Search NASA⌕ Search

SEARCH · Search NASA

Results for “mining techniques”

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 343 records · Page 19

Adaptive Neuron Apoptosis for Accelerating Deep Learning on Large Scale Systems

Machine Learning and Data Mining (MLDM) algorithms are becoming ubiquitous in {\em model learning} from the large volume of data generated using simulations, experiments and handheld devices. Deep Learning algorithms -- a class of MLDM algorithms -- are applied for automatic feature extraction, and learning non-linear models for unsupervised and supervised algorithms. Naturally, several libraries which support large scale Deep Learning -- such as TensorFlow and Caffe -- have become popular. In this paper, we present novel techniques to accelerate the convergence of Deep Learning algorithms by conducting low overhead removal of redundant neurons -- {\em apoptosis} of neurons -- which do not contribute to model learning, during the training phase itself. We provide in-depth theoretical underpinnings of our heuristics (bounding accuracy loss and handling apoptosis of several neuron types), and present the methods to conduct adaptive neuron apoptosis. We implement our proposed heuristics with the recently introduced TensorFlow and using its recently proposed extension with MPI. Our performance evaluation on two difference clusters -- one connected with Intel Haswell multi-core systems, and other with nVIDIA GPUs -- using InfiniBand, indicates the efficacy of the proposed heuristics and implementations. Specifically, we are able to improve the training time for several datasets by 2-3x, while reducing the number of parameters by 30x (4-5x on average) on datasets such as ImageNet classification. For the Higgs Boson dataset, our implementation improves the accuracy (measured by Area Under Curve (AUC)) for classification from 0.88/1 to 0.94/1, while reducing the number of parameters by 3x in comparison to existing literature, while achieving a 2.44x speedup in comparison to the default (no apoptosis) algorithm.

Siegel, Charles M.↗

Micromechanical processes in consolidated granular salt

Here, granular salt is likely to be used as backfill material and a seal system component within geologic salt formations serving as a repository for long-term isolation of nuclear waste. Pressure from closure of the surrounding salt formation will promote consolidation of granular salt, eventually resulting in properties comparable to native salt. Understanding dependence of consolidation processes on stress state, moisture availability, temperature, and time is important for demonstrating sealing functions and long-term repository performance. This study characterizes laboratory-consolidated granular salt by means of microstructural observations. Granular salt material from mining operations was obtained from the bedded Salado Formation hosting the Waste Isolation Pilot Plant and the Avery Island salt dome. Laboratory test conditions included hydrostatic consolidation of jacketed granular salt with varying conditions of confining isochoric stress to 38 MPa, temperature to 250 °C, moisture additions of 1% by weight, time duration, and vented and non-vented states. Resultant porosities ranged between 1% and 22%. Optical and scanning electron microscopic techniques were used to ascertain consolidation mechanisms. From these investigations, samples with 1% added moisture or unvented during consolidation, exhibit clear pressure solution processes with tightly cohered grain boundaries and occluded fluid pores. Samples with only natural moisture content consolidated by a combination of brittle, cataclastic, and crystal plastic deformation. Recrystallization at 250 °C irrespective of moisture conditions was also observed. The range and variability of conditions applied in this study, combined with the techniques used to display microstructural features, are unique, and provide insight into an important area of governing deformation mechanism(s) occurring within salt repository applications.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Materials Engineering for High Performance and Durability Proton Exchange Membrane Water Electrolyzers

Proton exchange membrane water electrolyzers (PEMWEs) are expected to play a crucial role in the global green energy transition during the 21st century. They provide a versatile and sustainable solution for generating hydrogen with very high purity in combination with renewable energies, such as solar and wind. Despite their promise, PEMWEs face several critical problems, including high costs, performance limitations, and durability challenges, particularly at low iridium (Ir) loading on the anode. Advancing next-generation PEMWEs requires extensive work on materials engineering of all cell components, including the catalyst layer (CL), membrane, porous transport layer (PTL), bipolar plate (BPP), and gasket. This task must be performed with the complementary contribution of different modeling and characterization techniques. This review presents a critical perspective from academia, research centers, and industry, mapping main developments, remaining gaps, and strategic pathways to advance PEMWE technology. A focus is devoted to key aspects, such as operation at low Ir loading, membrane durability, multiscale transport layers, porous and non-porous flow fields, multiphysics modeling, and multipurpose characterization techniques, which are thoroughly discussed. By unifying these topics, this review provides readers with the essential knowledge to grasp current developments and tackle tomorrow's challenges in PEMWE engineering.

36 MATERIALS SCIENCE↗

Where’s Swimmy?: Mining unique color features buried in galaxies by deep anomaly detection using Subaru Hyper Suprime-Cam data

Abstract We present the Swimmy (Subaru WIde-field Machine-learning anoMalY) survey program, a deep-learning-based search for unique sources using multicolored (grizy) imaging data from the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP). This program aims to detect unexpected, novel, and rare populations and phenomena, by utilizing the deep imaging data acquired from the wide-field coverage of the HSC-SSP. This article, as the first paper in the Swimmy series, describes an anomaly detection technique to select unique populations as “outliers” from the data-set. The model was tested with known extreme emission-line galaxies (XELGs) and quasars, which consequently confirmed that the proposed method successfully selected $\sim\!\! 60\%$–$70\%$ of the quasars and $60\%$ of the XELGs without labeled training data. In reference to the spectral information of local galaxies at z = 0.05–0.2 obtained from the Sloan Digital Sky Survey, we investigated the physical properties of the selected anomalies and compared them based on the significance of their outlier values. The results revealed that XELGs constitute notable fractions of the most anomalous galaxies, and certain galaxies manifest unique morphological features. In summary, deep anomaly detection is an effective tool that can search rare objects, and, ultimately, unknown unknowns with large data-sets. Further development of the proposed model and selection process can promote the practical applications required to achieve specific scientific goals.

Astronomy & Astrophysics↗

Fabrication of Biofunctionalized PVA/PEI Nanofibers for Rare Earth Element Extraction from Coal Mine Tailings

The need for rare earth elements (REEs) for new technology is becoming increasingly stiff with demand at a record high. Biological ligands like lanmodulin and lanthanide binding tags (LBT) have gained noticeable attention for their selective extraction of valuable lanthanide metals in aqueous solution. Building on previous work using microbeads as the extraction media, we propose that nanofibers can improve extraction capabilities. Nanofibers could provide more efficient flow and higher extraction yields due to their high surface area to volume ratio. Relying on thiolmaleimide click chemistry as the biological ligand immobilization route, electrospinning as the nanofiber manufacturing method, and numerous characterization techniques, we produced electrospinning polymer solutions and nanofibers with a maximum neodymium adsorbent capacity of 12.1 μmol/g of extraction media immobilized with LBT. Although the prior work produced an extraction media with an equivalent capacity of ~40 μmol/g, the electrospinning process is much more economical and scalable. This work paves the way for an industrially relevant, environmentally-friendly REE extraction material. Further work is required to produce higher quality nanofibers, to investigate common ion interference with REE adsorbance, and to scale production up to a relevant scale.

01 COAL, LIGNITE, AND PEAT↗

Characterization of Lanthanum Monazite Surface Chemistry and Crystal Morphology through Density Functional Theory and Experimental Approaches

Monazite is a rare earth element (REE)-containing mineral that consists of (REE)PO 4 formal units and is one of the most important sources of these critical materials. The concentration of REEs from mined monazite ore often involves froth flotation, which is a beneficiation process that enhances the efficiency of downstream processing. The effectiveness of froth flotation is largely governed by the ability of collector agents to selectively bind to monazite particles. Thus, a molecular-level understanding of monazite interfacial chemistry is integral to the design of effective collector agents. To address this need, we performed density functional theory (DFT) calculations and a variety of experimental techniques to characterize La-monazite and elucidate its crystal morphology. Interestingly, we find minimal differences in the predicted morphologies of La-monazite for hydrous and anhydrous environments, which are largely dominated by low-index facets (e.g., {110}, {100}, and {010}). Indexing of synthesized La-monazite crystals via X-ray diffraction also uncovers {110} and {100} as the predominant facets. The average surface energies of 0% and 100% water coverage La-monazite crystals were predicted to be 0.87 and 0.76 J/m 2 , respectively, while calorimetry suggests values of 1.30 and 1.15 J/m 2 , respectively. The apparent discrepancies between the theoretical and experimental values are expected and attributed to defects present in physical crystals, in contrast to the perfect mineral surfaces in simulations. The difference in surface energy between the 0% and 100% water coverage morphologies predicted by theory is consistent with the value measured via calorimetry. DFT reveals a wide range of adsorption energies for water across the studied facets, but in all cases, water is predicted to strongly bind to monazite surfaces with an average adsorption energy of -92.7 kJ/mol for a La-monazite single crystal. Finally, this study provides the groundwork necessary for the rational design of froth flotation collector agents by granting molecular-level insight into the predominant facets of monazite.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Arsenic Speciation in Plankton Organisms from Contaminated Lakes: Transformations at the Base of the Freshwater Food Chain

The two complementary techniques high performance liquid chromatography-inductively coupled plasma-mass spectrometry (HPLC-ICP-MS) and X-ray absorption near edge structure (XANES) analysis were used to assess arsenic speciation in freshwater phytoplankton and zooplankton collected from arsenic-contaminated lakes in Yellowknife (Northwest Territories, Canada). Arsenic concentrations in lake water ranged from 7 {micro}g L{sup -1} in a noncontaminated lake to 250 {micro}g L{sup -1} in mine-contaminated lakes, which resulted in arsenic concentrations ranging from 7 to 340 mg kg{sup -1} d.w. in zooplankton organisms (Cyclops sp.) and from 154 to 894 mg kg{sup -1} d.w. in phytoplankton. The main arsenic compounds identified by HPLC-ICP-MS in all plankton were inorganic arsenic (from 38% to 98% of total arsenic). No other arsenic compounds were found in phytoplankton, but zooplankton organisms showed the presence of organoarsenic compounds, the most common being the sulfate arsenosugar, up to 47% of total arsenic, with traces of phosphate sugar, glycerol sugar, methylarsonate (MMA), and dimethylarsinate (DMA). In the uncontaminated Grace Lake, zooplankton also contained arsenobetaine (AB). XANES characterization of arsenic in the whole plankton samples showed AsV-O as the only arsenic compound in phytoplankton, and AsIII-S and AsV-O compounds as the two major inorganic arsenic species in zooplankton. The proportion of organoarsenicals and inorganic arsenic in zooplankton depends upon the arsenic concentration in lakes and shows the impact of arsenic contamination: zooplankton from uncontaminated lake has higher proportions of organoarsenic compounds and contains arsenobetaine, while zooplankton from contaminated area contains mostly inorganic arsenic.

54 ENVIRONMENTAL SCIENCES↗

A measurement of neutrino oscillations with muon neutrinos in the MINOS experiment

Experimental evidence has established that neutrino flavor states evolve over time. A neutrino of a particular flavor that travels some distance can be detected in a different neutrino flavor state. The Main Injector Neutrino Oscillation Search (MINOS) is a long-baseline experiment that is designed to study this phenomenon, called neutrino oscillations. MINOS is based at Fermilab near Chicago, IL, and consists of two detectors: the Near Detector located at Fermilab, and the Far Detector, which is located in an old iron mine in Soudan, MN. Both detectors are exposed to a beam of muon neutrinos from the NuMI beamline, and MINOS measures the fraction of muon neutrinos that disappear after traveling the 734 km between the two detectors. One can measure the atmospheric neutrino mass splitting and mixing angle by observing the energy-dependence of this muon neutrino disappearance. MINOS has made several prior measurements of these parameters. Here I describe recently-developed techniques used to enhance our sensitivity to the oscillation parameters, and I present the results obtained when they are applied to a dataset that is twice as large as has been previously analyzed. We measure the mass splitting Δm 23 2 = (2.32 -0.08 +0.12 ) x 10 -3 eV 2 /c 4 and the mixing angle sin 2 (2θ 32 ) > 0.90 at 90% C.L. These results comprise the world's best measurement of the atmospheric neutrino mass splitting. Alternative disappearance models are also tested. The neutrino decay hypothesis is disfavored at 7.2σ and the neutrino quantum decoherence hypothesis is disfavored at 9.0σ.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Time of Flight Secondary Ion Mass Spectrometry for Characterization of Pt-Coated Porous Transport Layers in PEM Water Electrolyzers

Titanium-based porous transport layers (PTLs) and iridium-based catalyst layers (CLs) are two main components of proton exchange membrane water electrolyzers (PEMWEs). PTLs are typically coated with platinum to minimize interfacial losses and to support long-term operation. Optimizing coatings and the PTL-CL interface requires comprehensive characterization. This study establishes time-of-flight secondary ion mass spectrometry (ToF-SIMS) as a valuable technique for PTL characterization, addressing capabilities and limitations related to PTL morphology. A methodology was developed that uses a Cs + sputter beam for dynamic depth profiling, with data collected in both positive-ion (MCs + ) and negative-ion modes to generate depth profiles, 2D ion maps, and 3D ion reconstructions. ToF-SIMS detected relative differences in platinum-layer thickness between samples; these trends were validated by cross-sectional scanning transmission electron microscope (STEM) measurements and flat-titanium substrate controls. Interfacial oxide layers are identified in both ion modes, with enhanced oxide sensitivity in negative mode. The technique’s high sensitivity enables detection of nanometer-scale coatings and trace impurities within the bulk PTL structure. These results provide a methodological framework for analyzing Pt-coated PTLs, with the potential to extend to other components in PEMWEs and other electrolyzer systems.

36 MATERIALS SCIENCE↗

Detection and Association of Operational Events using DAS and Seismometers (FY 2025 Mid-Year Report)

This mid-year report summarizes ongoing work to identify anomalous vibration signals indicative of potential containment breaches. This work includes compiling continuous seismic datasets and testing and refining underground detection and geolocation techniques. In the first two quarters of FY25, we have completed two project work plan tasks: (1) creating a database of continuous waveforms and ground truth event data from multiple modalities and (2) refining and implementing a detection and association algorithm to create a catalog of anomalous underground activities. This report contains a summary of the seismic database including the continuous seismic data collected by a dense array of surface seismic stations above Pleasant Gap Mine, and continuous seismic data collected using subsurface distributed acoustic sensing (DAS) in the subsurface at Sanford Underground Research Facility (SURF) and the ground truth information gathered from both sites. This report also includes results from refining and applying a dynamic power spectral density detector to both continuous seismic datasets. Finally, the report provides an initial catalog of subsurface operational events from both sensing modalities.

58 GEOSCIENCES↗

Aerial Scene Recognition using Efficient Sparse Representation

Advanced scene recognition systems for processing large volumes of high-resolution aerial image data are in great demand today. However, automated scene recognition remains a challenging problem. Efficient encoding and representation of spatial and structural patterns in the imagery are key in developing automated scene recognition algorithms. We describe an image representation approach that uses simple and computationally efficient sparse code computation to generate accurate features capable of producing excellent classification performance using linear SVM kernels. Our method exploits unlabeled low-level image feature measurements to learn a set of basis vectors. We project the low-level features onto the basis vectors and use simple soft threshold activation function to derive the sparse features. The proposed technique generates sparse features at a significantly lower computational cost than other methods~\cite{Yang10, newsam11}, yet it produces comparable or better classification accuracy. We apply our technique to high-resolution aerial image datasets to quantify the aerial scene classification performance. We demonstrate that the dense feature extraction and representation methods are highly effective for automatic large-facility detection on wide area high-resolution aerial imagery.

Computer Vision↗

Zinc speciation and desorption kinetics in a mining waste impacted tropical soil amended with phosphate

Mining is an important component of the Brazilian economy. However, it may also contribute to environmental problems such as the pollution of soils with zinc and other potentially toxic metals. Our objective was to evaluate changes in the chemical speciation and mobility of Zn in a soil amended with phosphate. Soil samples were collected from a deactivated mining area in the state of Minas Gerais, Brazil, and amended with NH 4 H 2 PO 4 saturated with deionized water to 70 % of maximum water retention and incubated at 25 ± 2 °C in open containers for 60 days. The soil was chemically and mineralogically characterized, and sequential extraction, desorption kinetics, and speciation were carried out using synchrotron bulk-sample and micro–X-ray Absorption Near-Edge Structure (XANES/μ-XANES) spectroscopy at the Zn K-edge, and X-ray fluorescence microprobe analysis (μ-XRF). The combination of μ-XRF and μ-XANES techniques made it possible to identify Zn hotspots in the main species formed after phosphate remediation. The best fit combination for bulk XANES and μ-XANES was observed in Zn-montmorillonite, Zn-kerolite, Zn-ferrihydrite, and gahnite. In the course of phosphate treatment, gahnite, Zn layered double hydroxides (Zn-LDH), Zn3(PO4), and ZnO were identified by bulk XANES, while Zn-ferrihydrite, Zn-montmorillonite, and scholzite were identified by μ-XANES. Zinc in the phosphate-amended soil had the strongest partial correlations (r' > 0.05) with Ni, Co, Fe, Cr, Mn, Si, P, Cd, Pb, and Cd, while the unamended soil showed the strongest correlation with Cu, Pb, Fe, and Si. The application of NH 4 H 2 PO 4 altered Zn speciation and favored an increase in Zn desorption. Here, the most available Zn contents after phosphate amendment were correlated with the release of exchangeable Zn fractions, associated with carbonate and organic matter.

54 ENVIRONMENTAL SCIENCES↗

Final Technical Report. Reactivity of Iron-Bearing Minerals and CO 2 Sequestration and Surface Chemistry of Pyrite. An Interdisciplinary Approach

Over the course of the scientific program, two areas of research were pursued: reactions of iron oxides with supercritical CO 2 and sulfide and surface reactivity of pyrite. The latter area of interest was to understand the chemistry that results when supercritical CO 2 (scCO 2 ) with H 2 S and/or SO 2 in deep saline formations (DFS) contacts iron bearing minerals. Understanding the complexities the sulfur co-injectants introduce is a critical step in developing CO 2 sequestration as a climate-mitigating strategy. The research strategy was to understand macroscopic observations of this chemistry with an atomic/molecular level view using surface analytical techniques. Research showed that the exposure of iron (oxyhdr)oxides (which included ferrihydrite, goethite, and hematite) to scCO 2 in the presence of sulfide led to reactions that formed siderite (FeCO 3 ). The results have important implications for the sequestration of CO 2 via carbonation reactions in the Earth’s subsurface. An earlier area of focus in the project was to understand pyrite oxidation in microscopic detail. This understanding was used to understand macroscopic observations of pyrite reactivity. Results obtained from this research led to a better understanding how pyrite reacts in a range of chemical environments. Geochemical and modern surface science techniques were used to understand the chemistry of pyrite in important environmental conditions. The program relied on a strong integration the results of these techniques to provide a fundamental understanding to the macroscopic chemistry exhibited by pyrite in the environment. Major achievements during these studies included developing an understanding of the surface sites on pyrite that controlled its reactivity under oxidizing conditions. In particular sulfur anion vacancies and/or ferric sites were sites of reactivity. Studies also showed that the adsorption of phospholipid on the surface to selectively suppress the reactivity of these sites could of potential importance for suppressing acid mine drainage in the environment (a problem common to coal-mining sites). Biotic studies showed that microbial activity that promotes the oxidation of pyrite to produce AMD could also be suppressed by the adsorption of phospholipid.

58 GEOSCIENCES↗

Fabrication of porous transport electrodes: Development of quantitative approach for quality control

This work focuses on porous transport electrodes (PTEs), which integrate the anodic catalyst with the adjacent Ti porous transport layer (PTL). Challenges in catalyst deposition on PTLs, particularly at low loadings, motivated this study to evaluate various fabrication methods and characterization approaches. This work investigated Pt-treated PTLs coated with Ir-based catalysts using several common methods, including airbrush coating, rod coating, ultrasonic spray coating, electrodeposition, and sputter deposition, with catalyst loadings ranging from 2.9 to 0.1 mg/cm 2 , providing the opportunity for comparisons across a large set of samples produced by different methods. Two widely accessible characterization techniques: X-ray computed tomography (XCT) and scanning electron microscopy energy dispersive X-ray spectroscopy (SEM-EDS) were explored. Initial evaluation of selected samples with XCT provided qualitative insights into catalyst distribution, however comprehensive quantitative analysis was limited. SEM-EDS enabled detailed information on the catalyst distribution both qualitatively and quantitatively using two metrics. Atomic and surface area % ratios of Pt:Ir and Ti:Ir revealed trends in catalyst loading and losses into the PTL pores, as well as evaluating the homogeneity of catalyst coatings. The analysis demonstrated that ultrasonic spray coating, electrodeposition, and sputter coating produced the most homogeneous coatings, with minimal catalyst losses observed for electrodeposition and sputter coating. By adapting common techniques with novel, standardized methodologies, this work establishes a universally applicable framework for cross-study comparison of PTEs. The SEM-EDS approach provides a practical, accessible tool for PTE characterization and contributes a reference dataset supporting both research development and rapid quality control.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Visualization of hydraulic fracture using physics-informed clustering to process ultrasonic shear waves

Ultrasonic transmission is sensitive to the spatial variation in mechanical properties of materials due to the presence of cracks/fractures. Wave propagation through fractured media introduces changes in the frequency content, travel time and transmission coefficient of the wave. A workflow based on physics-informed unsupervised learning is developed to process the transmitted ultrasonic-shear waveforms to non-invasively visualize the geomechanical alterations due to hydraulic fracturing. Novelty of the work involves the assignment of both statistically consistent and physically consistent clusters to the measurements of shear waveforms acquired across the one axial and two frontal planes. Physically consistent/relevant information is incorporated by considering the travel time of the peak of spectral energy and transmission coefficient of the transmitted waveform. The proposed workflow generates maps of geomechanical alterations across the frontal and axial planes of the sample. The outputs of the workflow are in good agreement with independent techniques viz. acoustic emission and X-ray computed tomography. Finally, the proposed workflow can be adapted for improved fracture characterization in the subsurface when processing sonic-logging, cross-wellbore seismic or surface seismic waveform data.

42 ENGINEERING↗

Ionic gold in calcrete revealed by LA-ICP-MS, SXRF and XANES

Highly anomalous Au concentrations in calcrete were discovered in 1987 at the Bounty Gold Deposit, Western Australia. A strong correlation was noted between the Ca, Mg, Sr and Au in soil profiles which have not only attracted the interest of mineral explorers but also chemists, soil scientists, metallurgists and climatologists. Gold has been considered an inert element and so its strong association with the alkaline earth group of relatively mobile elements is both remarkable and intriguing. Despite widespread interest, there have been few published papers on the Au-calcrete phenomenon. Here, we present work conducted on calcareous soil samples from above the Bounty mineralization in Western Australia, prior to mining. Using SXRF (synchrotron X-ray fluorescence) and XANES (micro-X-ray absorption near-edge structure), we have shown for the first time the distribution of Au in calcrete and that it occurs in both particulate and ionic form. Much of the ionic Au associated with Br is found in a root tubule. The observations are consistent with an evapotranspiration model for the formation of Au in the calcrete; Au has been mobilized then precipitated as vadose water has been removed from the soil by trees and shrubs. While the association between Au and Ca is very strong in bulk sample analyses down the soil profile, other detailed analyses on sub-samples using wet chemical, LA-ICP-MS (laser ablation inductively coupled mass spectrometry) and SXRF techniques show that it is not apparent at the sub-millimeter scale. This suggests that the Au and Ca are behaving similarly but independently and they do not (at the {micro}m scale) co-precipitate with carbonate minerals. These results corroborate other studies that suggest biotic influences can affect the mobilization and distribution of Au in surficial materials. Water-extractable Au in calcrete has been reported previously and the ionic Au described in this study likely represents that soluble component. The presence of easily solubilised Au in soils has been widely discussed and exploited for mineral exploration.

54 ENVIRONMENTAL SCIENCES↗

Detecting Silent Data Corruption for Extreme-Scale Applications through Data Mining

Supercomputers allow scientists to study natural phenomena by means of computer simulations. Next-generation machines are expected to have more components and, at the same time, consume several times less energy per operation. These trends are pushing supercomputer construction to the limits of miniaturization and energy-saving strategies. Consequently, the number of soft errors is expected to increase dramatically in the coming years. While mechanisms are in place to correct or at least detect some soft errors, a significant percentage of those errors pass unnoticed by the hardware. Such silent errors are extremely damaging because they can make applications silently produce wrong results. In this work we propose a technique that leverages certain properties of high-performance computing applications in order to detect silent errors at the application level. Our technique detects corruption solely based on the behavior of the application datasets and is completely application-agnostic. We propose multiple corruption detectors, and we couple them to work together in a fashion transparent to the user. We demonstrate that this strategy can detect the majority of the corruptions, while incurring negligible overhead. We show that with the help of these detectors, applications can have up to 80% of coverage against data corruption.

97 MATHEMATICS AND COMPUTING↗

Multi-Level Structural Damage Characterization Using Sparse Acoustic Sensor Networks and Knowledge Transferred Deep Learning

Standard structural health monitoring techniques face well-known difficulties for comprehensive defect diagnosis in real-world structures that have structural, material, or geometric complexity. This motivates the exploration of machine-learning-based structural health monitoring methods in complex structures. However, creating sufficient training data sets with various defects is an ongoing challenge for data-driven machine (deep) learning algorithms. The ability to transfer the knowledge of a trained neural network from one component to another or to other sections of the same component would drastically reduce the required training data set. Also, it would facilitate computationally inexpensive machine learning based inspection systems. In this work, a machine-learning-based multi-level damage characterization is demonstrated with the ability to transfer trained knowledge within the sparse sensor network. A novel network spatial assistance and an adaptive convolution technique are proposed for efficient knowledge transfer within the deep learning algorithm. Proposed structural health monitoring method is experimentally evaluated on an aluminum plate with artificially induced defects. It was observed that the method improves the performance of knowledge transferred damage characterization by 50% during localization and 24% during severity assessment. Further, experiments using time windows with and without multiple edge reflections are studied. Results reveal that multiply scattered waves contain rich and deterministic defect signatures that can be mined using deep learning neural networks, improving the accuracy of both identification and quantification. In the case of a fixed sensor network, using multiply scattered waves shows 100% prediction accuracy at all levels of damage characterization.

36 MATERIALS SCIENCE↗