Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “factorization machine”

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 685 records · Page 38
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Global Sensitivity Analysis of a Reactive Transport Model for Mineral Scale Formation During Hydraulic Fracturing

Injection of water-based hydraulic fracturing fluid (HFF) into tight shale gas/oil formations can increase formation permeability and enhance production rates, but this process frequently causes mineral scale formation that can occlude pore space and hinder flow. To identify the most important factors that control the formation of mineral scales, we applied a novel global sensitivity analysis method—distance-based generalized sensitivity analysis (DGSA)—to a reactive transport model (RTM) that was previously built and calibrated to simulate precipitation of barite [BaSO4] and iron (hydr)oxide [Fe(OH) 3 ] in shale matrices and on fracture surfaces. Reactive transport simulations were run with model parameters randomly sampled based on assigned uncertainties. Modeling results for barite and Fe(OH)3 formation were clustered using machine-learning algorithms. A list of ranked critical input parameters was obtained after statistical quantification of cumulative distribution functions of input parameters. We found that barite formation is most sensitive to the rate of sulfate ion generation, which is determined by the pyrite dissolution rate coefficient and oxidant availability. In addition, barite formation is sensitive to the initial amounts of barite in HFF and shale, followed by barite thermodynamics/kinetics. For Fe(OH) 3 formation, the ranked factors are Fe(OH)3 precipitation rate coefficients, initial HFF pH, initial Fe(OH) 3 amount in HFF, and oxidant availability. Overall, our results provide insights into managing mineral scale formation during hydraulic fracturing to enhance production. Meanwhile, this study serves as an example of global sensitivity analysis of RTMs using the efficient, straightforward, and open-source DGSA method.

58 GEOSCIENCES↗
DOE OSTI2026

Machine learning insights into microstructural origins of transport and mechanical properties in porous microstructures

Multifunctional porous materials are increasingly needed across various fields, but their complex microstructures create significant challenges due to the intricate microstructure-property relationships. This complexity, combined with limitations of traditional analysis methods, hinders efforts to understand and optimize microstructure–property relationships. Here, to address this, we integrate physics-based mesoscale modeling with interpretable machine learning (ML) to uncover how microstructural features govern effective diffusivity and elastic modulus. At constant porosity, we show diffusivity varies by over 150 × and modulus by ∼50 ×, highlighting the power of microstructure engineering. Statistical analysis reveals bimodal behavior in diffusivity and unimodal in modulus. ML identifies connectivity as the dominant factor, while modulus is also sensitive to domain size and feature interactions. Controlled simulations further highlight domain shape as a critical feature for modulus. This framework enables efficient exploration of microstructure-property correlations, offering new insights to guide the design of advanced porous materials.

Bicontinuous microstructure↗
DOE OSTI2022

Novel Application of Machine Learning Techniques for Rapid Source Apportionment of Aerosol Mass Spectrometer Datasets

In this work, we apply machine learning approaches sparse multinomial logistic regression to classify aerosol mass spectrometer (AMS) unit mass resolution (UMR) data followed by an ensemble regression technique for source apportionment of organic aerosols (OA). The classifier was trained on 60 well characterized laboratory and positive matrix factorization (PMF) deconvolved reference spectra to identify eight OA types. These include four laboratory-derived secondary organic aerosol (SOA) spectra, which include isoprene photooxidation SOA, isoprene epoxydiols (IEPOX) SOA, a monoterpene SOA type that includes a-pinene and ß-pinene SOA, and aromatic SOA from oxidation of naphthalene and m-xylene precursors, as well as PMF deconvolved spectra for three primary organic aerosol (POA) types, namely, hydrocarbon-like organic aerosol (HOA), biomass burning organic aerosol (BBOA), and cooking OA (COA), and a more oxidized oxygenated OA type (MO-OOA). A 5-fold cross-validation strategy, repeated 10 times, was used to assess the classifier’s performance. The classifier had high classification accuracy for COA, aromatic SOA, and isoprene SOA spectra but incorrectly classified ~9% by number of MO-OOA spectra as BBOA, 12% of BBOA spectra as HOA (and vice versa), and 18% of IEPOX-SOA spectra as aromatic SOA. Next, an ensemble regression model was trained on an artificially generated dataset consisting of mixtures of different OA types to assess its ability to predict fractional mass abundances from classification probabilities of various OA species obtained from the multinomial logistic regression classifier trained on the reference spectra. Ultimately, the proposed approach was applied for source apportionment of aircraft-based AMS measurements of OA UMR spectra during the HI-SCALE field campaign. On two representative days (May 6th and 18th, 2016), the algorithm determined that ~50-60% of OA by mass was MO-OOA, which represented a highly aged organic aerosol mixture from different sources. On both days, BBOA was determined to contribute less than 10% to OA by mass. However, on May 18th, the aromatic SOA fraction was higher compared to that on May 6th. The proposed approach is capable of rapidly analyzing AMS data in real time, making it suitable for applications where rapid source apportionment of AMS OA spectra is desirable.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

A Causal Inference Model Based on Random Forests to Identify the Effect of Soil Moisture on Precipitation

Soil moisture influences precipitation mainly through its impact on land–atmosphere interactions. Understanding and correctly modeling soil moisture–precipitation (SM–P) coupling is crucial for improving weather forecasting and subseasonal to seasonal climate predictions, especially when predicting the persistence and magnitude of drought. However, the sign and spatial structure of SM–P feedback are still being debated in the climate research community, mainly due to the difficulty in establishing causal relationships and the high degree of nonlinearity in land–atmosphere processes. To this end, we developed a causal inference model based on the Granger causality analysis and a nonlinear machine learning model. This model includes three steps: nonlinear anomaly decomposition, nonlinear Granger causality analysis, and evaluation of the quality of SM–P feedback, which eliminates the nonlinear response of interannual and seasonal variability and the memory effects of climatic factors and isolates the causal relationship of local SM–P feedback. We applied this model by using National Climate Assessment–Land Data Assimilation System (NCA-LDAS) datasets over the United States. Here, the results highlight the importance of nonlinear atmosphere responses in land–atmosphere interactions. In addition, the strong feedback over the southwestern United States and the Great Plains both highlight the impacts of topographic factors rather than only the sensitivity of evapotranspiration to soil moisture. Furthermore, the SM–P index defined by our framework is used to benchmark Earth system models (ESMs), which provides a new metric for efficiently identifying potential model biases in modeling local land–atmosphere interactions and may help the development of ESMs in improving simulations of water cycle variability and extremes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Physiochemical Machine Learning Models Predict Operational Lifetimes of CH3NH3PbI3 Perovskite Solar Cells

Halide perovskites are promising photovoltaic (PV) materials with the potential to lower the cost of electricity and greatly expand the penetration of PV if they can demonstrate long-term stability under illumination in the presence of moisture and oxygen. The solar cell service lifetime as quantified by the T80 (the time required for the power conversion efficiency to drop to 80% of its starting value) is a useful metric to assess stability. The T80 for utility, commercial, or residential PV systems needs to be several decades in order to yield low-cost electricity, and thus it is not practical to directly measure the T80. It would be useful if T80 could be predicted from the initial dynamics of a solar cell’s performance, but until now no models have been developed to forecast T80. In this work, we report the development of machine learning models to predict T80 of ITO/NiOx/CH3NH3PbI3/C60/BCP/Ag solar cells operating at maximum power point under 1-sun equivalent photon flux in air at varying temperatures and relative humidities. Efficiency losses are driven by short-circuit current and fill factor, indicating that chemical decomposition of the perovskite is a major contributor to degradation. Spatial patterns evident from in situ dark field optical microscopy suggest that the electric field gradient at device edges plays a significant role in perovskite decomposition, along with photochemical reactions with O2 and H2O. Models are trained using a menu of features from three distinct categories: (i) features based on measurements of the initial rates of change of device parameters, (ii) features based on the ambient conditions during operation (temperature, & partial pressure of H2O), and (iii) features based on underlying physics and chemistry. We show that a theory-based physiochemical feature derived from a model of the chemical reaction kinetics of the rate of degradation of the CH3NH3PbI3 is particularly valuable for prediction. This physiochemical feature was selected as the first or second most dominant feature in the best performing models. With a dataset consisting of 45 accelerated degradation experiments with T80 that range over a factor of 30, the model predicts T80 with an accuracy of about 40% (|predicted T80 - observed T80| / observed T80) on samples not used in training. This hybrid ML approach should be effective when applied to other compositions, device architectures, and advanced packaging schemes.

14 SOLAR ENERGY↗
DOE OSTI2022

Physiochemical Machine Learning Models Predict Operational Lifetimes of CH3NH3PbI3 Perovskite Solar Cells

Halide perovskites are promising photovoltaic (PV) materials with the potential to lower the cost of electricity and greatly expand the penetration of PV if they can demonstrate long-term stability under illumination in the presence of moisture and oxygen. The solar cell service lifetime as quantified by the T80 (the time required for the power conversion efficiency to drop to 80% of its starting value) is a useful metric to assess stability. The T80 for utility, commercial, or residential PV systems needs to be several decades in order to yield low-cost electricity, and thus it is not practical to directly measure the T80. It would be useful if T80 could be predicted from the initial dynamics of a solar cell’s performance, but until now no models have been developed to forecast T80. In this work, we report the development of machine learning models to predict T80 of ITO/NiOx/CH3NH3PbI3/C60/BCP/Ag solar cells operating at maximum power point under 1-sun equivalent photon flux in air at varying temperatures and relative humidities. Efficiency losses are driven by short-circuit current and fill factor, indicating that chemical decomposition of the perovskite is a major contributor to degradation. Spatial patterns evident from in situ dark field optical microscopy suggest that the electric field gradient at device edges plays a significant role in perovskite decomposition, along with photochemical reactions with O2 and H2O. Models are trained using a menu of features from three distinct categories: (i) features based on measurements of the initial rates of change of device parameters, (ii) features based on the ambient conditions during operation (temperature, & partial pressure of H2O), and (iii) features based on underlying physics and chemistry. We show that a theory-based physiochemical feature derived from a model of the chemical reaction kinetics of the rate of degradation of the CH3NH3PbI3 is particularly valuable for prediction. This physiochemical feature was selected as the first or second most dominant feature in the best performing models. With a dataset consisting of 45 accelerated degradation experiments with T80 that range over a factor of 30, the model predicts T80 with an accuracy of about 40% (|predicted T80 - observed T80| / observed T80) on samples not used in training. This hybrid ML approach should be effective when applied to other compositions, device architectures, and advanced packaging schemes.

14 SOLAR ENERGY↗
DOE OSTI2024

Error and Correction Analysis for the FFA@CEBAF Energy Upgrade

An energy upgrade design for the Continuous Electron Beam Accelerator Facility (CEBAF) is under development, using fixed field alternating gradient (FFA) return arcs to recirculate electron beam up to an additional five times through the accelerating structures at CEBAF. A necessary component of any large accelerator is a beam steering and optical correction system. Small environmental changes and system errors can lower beam quality or even shut down the machine; and in pursuit of the scientific mission of JLab, high quality electron beams must be delivered to the experimental halls on a predictable schedule. Correction in the novel FFA arcs of the current upgrade design is complicated by several factors. These complexities inform the choice of correction algorithm structure and parameter values. A baseline algorithm in addition to diagnostic and correction hardware configuration is presented. The effect of this correction protocol is shown with respect to estimated errors, and several possible extensions of the algorithm are discussed. This work presents an important proof of concept for the FFA@CEBAF design effort, and provides a functional correction strategy which may be simply adjusted and optimized for future design changes.

Coxe, Alex [Old Dominion Univ., Norfolk, VA (Unite↗
DOE OSTI2022

Cracking Behavior of Irradiated Ex-Plant Materials

The cracking behavior of reactor core internal materials is a key factor for the long-term safety and availability of light water reactors. Subject to high-temperature coolant, thermal and mechanical loading, and neutron irradiation, reactor core internal materials are susceptible to several degradation mechanisms during power operation. To understand the long-term effects of neutron irradiation, irradiated materials harvested from a decommissioned reactor were studied for their cracking susceptibility and fracture resistance. Small compact-tension specimens were machined from different locations on the baffle plates with irradiation doses ranging from 0.06 to 48.5 dpa. Crack growth rate and fracture toughness J-resistance (J-R) curve tests were performed on these specimens in a simulated light water reactor environment, and the fracture morphologies of the tested samples were examined with a scanning electron microscope. All samples behaved similarly under cyclic loading, and no deteriorated corrosion-fatigue behavior was observed in the test environments. Under constant load, no elevated cracking susceptibility was observed either in these samples at low stress intensity factors. However, a rapid cracking behavior featuring very high crack growth rates was observed in several high-dose samples. This rapid crack growth behavior was activated when the applied stress intensity factor, K, was sufficiently high. Once activated, the rapid crack growth remained steady even when the K was lowered to the pre-activation level and stopped only when the K was reduced drastically. Yet, the rapid crack growth could be re-activated repeatedly in these samples by raising the K above the initial activation level. This rapid cracking response may be related to the severe irradiation embrittlement experienced by these samples. The J-R curve tests performed with these samples showed a significant decline in fracture toughness with increasing neutron irradiation. Local cracking at the crack tip facilitated by the severe irradiation embrittlement of this material may be responsible for the rapid crack growth behavior demonstrated by this decommissioned material.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Micromechanical Surrogate Machine Learning Model for Creep Deformation Modeling

Process variability during the manufacture of gas turbine engine hot section components can significantly affect the material’s resulting microstructure. In casting, for instance, geometric variation within a component (thin sections versus thick sections, radial location) influences cooling rates and the resulting grain size. The high temperature creep response is known to be sensitive to grain size owing to a diffusional creep mechanism which occurs more readily along grain boundaries. Microstructural variation correspondingly drives mechanical behavior which propagates into component scale performance uncertainty. These factors are essential when planning inspection, maintenance, and repair strategies within a reliability framework. These benefits provide opportunities to increase overall energy efficiency through refined margins. Critically, there is an opportunity to bolster existing data-driven reliability models using physics-driven process-structure-property relations. Here we present recent work establishing a framework for evaluating the probabilistic creep performance of high-temperature materials. A novel microstructure-sensitive crystal plasticity finite element model is established that captures both grain boundary and crystallographic deformation effects. The computationally expensive physics model is calibrated using a statistical approach and this high-fidelity model is subsequently used to train a computationally efficient machine learning surrogate model. The surrogate model is essential for sampling a large ensemble of simulated structure-property pair results. The ensemble data are then mined to extract salient trends to be incorporated into a microstructure-sensitive reliability model. The proposed approach represents a novel way to capture microstructure-sensitive trends from physics-based models within a modern reliability framework.

Fernandez-Zelaia, Patxi [ORNL]↗
DOE OSTI2022

Machine learning and shallow groundwater chemistry to identify geothermal prospects in the Great Basin, USA

This study discovers various geothermal prospects in the Great Basin, USA based on shallow groundwater chemical (geochemical) data. The geochemical data are expected to include hidden (latent) information that is a proxy for geothermal prospectivity. We processed the sparse geochemical data in the Great Basin at 14,341 locations including 18 attributes. Next, a non-negative matrix factorization with customized k-means clustering is applied to the geochemical data matrix that automatically finds three hidden geothermal signatures representing modestly, moderately, and highly confident geothermal prospects. The algorithm also evaluated the probability of occurrence of these types of resources through the studied region. There is a consistency between regional geothermal prospectivity as estimated by our ML methodology and the traditional play fairway analysis conducted over a portion of the study area. We also identify the dominant data attributes associated with each signature. Finally, our ML analyses allow us to reconstruct attributes from sparse into continuous over the study domain. The predicted continuous attributes can be used for future detailed geothermal explorations in the Great Basin.

15 GEOTHERMAL ENERGY↗
DOE OSTI2020

Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization

Abstract Atmospheric radiation is the main driver of weather and climate, yet due to a complicated absorption spectrum, the precise treatment of radiative transfer in numerical weather and climate models is computationally unfeasible. Radiation parameterizations need to maximize computational efficiency as well as accuracy, and for predicting the future climate many greenhouse gases need to be included. In this work, neural networks (NNs) were developed to replace the gas optics computations in a modern radiation scheme (RTE+RRTMGP) by using carefully constructed models and training data. The NNs, implemented in Fortran and utilizing BLAS for batched inference, are faster by a factor of 1–6, depending on the software and hardware platforms. We combined the accelerated gas optics with a refactored radiative transfer solver, resulting in clear‐sky longwave (shortwave) fluxes being 3.5 (1.8) faster to compute on an Intel platform. The accuracy, evaluated with benchmark line‐by‐line computations across a large range of atmospheric conditions, is very similar to the original scheme with errors in heating rates and top‐of‐atmosphere radiative forcings typically below 0.1 K day −1 and 0.5 W m −2 , respectively. These results show that targeted machine learning, code restructuring techniques, and the use of numerical libraries can yield material gains in efficiency while retaining accuracy.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2016

Finding new perovskite halides via machine learning

Advanced materials with improved properties have the potential to fuel future technological advancements. However, identification and discovery of these optimal materials for a specific application is a non-trivial task, because of the vastness of the chemical search space with enormous compositional and configurational degrees of freedom. Materials informatics provides an efficient approach toward rational design of new materials, via learning from known data to make decisions on new and previously unexplored compounds in an accelerated manner. Here, we demonstrate the power and utility of such statistical learning (or machine learning, henceforth referred to as ML) via building a support vector machine (SVM) based classifier that uses elemental features (or descriptors) to predict the formability of a given ABX 3 halide composition (where A and B represent monovalent and divalent cations, respectively, and X is F, Cl, Br, or I anion) in the perovskite crystal structure. The classification model is built by learning from a dataset of 185 experimentally known ABX 3 compounds. After exploring a wide range of features, we identify ionic radii, tolerance factor, and octahedral factor to be the most important factors for the classification, suggesting that steric and geometric packing effects govern the stability of these halides. As a result, the trained and validated models then predict, with a high degree of confidence, several novel ABX 3 compositions with perovskite crystal structure.

36 MATERIALS SCIENCE↗
DOE OSTI2015

Impact of a narrow limiter SOL heat flux channel on the ITER first wall panel shaping

We report the inboard limiters for ITER were initially designed on the assumption that the parallel heat flux density in the scrape-off layer (SOL) could be approximated by a single exponential with decay length λ q . This assumption was found not to be adequate in 2012, when infra-red (IR) thermography measurements on the inner column during JET limiter discharges clearly revealed the presence of a narrow heat flux channel adjacent to the last closed flux surface. This near-SOL decay occurs with λ q ~ few mm, much shorter than the main SOL λ q , and can raise the heat flux at the limiter apex a factor up to ~4 above the value expected from a single, broader exponential. The original logarithmically shaped ITER inner wall first wall panels (FWPs) would be unsuited to handling the power loads produced by such a narrow feature. A multi-machine study involving the C-Mod, COMPASS, DIII-D and TCV tokamaks, employing inner wall IR measurements and/or inner wall reciprocating probes, was initiated to investigate the narrow limiter SOL heat flux channel. This paper describes the new results which have provided an experimental database for the narrow feature and presents an ITER inner wall FWP toroidal shape optimized for a double-exponential profile with λ q = 4 (narrow feature) and 50 mm (main-SOL), the latter also derived from a separate multi-machine database constituted recently within the International Tokamak Physics Activity. It is shown that the new shape allows the power handling capability of the original shape design to be completely recovered for a wide variety of limiter start-up equilibria in the presence of a narrow feature, even taking assembly tolerances into account. Additionally, it is further shown that the new shape has the interesting property of both mitigating the impact of the narrow feature and resulting in only a very modest increase in heat load, compared to the current design, if the narrow feature is not eventually found on ITER.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2017

Investigating the radial structure of axisymmetric fluctuations in the TCV tokamak with local and global gyrokinetic GENE simulations

Axisymmetric (n=0) density fluctuations measured in the TCV tokamak are observed to possess a frequency f<sub>0</sub> which is either varying (radially dispersive oscillations) or a constant over a large fraction of the plasma minor radius (radially global oscillations) as reported in a companion paper [Z. Huang <i>et al.</i>, this issue]. Given that f<sub>0</sub> scales with the sound speed and given the poloidal structure of density fluctuations, these oscillations were interpreted as Geodesic Acoustic Modes, even though f<sub>0</sub> is in fact smaller than the local linear GAM frequency f<sub>GAM</sub> . In this work we employ the Eulerian gyrokinetic code GENE to simulate TCV relevant conditions and investigate the nature properties of these oscillations, in particular their relation to the safety factor profile. Local and global simulations are carried out and a good qualitative agreement is observed between experiments and simulations. By varying also the plasma temperature and density profiles, we conclude that a variation of the edge safety factor alone is not sufficient to induce a transition from global to radially inhomogeneous oscillations, as was initially suggested by experimental results. This transition appears instead to be the combined result of variations in the different plasma profiles, collisionality and finite machine size effects. In conclusion, simulations also show that radially global GAM-like oscillations can be observed in all fluxes and fluctuation fields, suggesting that they are the result of a complex nonlinear process involving also finite toroidal mode numbers and not just linear global GAM eigenmodes.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Probabilistic simulation of quantum circuits using a deep-learning architecture

The fundamental question of how to best simulate quantum systems using conventional computational resources lies at the forefront of condensed matter and quantum computation. It impacts both our understanding of quantum materials and our ability to emulate quantum circuits. Here we present an exact formulation of quantum dynamics via factorized generalized measurements which maps quantum states to probability distributions with the advantage that local unitary dynamics and quantum channels map to local quasistochastic matrices. This representation provides a general framework for using state-of-the-art probabilistic models in machine learning for the simulation of quantum many-body dynamics. Using this framework, we have developed a practical algorithm to simulate quantum circuits using an attention network based on a powerful neural network ansatz responsible for the most recent breakthroughs in natural language processing. We demonstrate our approach by simulating circuits that build Greenberger-Horne-Zeilinger and linear graph states of up to 60 qubits, as well as a variational quantum eigensolver circuit for preparing the ground state of the transverse field Ising model on several system sizes. Our methodology constitutes a modern machine learning approach to the simulation of quantum physics with applicability both to quantum circuits as well as other quantum many-body systems.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

First measurement of Λ c + production down to p T = 0 in p p and p -Pb collisions at s N N = 5.02 TeV

The production of prompt $Λ_{c}^{+}$ baryons has been measured at midrapidity in the transverse momentum interval 0 < p T < 1 GeV/c for the first time, in pp and p–Pb collisions at a center-of-mass energy per nucleon-nucleon collision $\sqrt{s_{NN}}$ = 5.02 TeV . The measurement was performed in the decay channel $Λ_{c}^{+}$ → p$K_{S}^{0}$ by applying new decay reconstruction techniques using a Kalman-Filter vertexing algorithm and adopting a machine-learning approach for the candidate selection. The p T -integrated $Λ_{c}^{+}$ production cross sections in both collision systems were determined and used along with the measured yields in Pb–Pb collisions to compute the p T -integrated nuclear modification factors R pPb and R AA of $Λ_{c}^{+}$ baryons, which are compared to model calculations that consider nuclear modification of the parton distribution functions. The $Λ_{c}^{+}$/D 0 baryon-to-meson yield ratio is reported for pp and p–Pb collisions. Comparisons with models that include modified hadronization processes are presented, and the implications of the results on the understanding of charm hadronization in hadronic collisions are discussed. A significant (3.7σ) modification of the mean transverse momentum of $Λ_{c}^{+}$ baryons is seen in p –Pb collisions with respect to pp collisions, while the p T -integrated $Λ_{c}^{+}$/D 0 yield ratio was found to be consistent between the two collision systems within the uncertainties.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2023

Journey Over Destination: Differentiable Sensor Placement Enhances Generalization [Poster]

The challenge of reconstructing spatial fields that change over time from limited sensor data has been a focal point for many research studies. Various machine learning methods have been used in attempts to address this complex issue, including convolutional neural networks. All of the proposed methods share a common requirement that the user needs to manually determine the sensor positions. This requirement remains a limiting factor in the ongoing quest for efficient learning and accurate field reconstruction. This study aims to present a method that enables a model to optimize sensor positions via backpropagation, thereby facilitating the model’s exploration of the spatial domain and enhancing sensor positioning effectively. Indexing naturally incorporates discrete decisions. This operation is nondifferentiable which is a requirement for the application of gradient-based optimization methods. We showcased its effectiveness by training an attention-based neural network, which achieved top-tier performance on two separate datasets. To our knowledge, this represents the first fully end-to-end differentiable workflow for enhancing sensor placement within a neural network model.

58 GEOSCIENCES↗
DOE OSTI2022

Ch3MS-RF: a random forest model for chemical characterization and improved quantification of unidentified atmospheric organics detected by chromatography–mass spectrometry techniques

Abstract. The chemical composition of ambient organic aerosols plays a critical role in driving their climate and health-relevant properties and holds important clues to the sources and formation mechanisms of secondary aerosol material. In most ambient atmospheric environments, this composition remains incompletely characterized, with the number of identifiable species consistently outnumbered by those that have no mass spectral matches in the literature or the National Institute of Standards and Technology/National Institutes of Health/Environmental Protection Agency (NIST/NIH/EPA) mass spectral databases, making them nearly impossible to definitively identify. This creates significant challenges in utilizing the full analytical capabilities of techniques which separate and generate spectra for complex environmental samples. In this work, we develop the use of machine learning techniques to quantify and characterize novel, or unidentifiable, organic material. This work introduces Ch3MS-RF (Chemical Characterization by Chromatography–Mass Spectrometry Random Forest Modeling), an open-source, R-based software tool, for efficient machine-learning-enabled characterization of compounds separated in chromatography–mass spectrometry applications but not identifiable by comparison to mass spectral databases. A random forest model is trained and tested on a known 130 component representative external standard to predict the response factors of novel environmental organics based on position in volatility–polarity space and mass spectrum, enabling the reproducible, efficient, and optimized quantification of novel environmental species. Quantification accuracy on a reserved 20 % test set randomly split from the external standard compound list indicates that random forest modeling significantly outperforms the commonly used methods in both precision and accuracy, with a median response factor percent error of −2 %, for modeled response factors, compared to > 15 %, for typically used proxy assignment-based methods. Chemical properties modeling, evaluated on the same reserved 20 % test set and an extrapolation set of species identified in ambient organic aerosol samples collected in the Amazon rainforest, also demonstrate robust performance. Extrapolation set property prediction mean absolute errors for carbon number, oxygen to carbon ratio (O : C), average carbon oxidation state (OSc‾), and vapor pressure are 1.8, 0.15, 0.25, and 1.0 (log(atm)), respectively. Extrapolation set out-of-sample R2 for all properties modeled are above 0.75, with the exception of vapor pressure. While predictive performance for vapor pressure is less robust compared to the other chemical properties modeled, random-forest-based modeling was significantly more accurate than other commonly used methods of vapor pressure prediction, decreasing the mean vapor pressure prediction error to 0.24 (log(atm)) from 0.55 (log(atm)) (chromatography-based vapor pressure prediction) and 1.2 (log(atm)) (chemical formula-based vapor pressure prediction). The random forest model significantly advances an untargeted analysis of the full scope of chemical speciation yielded by two-dimensional gas chromatography (GCxGC-MS) techniques and can be applied to gas chromatography coupled with electron ionization mass spectrometry (GC-MS) as well. It enables the accurate estimation of key chemical properties commonly utilized in the atmospheric chemistry community, which may be used to more efficiently identify important tracers for further individual analysis and to characterize compound populations uniquely formed under specific ambient conditions.

54 ENVIRONMENTAL SCIENCES↗