multiple datasets, including a field reservoir model.

15 GEOTHERMAL ENERGY↗
DOE OSTI2020

Intelligent multi-zone residential HVAC control strategy based on deep reinforcement learning

Residential heating, ventilation, and air conditioning (HVAC) has been considered as an important demand response resource. However, the optimization of residential HVAC control is no trivial task due to the complexity of the thermal dynamic models of buildings and uncertainty associated with both occupant-driven heat loads and weather forecasts. In this paper, we apply a novel model-free deep reinforcement learning (RL) method, known as the deep deterministic policy gradient (DDPG), to generate an optimal control strategy for a multi-zone residential HVAC system with the goal of minimizing energy consumption cost while maintaining the users’ comfort. Here, the applied deep RL-based method learns through continuous interaction with a simulated building environment and without referring to any prior model knowledge. Simulation results show that compared with the state-of-art deep Q network (DQN), the DDPG-based HVAC control strategy can reduce the energy consumption cost by 15% and reduce the comfort violation by 79%; and when compared with a rule-based HVAC control strategy, the comfort violation can be reduced by 98%. In addition, experiments with different building models and retail price models demonstrate that the well-trained DDPG-based HVAC control strategy has high generalization and adaptability to unseen environments, which indicates its practicability for real-world implementation.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗