Machine learning results and code
Results for all the Four MLP-NN, RF, XGBR, and RF algorithms for each of geophysical array`s and machine learning python code is provided.
58 GEOSCIENCES↗
SEARCH · Engineering Papers
Results for “Random forests”
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.
MLP-NN vs Gauss-Newton files
-Input files for randomly selected subset data (30 instances) from primary dipole-dipole forward modeling data.-Inversion results in surfer grid format as well as in .DAT format.-Scatterplots for MLP-NN vs Gauss-Newton present in the excel sheet.
58 GEOSCIENCES↗
Results of Uncertainty Quantification of data , and Pi3NN code
Uncertainty Quantification results along with Pi3NN code and forward model data for dipole-dipole on which it was performed is attached.
58 GEOSCIENCES↗
Implementation of MLP-NN on Field data for dipole dipole and Alt3 Wenner configurations
Training data for both dipole dipole and Alt3 wenner configurations included. Field dataset for both dipole dipole and Alt3 wenner configurations included (Test dataset). Python code for implementation of MLP-NN included. Results for inversion of both configurations included.
58 GEOSCIENCES↗
Supplemental Tables for "Machine learning enables identification of a novel yeast galactose utilization pathway"
Supplemental tables for the paper "Machine learning enables identification of a novel yeast galactose utilization pathway" by Marie-Claire Harrison et al. 2024.
artificial intelligence↗
Chronic obstructive pulmonary disease among former United States Department of Energy workers: comorbidities and lung function changes
Chronic Obstructive Pulmonary Disease (COPD) is a major cause of morbidity and mortality in the United States and is frequently associated with multiple comorbidities which lead to poor COPD outcomes in the general population. However, little is known regarding COPD comorbidities in occupational cohorts whose exposure experiences could result in differences in comorbidities compared to the general population. These differences may also be important for assessing COPD outcomes such as lung function changes or decline. Therefore, the objectives of this study were to: (1) identify and describe clusters of COPD comorbidities among Department of Energy (DOE) former workers; (2) assess if the attributes of the identified clusters differ from those identified among the general population based on the published literature, and (3) identify predictors of lung function changes and decline among DOE former workers.
60 APPLIED LIFE SCIENCES↗
Data supporting manuscript from L. Sheneman, G. Stephanopoulos, A.E. Vasdekis titled "Deep learning classification of lipid droplets in quantitative phase images" as currently under review at PLOS ONE. This includes: 1) raw and binary labeled Quantitative Phase Images (QPI) of Y. lipolytica cells used in the analyses described within the manuscript. 2) various derived data including classifier scores, etc.
Data supporting manuscript from L. Sheneman, G. Stephanopoulos, A.E. Vasdekis titled "Deep learning classification of lipid droplets in quantitative phase images" as currently under review at PLOS ONE. This includes: 1) raw and binary labeled Quantitative Phase Images (QPI) of Y. lipolytica cells used in the analyses described within the manuscript. 2) various derived data including classifier scores, etc.
ANN↗
A Photometric Machine-Learning Method to Infer Stellar Metallicity
Following its formation, a star's metal content is one of the few factors that can significantly alter its evolution. Measurements of stellar metallicity ([Fe/H]) typically require a spectrum, but spectroscopic surveys are limited to a few x 10(exp 6) targets; photometric surveys, on the other hand, have detected > 10(exp 9) stars. I present a new machine-learning method to predict [Fe/H] from photometric colors measured by the Sloan Digital Sky Survey (SDSS). The training set consists of approx. 120,000 stars with SDSS photometry and reliable [Fe/H] measurements from the SEGUE Stellar Parameters Pipeline (SSPP). For bright stars (g' < or = 18 mag), with 4500 K < or = Teff < or = 7000 K, corresponding to those with the most reliable SSPP estimates, I find that the model predicts [Fe/H] values with a root-mean-squared-error (RMSE) of approx.0.27 dex. The RMSE from this machine-learning method is similar to the scatter in [Fe/H] measurements from low-resolution spectra..
machine learning↗
Multisensor Approach to Land use and Land Cover Mapping in Brazilian Amazon
Explore the source record for details and available documents.
Classification, Sentinel images, Random Forest, Mu↗
Porosity Area Fraction Analysis of Bonded Borosilicate Specimens From in-Situ Optical Imagery
Image segmentation methods are routinely used to analyze data derived in nondestructive testing scenarios. A typical problem in this area is to estimate porosity from X-ray Computed Tomography (CT) inspections. This investigation aims to understand the effectiveness of open-source image segmentation methods for porosity estimation from optical images in the context of strength analysis of bonded materials and provides a benchmark of performance against a standard commercial software package. The rank ordering of the specimens by porosity area fraction is consistent across both methods.
Porosity area fraction↗
Porosity Area Fraction Analysis of Bonded Borosilicate Specimens From in-Situ Optical Imagery
Image segmentation methods are routinely used to analyze data derived in nondestructive testing scenarios. A typical problem in this area is to estimate porosity from X-ray Computed Tomography (CT) inspections. This investigation aims to understand the effectiveness of open-source image segmentation methods for porosity estimation from optical images in the context of strength analysis of bonded materials and provides a benchmark of performance against a standard commercial software package. The rank ordering of the specimens by porosity area fraction is consistent across both methods.
Porosity area fraction↗
Marin County Wildfires Ii: Improving Fire Suppression Modeling to Inform Fire Prevention and Suppression Decisions in Marin County, Ca
A future of increased wildfires requires greater integration of spatial analysis and local knowledge of emergency responders. We examine the application of a Potential Operational Delineations (PODs) framework for strategic pre-fire planning in Marin County. PODs are spatial units for wildfire management that combine predictive modeling and local firefighter expertise to identify potential control locations as unit boundaries and assess the difficulty of suppression within units. Additionally, this project explores the integration of road networks and social vulnerability to assess environmental justice in evacuation safety. This project constitutes a novel application of the PODs framework as it integrates expertise from Marin County senior firefighters with a Fireline Location Model (FLM) to achieve POD definition and uses a Suppression Difficulty Score (SDS) to rank each POD. The FLM uses network analysis and hydrologic modeling to identify key roads and ridgelines as boundaries and combines them with expert knowledge, in the form of workshops, to construct PODs. Once identified, PODs are classified using SDS, which includes processed inputs such as LiDAR-derived aboveground biomass, ECOSTRESS Evaporative Stress Index, land use cover type from Sentinel-2 Imagery, and a digital elevation model. Environmental justice for evacuation safety incorporated three key road metrics such as connectivity, travel area, and exit capacity, the Social Vulnerability Index from the Center for Disease Control, and cell coverage to determine a final Evacuation Difficulty Score. Results indicate a strong link between road networks as primary POD boundaries, with ridgelines and waterways as secondary and tertiary locations. Specifically, we find 78.5% of expertise-identified POD boundaries align with FLM-determined boundaries. More validation is needed to support this process; however, initial results signal a feasible framework to integrate expertise and spatial analysis in local level strategic fire planning
Wildfire modeling↗
Córdoba Wildland Fires: Assessing Fire Risk Factors in Córdoba, Argentina using Earth Observations
In recent years, Córdoba, Argentina has experienced intensified wildfire activity, with fires in 2020 alone scorching over 300,000 hectares within the province. Potential causes for the increased burn area include climate change, the expanding wildland-urban interface (WUI), and inadequate fire management practices. Previous studies have produced fire frequency maps for the region, but gaps remain in understanding the environmental parameters influencing fire behavior and growth. We partnered with the Instituto Nacional de Tecnología Agropecuaria (INTA) to address these gaps by utilizing NASA Earth observing data to analyze key wildfire risk factors. Using a combination of data inputs from Soil Moisture Active Passive (SMAP), Shuttle Radar Topography Mission (SRTM), Global Precipitation Measurement (GPM) Integrated Multi-satellite Retrievals for GPM (IMERG), and Aqua/Terra Moderate Resolution Imaging Spectroradiometer (MODIS), we created a ten-year baseline using environmental variables to determine anomalies that influenced the fires of 2020. Baseline data were used to calculate the statistical significance of the environmental factors as precursors to wildfires. We found that the normalized difference vegetation index (NDVI) and precipitation were the strongest indicators for the September 2020 wildfires. Using the environmental risk factors, we created a wildfire risk map for the province of Córdoba, which can be used to enhance our partner’s fire management strategies and decision-making processes.
wildland fires↗
Reducing Uncertainty in Offshore Wind Energy Yield Estimates via a Metocean Reference Site
The offshore wind industry is burgeoning in the coastal waters of the United States, specifically along the Atlantic. For wind energy to be successful, reliable observations and model simulations are needed for resource assessment and forecasting. While many of these activities have already begun, there is currently an absence of observations at hub-height in these waters, with the closest available hub-height measurements usually taken onshore. Deployment of floating lidars has occurred through various federally funded projects, but only encapsulates time periods of a couple of years at best. Private industry is also beginning to leverage floating lidars, but this data is often proprietary, and not shared with the general public. In this work, we make the case for a metocean reference site for long-term offshore wind energy. Specifically, we quantify the impact of having a metocean reference site compared to other methods of determining hub-height winds and energy production. We use an offshore floating lidar to directly measure the wind resource, and compare these measurements to predictions derived from other widely-available surface meteorological variables. These prediction methods (vertical extrapolation, machine learning, and NWP output) produce a variety of vertical wind speed profiles, of which produce different energy yield estimates for a reference offshore turbine (Figure 1). While some methods perform reasonably well against the lidar, the uncertainty in these energy yield estimates has financial implications, further illustrating the need for long-term measurements in coastal waters.
machine learning↗
AI-Enabled Operations at Fermi Complex: Multivariate Time Series Prediction for Outage Prediction and Diagnosis
The Main Control Room of the Fermilab accelerator complex continuously gathers extensive time-series data from thousands of sensors monitoring the beam. However, unplanned events such as trips or voltage fluctuations often result in beam outages, causing operational downtime. This downtime not only consumes operator effort in diagnosing and addressing the issue but also leads to unnecessary energy consumption by idle machines awaiting beam restoration. The current threshold-based alarm system is reactive and faces challenges including frequent false alarms and inconsistent outage-cause labeling. To address these limitations, we propose an AI-enabled framework that leverages predictive analytics and automated labeling. Using data from $2,703$ Linac devices and $80$ operator-labeled outages, we evaluate state-of-the-art deep learning architectures, including recurrent, attention-based, and linear models, for beam outage prediction. Additionally, we assess a Random Forest-based labeling system for providing consistent, confidence-scored outage annotations. Our findings highlight the strengths and weaknesses of these architectures for beam outage prediction and identify critical gaps that must be addressed to fully harness AI for transitioning downtime handling from reactive to predictive, ultimately reducing downtime and improving decision-making in accelerator management.
Jain, Milan [PNL, Richland] (ORCID:000000021676111↗
Validating Protection System Behavior with Machine Learning in a Master State Overseer
As power system protection devices continue the widespread transition from analog to digital, they become increasingly intricate. The internal functions and communication between critical grid components must now be significantly more complex to keep up with the demands of the modern smart grid. This brings increased difficulty in maintenance and monitoring, making it harder to identify potential misoperation, power anomalies, and cyber threats. Such issues are often only pinpointed after an exhaustive and costly post-mortem analysis, when a major outage or damage has already occurred. A solution is needed for validating protection systems as they operate, independently evaluating grid state and confirming whether the protection system is behaving accordingly. As opposed to incident response, this acts as a constant verification mechanism that raises a flag when subtler issues are noticed, catching them earlier and preventing larger incidents. This work presents the implementation of such a system, expanding on the prototype developed by the authors in a previous paper. This is accomplished with a machine learning (ML) system capable of validating the performance of protection systems by classifying anomalous events and characterizing protection system responses based solely on available current and voltage measurements. Additionally, this system is contextualized within a larger, modular Master State awareness Overseer (MSO) framework, responsible for monitoring, analyzing, and managing an electric grid.
24 - POWER TRANSMISSION AND DISTRIBUTION↗
The two-point correlation function of randomly distributed Lyman alpha clouds
It is often assumed that Ly-alpha forest clouds are randomly distributed, intergalactic objects that are highly ionized by the UV background produced by quasars. If these assumptions are true, fluctuations in the UV background should produce a nonzero two point correlation function in the Ly-alpha forest. This effect, which is really just a generalization of the proximity effect, is more significant at high redshift (z is approximately 3-4) because the mean free path for UV photons is smaller there, and the fluctuations correspondingly larger. This effect was studied using both the semi-analytic techniques of Zuo's recent papers and Monte Carlo simulations. The correlation function is expected to have a small yet potentially measurable amplitude that is consistent with current upper limits. Furthermore, the signature of this effect is distinctive because the nonzero correlation function extends over the photon mean free path, which is larger than the expected scale of large-scale structure. Observations or upper limits on this effect could provide information about the source of the ionizing background at high redshifts and the nature of the Ly-alpha forest clouds.
Fardal, Mark A.↗
Efficient Sampling of Complex Interdependent and Multiplex Networks
Efficient sampling of interdependent and multiplex infrastructure networks is critical for effectively applying failure and recovery algorithms in real-world settings, as well as to generate property-preserving reduced-order graph-based ensembles that address topological uncertainties. In this paper, we first explore the performance, i.e. the success in preserving graph properties, of graph sampling algorithms for interdependent and multiplex networks with synthetic and real-world graphs. We simulate sampling algorithms under different parameter settings. These settings include probabilistic graph generators, coupling patterns, and various performance metrics. Our results show that while Random Node and Random Walk sampling algorithms perform best for interdependent networks, Random Edge and Forest Fire sampling algorithms perform best for multiplex networks. Second, we propose and implement a novel similarity-based sampling algorithm for multiplex networks that samples only log(N) number of layers of an N-layer multiplex network while yielding computational savings with performance guarantees. Experimental results show that similarity sampling outperforms complete sampling of all layers while decreasing performance costs from a linear scale to a logarithmic one. Our results also indicate that similarity-based sampling outperforms complete sampling and random selection in nearly all scenarios when tested with real-world data.
Subasi, Omer↗