Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Sensor Network”

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 1,279 records · Page 71

Machine learning without a processor: Emergent learning in a nonlinear analog network

Standard deep learning algorithms require differentiating large nonlinear networks, a process that is slow and power-hungry. Electronic contrastive local learning networks (CLLNs) offer potentially fast, efficient, and fault-tolerant hardware for analog machine learning, but existing implementations are linear, severely limiting their capabilities. These systems differ significantly from artificial neural networks as well as the brain, so the feasibility and utility of incorporating nonlinear elements have not been explored. Here, we introduce a nonlinear CLLN—an analog electronic network made of self-adjusting nonlinear resistive elements based on transistors. We demonstrate that the system learns tasks unachievable in linear systems, including XOR (exclusive or) and nonlinear regression, without a computer. We find our decentralized system reduces modes of training error in order (mean, slope, curvature), similar to spectral bias in artificial neural networks. The circuitry is robust to damage, retrainable in seconds, and performs learned tasks in microseconds while dissipating only picojoules of energy across each transistor. This suggests enormous potential for fast, low-power computing in edge systems like sensors, robotic controllers, and medical devices, as well as manufacturability at scale for performing and studying emergent learning.

Science & Technology - Other Topics↗

Smart Pixels: In-pixel AI for on-sensor data filtering

We present a smart pixel prototype readout integrated circuit (ROIC) designed in CMOS 28 nm bulk process, with in-pixel implementation of an artificial intelligence (AI) / machine learning (ML) based data filtering algorithm designed as proof-of-principle for a Phase III upgrade at the Large Hadron Collider (LHC) pixel detector. The first version of the ROIC consists of two matrices of 256 smart pixels, each 25$\times$25 µm\textsuperscript{2} in size. Each pixel consists of a charge-sensitive preamplifier with leakage current compensation and three auto-zero comparators for a 2-bit flash-type ADC. The frontend is capable of synchronously digitizing the sensor charge within 25 ns. Measurement results show an equivalent noise charge (ENC) of $\sim$30e\textsuperscript{-} and a total dispersion of $\sim$100e\textsuperscript{-} The second version of the ROIC uses a fully connected two-layer neural network (NN) to process information from a cluster of 256 pixels to determine if the pattern corresponds to highly desirable high-momentum particle tracks for selection and readout. The digital NN is embedded in-between analog signal processing regions of the 256 pixels without increasing the pixel size and is implemented as fully combinatorial digital logic to minimize power consumption and eliminate clock distribution, and is active only in the presence of an input signal. The total power consumption of the neural network is $\sim$ 300 $\mu$W. The NN performs momentum classification based on the generated cluster patterns and even with a modest momentum threshold, it is capable of 54.4\% – 75.4\% total data rejection, opening the possibility of using the pixel information at 40MHz for the trigger. The total power consumption of analog and digital functions per pixel is $\sim$ 6 $\mu$W per pixel, which corresponds to $\sim$ 1 W/cm\textsuperscript{2} staying within the experimental constraints.

Parpillon, Benjamin↗

GLOBAL MONITORING OF URANIUM HEXIFLORIDE CYLINDERS NEXT STEPS IN DEVELOPMENT OF AN ACTION PLAN

Over 40 industrial facilities world-wide use standardized uranium hexafluoride (UF{sub 6}) cylinders for transport, storage and in-process receiving in support of uranium conversion, enrichment and fuel fabrication processes. UF{sub 6} is processed and stored in the cylinders, with over 50,000 tU of UF{sub 6} transported each year in these International Organization for Standardization (ISO) qualified containers. Although each cylinder is manufactured to an ISO standard that calls for a nameplate with the manufacturer's identification number (ID) and the owner's serial number engraved on it, these can be quite small and difficult to read. Recognizing that each facility seems to use a different ID, a cylinder can have several different numbers recorded on it by means of metal plates, sticky labels, paint or even marker pen as it travels among facilities around the world. The idea of monitoring movements of UF{sub 6} cylinders throughout the global uranium fuel cycle has become a significant issue among industrial and safeguarding stakeholders. Global monitoring would provide the locations, movements, and uses of cylinders in commercial nuclear transport around the world, improving the efficiency of industrial operations while increasing the assurance that growing nuclear commerce does not result in the loss or misuse of cylinders. It should be noted that a unique ID (UID) attached to a cylinder in a verifiable manner is necessary for safeguarding needs and ensuring positive ID, but not sufficient for an effective global monitoring system. Modern technologies for tracking and inventory control can pair the UID with sensors and secure data storage for content information and complete continuity of knowledge over the cylinder. This paper will describe how the next steps in development of an action plan for employing a global UF{sub 6} cylinder monitoring network could be cultivated using four primary UID functions - identification, tracking, controlling, and accounting.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Diffusion dynamics of the Keap1–Cullin3 interaction in single live cells

Highlights: ► We developed a quantitative FRAP-based system to study the Keap1–Cul3 interaction. ► We show that Keap1–EGFP and mCherry–Cul3 interact in single live cells. ► We used inducers which target distinct cysteine sensors of Keap1 and differ 4000-fold in potency. ► Inducers cause Nrf2 stabilization, nuclear translocation, and target gene expression. ► Inducers of four different types do not dissociate the Keap1–EGFP:mCherry–Cul3 complex. -- Abstract: Transcription factor NF-E2 p45-related factor 2 (Nrf2) regulates the expression of a network of genes encoding drug-detoxification, anti-inflammatory, and metabolic enzymes, as well as proteins involved in the regulation of cellular redox homeostasis. Under basal conditions, Kelch-like ECH associated protein 1 (Keap1) targets Nrf2 for ubiquitination and proteasomal degradation via association with Cullin3 (Cul3)-based Rbx1 E3 ubiquitin ligase. Various small molecules (inducers) activate Nrf2 leading to upregulation of cytoprotective gene expression. Inducers chemically modify specific cysteine residues of Keap1 which ultimately loses its ability to target Nrf2 for degradation. Dissociation of the Keap1–Cul3 complex by inducers is one possible mechanism, but evidence in single live cells is lacking. To investigate the diffusion dynamics of the Keap1–Cul3 interaction and the effect of inducers, we developed a quantitative fluorescence recovery after photobleaching (FRAP)-based system using Keap1–EGFP and mCherry–Cul3 fusion proteins. We show that Keap1–EGFP and mCherry–Cul3 interact in single live cells. Exposure for 1 h to small-molecule inducers of 4 different types, the oleanane triterpenoid CDDO, the isothiocyanate sulforaphane, the sulfoxythiocarbamate STCA, and the oxidant hydrogen peroxide which target distinct cysteine sensors within Keap1 with potencies which differ by nearly 4000-fold, does not dissociate the Keap1–Cul3 complex. As inducers cause conformational changes in Keap1, we conclude that changes in conformation rather than dissociation from Cul3 inactivate the repressor function of Keap1 leading to Nrf2 stabilization.

60 APPLIED LIFE SCIENCES↗

A hardware-in-the-loop (HIL) testbed for cyber-physical energy systems in smart commercial buildings

In recent years, there has been a growing trend toward the development of smart buildings that rely on cyber-physical systems (CPS) to optimize occupant comfort, safety, and energy efficiency. To ensure the reliable and efficient operation of CPS with designed control strategies, it is important to evaluate their performance under various scenarios before deploying them in the real world. This is where a Hardware-in-the-loop (HIL) testbed designed for studying sensor and control-related studies in smart buildings can be highly valuable. With the growing threat of cyber-attacks and physical faults targeting smart buildings, it is essential to ensure the security of building operations. A HIL testbed can emulate cyber-attack and physical fault scenarios, allowing researchers to develop and test threat detection and mitigation algorithms. This enables researchers to identify potential issues and optimize the algorithms in a safe and controlled environment before they are deployed in real-world settings, reducing the risk of failures that can negatively impact occupant comfort, safety, and energy efficiency. Therefore, this paper developed a HIL testbed designed for cyber-physical energy systems (e.g. buildings automation system (BAS)) in smart commercial buildings. The HIL testbed is comprised of a real-time building and Heating, Ventilation, and Air-Conditioning (HVAC) emulator using Modelica-based dynamic models, a set of BAS controllers, and a BAS computer server. The data generation capability of the HIL testbed is demonstrated by tracking normal and faulty operating data in the BAS, as well as monitoring detailed network traffic in the local BAS network. Here, this study further demonstrates the HIL testbed’s capability by conducting case studies on real-time physical fault and cyber-attack experiments using a Department of Energy (DOE) prototype commercial building. It is anticipated that the fully functional HIL testbed will be utilized for a variety of sensor and control-related studies, including but not limited to testing, developing, validating of different HVAC control strategies, fault detection & diagnosis, energy monitoring and analysis, cyber security study, etc.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

D2NO: Efficient handling of heterogeneous input function spaces with distributed deep neural operators

Neural operators have been applied in various scientific fields, such as solving parametric partial differential equations, dynamical systems with control, and inverse problems. However, challenges arise when dealing with input functions that exhibit heterogeneous properties, requiring multiple sensors to handle functions with minimal regularity. To address this issue, discretization-invariant neural operators have been used, allowing the sampling of diverse input functions with different sensor locations. However, existing frameworks still require an equal number of sensors for all functions. We propose a novel distributed approach to further relax the discretization requirements and solve the heterogeneous dataset challenges. Our method involves partitioning the input function space and processing individual input functions using independent and separate neural networks. A centralized neural network is used to handle shared information across all output functions. This distributed methodology reduces the number of gradient descent back-propagation steps, improving efficiency while maintaining accuracy. Here, we demonstrate that the corresponding neural network is a universal approximator of continuous nonlinear operators and present three numerical examples to validate its performance.

97 MATHEMATICS AND COMPUTING↗

All-polymer nanocomposites having superior strength, toughness and ultralow energy dissipation

Toughening polymers has attracted significant interest. Traditionally, polymer toughness is enhanced by constructing polymer networks or introducing sacrificial bonds into the chains between crosslink points. These strategies, though, introduce pronounced energy dissipation and associated heat, both of which are undesirable under long-term cyclic loading, for example at the interface of implants in the human body. By incorporating single-chain nanoparticles (SCNPs) into linear polymer chains to generate all-polymer nanocomposites (APNCs), we have been able to achieve high strength, high toughness with low energy dissipation. Using a combination of simulation and experimental results, we are advancing a “SCNPs effect” where tightly cross-linked SCNPs produce a modulus contrast to achieve strengthening and toughening. Benefitting from the soft interface, the penetrable and deformable SCNPs cause the surrounding polymer chains to move in concert, significantly reducing the interfacial friction to achieve low energy dissipation. The intramolecular cross-linking of the SCNPs and adhesion between the SCNPs and polymer matrix are critical for realizing such high-performance systems. Based on a Gaussian regression model and back propagation (BP) neural network, the mechanical strength can be predicted and is supported by simulations. Finally, the APNC concept described can be applied to elastomers and gels, broadening its utilization in high-cycle and low-dissipation applications, like soft robots, flexible sensors and cartilage replacements, and artificial heart valves.

36 MATERIALS SCIENCE↗

Predicting battery capacity from impedance at varying temperature and state of charge using machine learning

Prediction of battery health from electrochemical impedance spectroscopy (EIS) data can enable rapid measurement of battery state in real-world applications without using additional sensors or time-consuming performance measurements. However, deconvoluting the effect of capacity, state of charge, and temperature on EIS response is complicated analytically. Here, various machine-learning models, such as linear, Gaussian process, random forest, and artificial neural network regression, are utilized to predict capacity from EIS using hundreds of capacity, direct current (DC) resistance, and EIS measurements recorded under varying conditions of health, temperature, and state of charge (SOC). Several feature extraction and selection methods from traditional electrochemical analysis and statistical modeling are explored using machine-learning pipelines. EIS data from just two frequencies can accurately predict capacity, and interrogation shows that the optimal set of frequencies is not usually intuitive. Best results are achieved with an ensemble model, which predicts battery capacity with a mean absolute error of 1.9% on data from unobserved cells.

25 ENERGY STORAGE↗

Demonstration of Thermoacoustic Power Sensor

The goal of this project is to develop and demonstrate a thermoacoustic power sensor (TAPS) for Sodium-cooled Fast Reactor (SFR), with potential application also envisioned to other nuclear technologies such as Lead-Cooled Fast Reactor (LFR), Molten Salt Reactor (MSR), in addition to Light Water Reactors (LWRs). The project was led by Westinghouse Electric Company, LLC (Westinghouse) and carried out in collaboration with Argonne National Laboratory (ANL) and the University of Pittsburgh. A TAPS is a passive (self-powered), non-invasive (wireless) sensor envisioned for measuring key parameters, such as local temperature and neutron flux, in a nuclear reactor core. The sensor generates pressure waves (i.e., sound waves) with a frequency and amplitude dependent upon nuclear operating conditions (coolant temperature or power changes). The acoustic waves are able to travel through the core and associated structures, and reach to the detector network placed outside of the reactor vessel. These detectors require a very small amount of power which, during loss of power events, can be provided for example by harvesting gamma radiation energy, thus resulting in a monitoring system that can function both during normal operation and during loss of power events.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

SQMS Quantum R&D in Machine Learning, Optimization and Sensing beyond Fundamental Physics Applications

This newly formed team at SQMS under the Ecosystem Thrust is looking to develop capabilities impacting societal advances outside the core domain of HEP and condensed matter physics. We explicitly leverage the experimental and algorithmic innovations developed across all groups as well as connect to broad-scope external projects of the diverse team of PIs. As the inaugural set of projects, we are studying numerically quantum machine learning models inspired by efficiently trainable echo-state and orthogonal neural networks and developing designs for related experiments to be performed on quantum processors based on SQMS SRF cQED technology and Rigetti s transmon arrays. Investigated models exploit ideas and lessons learned from multiple prior work by SQMS team members in a variety of internal and external activities [R1]. Target initial applications include noisy signal processing, potentially captured by quantum sensors or noisy QPUs, as well as simulation and classification of healthcare data. For instance, image reconstruction of the brain s electrical properties by solving the inverse Maxwell equation problem with uncertainty [R2] through a hybrid quantum-classical physics-informed architecture for time-dependent processes [R3]. The group is also investigating the application and development of novel quantum sensors based on magnetic levitation of a superconducting sphere coupled to a superconducting qubit. This coupling enables high-precision measurements of the position of the sphere, which can be used for sensitive detection of forces, enabling practical applications such as gravimetry for geophysics analysis, or accelerometry for GPS-denied navigation [R4] [R1] Rieffel, Eleanor G., Ata Akbari Asanjan, M. Sohaib Alam, Namit Anand, David E. Bernal Neira, Sophie Block, Lucas T. Brady et al. "Assessing and advancing the potential of quantum computing: A NASA case study." Future Generation Computer Systems (2024). [R2] Yu, X., Serrall s, J.E., Giannakopoulos, I.I., Liu, Z., Daniel, L., Lattanzi, R. and Zhang, Z., 2023. Pifon-ept: Mr-based electrical property tomography using physics-informed fourier networks. IEEE Journal on Multiscale and Multiphysics Computational Techniques. [R3] Wudarski, Filip, Daniel OConnor, Shaun Geaney, Ata Akbari Asanjan, Max Wilson, Elena Strbac, P. Aaron Lott, and Davide Venturelli. "Hybrid quantum-classical reservoir computing for simulating chaotic systems." arXiv preprint arXiv:2311.14105 (2023). [R4] Higgins, Gerard, Saarik Kalia, and Zhen Liu. "Maglev for dark matter: Dark-photon and axion dark matter sensing with levitated superconductors." Physical Review D 109.5 (2024): 055024.

Venturelli, Davide↗

Reinforced double-threaded slide-ring networks for accelerated hydrogel discovery and 3D printing

Traditionally, slide-ring gels are stretchable but soft as a result of an elasticity-stretchability trade-off. Herein, we introduce a new approach to breaking this trade-off and creating reinforced slide-ring networks with mobile crosslinkers. Our approach involves the construction of a polyethylene glycol double-threaded γ-cyclodextrin-based pro-slide-ring crosslinker that serves as a modular component for 3D printing and copolymerization. The resulting crystalline-domain-reinforced slide-ring hydrogels, or CrysDoS-gels, exhibit both high elasticity and high stretchability. The modular synthesis allows for high-throughput synthesis of CrysDoS-gels, generating a large amount of data for structure-property analysis. Here, by employing data science techniques, such as machine learning and linear regression, not only were we able to identify which chemical components influence the mechanical properties of CrysDoS-gels, but this analysis also aided in the discovery of better-performing CrysDoS-gels. Finally, we demonstrate the potential application of the newly discovered CrysDoS-gels as sensing devices by 3D printing them as stress sensors with high sensitivity and a broad detection range.

3D-printing↗

Single Bimodular Sensor for Differentiated Detection of Multiple Oxidative Gases

Semiconductive metal-oxide sensors suffer from cross-sensitivities under mixed chemical condition, specifically upon mixture of multiple oxidative or reductive gases. Herein, a single bimodular sensor is demonstrated for smart differentiation of multiple oxidative analytes by relating the resistance-metric mode to impedance-metric mode. The sensor construct based on ZnO nanorods readily outputs three response datasets upon exposure of oxidative-gas mixture including O 2 , SO 2 , and NO 2 , the resistance, real part impedance, and imaginary part impedance. The differentiative and correlated nature between these response signals allows such a single sensor platform to differentiate these oxidative gases accurately and robustly. Linear and non-linear decision boundaries are established over a large gas-concentration range from 2 ppm to 3% through a combination of principal component analysis and artificial neural network training. A facile user interface is demonstrated for recognition and measurement of unknown gas analytes, with the error of the predicted analyte-concentration as low as 2%.

36 MATERIALS SCIENCE↗

The Dynamic Networks Experiments: Virtual Experiments to Quantify Gains in Nuclear Explosion Monitoring

We describe an ongoing series of virtual experiments conducted collaboratively by four United States National Laboratories: Sandia National Laboratories, Los Alamos National Laboratory, Lawrence Livermore National Laboratory, and Pacific Northwest National Laboratory. These Dynamic Network Experiments (DNEs) provide an experimental framework to evaluate the potential impact of new research tools on nuclear explosion monitoring. The second DNE (DNE2), completed in 2024, exploited waveform data (seismic, infrasound, and electromagnetic) that was recorded by multi-modal sensors within and near the Nevada National Security Site and synthetic radionuclide signatures over multiple time periods. During the execution of DNE2, we processed and analyzed data through a multi-stage event processing pipeline that ingested raw data, performed quality control, detected signals, built events from these signals, located these events, and characterized the events’ source types and sizes. For each stage and over the entire event processing pipeline, we evaluated performance changes by comparing the performance of new data processing methods, models, and algorithms against a baseline. We also performed an additional execution phase to assess event processing pipeline function, speed, and efficiency against that of an expert analyst, including computational and manual efforts. Finally, we assessed the impact and effort of modern computing infrastructure on the monitoring pipeline. This paper describes key elements of the DNEs, from formulation through execution, as demonstrated in DNE2. The DNEs introduce several novel concepts to quantitatively measure the potential impact of new methods on explosion monitoring, including the collaborative design of multi-modal datasets, performance and logistical metrics, and integrated analyses.

42 ENGINEERING↗

Developing a heterogeneous ensemble learning framework to evaluate Alkali-silica reaction damage in concrete using acoustic emission signals

The monitoring and evaluation of Alkali-silica reaction (ASR) damage in concrete structures are required to ensure the serviceability and integrity of concrete infrastructures such as bridges and dams. The innovation of this paper lies in the development of an automatic ASR monitoring and evaluation approach by leveraging acoustic emission (AE) and a heterogeneous ensemble learning framework. Here, in this paper, ASR was monitored by AE sensors attached to a concrete specimen, which was placed in a chamber with high humidity and temperature. The recorded AE signals were filtered and divided by four ASR phases according to signal strength, crack width and expansion strain. A heterogeneous ensemble network including convolutional neural networks (CNN) and random forest models was employed to learn different features from AE signals and classify the AE signals into their corresponding phases. The results suggest that the proposed model has a high performance and classifies the signals into the assigned phases with high accuracy.

42 ENGINEERING↗

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy's National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real-time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI's Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: Run in real-time on the edge (≥ 12 Hz), Classification: Achieve less than 5% false positive detection, Classification: Achieve ≥ 95% methane plume detection rate, Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time. In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy’s National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real- time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI’s Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: 1. Run in real-time on the edge (≥ 12 Hz) 2. Classification: Achieve less than 5% false positive detection 3. Classification: Achieve ≥ 95% methane plume detection rate 4. Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗

Assessing residential PM 2.5 concentrations and infiltration factors with high spatiotemporal resolution using crowdsourced sensors

Building conditions, outdoor climate, and human behavior influence residential concentrations of fine particulate matter (PM 2.5 ). To study PM 2.5 spatiotemporal variability in residences, we acquired paired indoor and outdoor PM 2.5 measurements at 3,977 residences across the United States totaling >10,000 monitor-years of time-resolved data (10-min resolution) from the PurpleAir network. Time-series analysis and statistical modeling apportioned residential PM 2.5 concentrations to outdoor sources (median residential contribution = 52% of total, coefficient of variation = 69%), episodic indoor emission events such as cooking (28%, CV = 210%) and persistent indoor sources (20%, CV = 112%). Residences in the temperate marine climate zone experienced higher infiltration factors, consistent with expectations for more time with open windows in milder climates. Likewise, for all climate zones, infiltration factors were highest in summer and lowest in winter, decreasing by approximately half in most climate zones. Large outdoor–indoor temperature differences were associated with lower infiltration factors, suggesting particle losses from active filtration occurred during heating and cooling. Absolute contributions from both outdoor and indoor sources increased during wildfire events. Infiltration factors decreased during periods of high outdoor PM 2.5 , such as during wildfires, reducing potential exposures from outdoor-origin particles but increasing potential exposures to indoor-origin particles. Time-of-day analysis reveals that episodic emission events are most frequent during mealtimes as well as on holidays (Thanksgiving and Christmas), indicating that cooking-related activities are a strong episodic emission source of indoor PM 2.5 in monitored residences.

54 ENVIRONMENTAL SCIENCES↗

Multiscale graph neural network autoencoders for interpretable scientific machine learning

The goal of this work is to address two limitations in autoencoder-based models: latent space interpretability and compatibility with unstructured meshes. This is accomplished here with the development of a novel graph neural network (GNN) autoencoding architecture with demonstrations on complex fluid flow applications. To address the first goal of interpretability, the GNN autoencoder achieves reduction in the number nodes in the encoding stage through an adaptive graph reduction procedure. Further, this reduction procedure essentially amounts to flowfieldconditioned node sampling and sensor identification, and produces interpretable latent graph representations tailored to the flowfield reconstruction task in the form of so-called masked fields. These masked fields allow the user to (a) visualize where in physical space a given latent graph is active, and (b) interpret the time-evolution of the latent graph connectivity in accordance with the time-evolution of unsteady flow features (e.g. recirculation zones, shear layers) in the domain. To address the goal of unstructured mesh compatibility, the autoencoding architecture utilizes a series of multi-scale message passing (MMP) layers, each of which models information exchange among node neighborhoods at various lengthscales. The MMP layer, which augments standard single-scale message passing with learnable coarsening operations, allows the decoder to more efficiently reconstruct the flowfield from the identified regions in the masked fields. Analysis of latent graphs produced by the autoencoder for various model settings are conducted using unstructured snapshot data sourced from large-eddy simulations in a backward-facing step (BFS) flow configuration with an OpenFOAM-based flow solver at high Reynolds numbers.

97 MATHEMATICS AND COMPUTING↗