Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “error performance”

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 1,783 records · Page 99
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Cross-correlation of CMB Polarization Lensing with High-$z$ Submillimeter $Herschel$-ATLAS Galaxies

Herein we report a 4.8σ measurement of the cross-correlation signal between the cosmic microwave background (CMB) lensing convergence reconstructed from measurements of the CMB polarization made by the P OLARBEAR experiment and the infrared-selected galaxies of the $Herschel$-ATLAS survey. This is the first measurement of its kind. We infer a best-fit galaxy bias of $b$ = 5.76 ± 1.25, corresponding to a host halo mass of log 10 (M h /M ⊙ ). =13.5$^{+0.2}_{-0.3}$ at an effective redshift of $z$ ~ 2 from the cross-correlation power spectrum. Residual uncertainties in the redshift distribution of the submillimeter galaxies are subdominant with respect to the statistical precision. We perform a suite of systematic tests, finding that instrumental and astrophysical contaminations are small compared to the statistical error. This cross-correlation measurement only relies on CMB polarization information that, differently from CMB temperature maps, is less contaminated by galactic and extragalactic foregrounds, providing a clearer view of the projected matter distribution. This result demonstrates the feasibility and robustness of this approach for future high-sensitivity CMB polarization experiments.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Fault Tolerant Decoding of QLDPC-GKP Codes with Circuit Level Soft Information

Concatenated bosonic-stabilizer codes have recently gained prominence as promising candidates for achieving low-overhead fault-tolerant quantum computing in the long term. In such systems, analog information obtained from the syndrome measurements of an inner bosonic code is used to inform decoding for an outer code layer consisting of a discrete-variable stabilizer code such as a surface code. The use of Quantum Low-Density Parity Check (QLDPC) codes as an outer code is of particular interest due to the significantly higher encoding rates offered by these code families, leading to a further reduction in overhead for large-scale quantum computing. Recent works have investigated the performance of QLDPC-GKP codes in detail, and the use of analog information from the inner code significantly boosts decoder performance. However, the noise models assumed in these works are typically limited to depolarizing or phenomenological noise. In this paper, we investigate the performance of QLDPC-GKP concatenated codes under circuit-level noise, based on a model introduced by Noh et al. in the context of the surface-GKP code. To demonstrate the performance boost from analog information, we investigate three scenarios: (a) decoding without soft information, (b) decoding with precomputed error probabilities but without real-time soft information, and (c) decoding with real-time soft information obtained from round-to-round decoding of the inner GKP code. Results show minimal improvement between (a) and (b), but a significant boost in (c), indicating that real-time soft information is critical for concatenated decoding under circuit-level noise. We also study the effect of measurement schedules with varying depths and show that using a schedule with minimum depth is essential for obtaining reliable soft information from the inner code.

Borah, Shantom K. [Arizona U. (main)]↗
DOE OSTI2018

Mass imbalances in EPANET water-quality simulations

EPANET is widely employed to simulate water quality in water distribution systems. However, in general, the time-driven simulation approach used to determine concentrations of water-quality constituents provides accurate results only for short water-quality time steps. Overly long time steps can yield errors in concentration estimates and can result in situations in which constituent mass is not conserved. The use of a time step that is sufficiently short to avoid these problems may not always be feasible. The absence of EPANET errors or warnings does not ensure conservation of mass. This paper provides examples illustrating mass imbalances and explains how such imbalances can occur because of fundamental limitations in the water-quality routing algorithm used in EPANET. In general, these limitations cannot be overcome by the use of improved water-quality modeling practices. This paper also presents a preliminary event-driven approach that conserves mass with a water-quality time step that is as long as the hydraulic time step. Results obtained using the current approach converge, or tend to converge, toward those obtained using the preliminary event-driven approach as the water-quality time step decreases. Improving the water-quality routing algorithm used in EPANET could eliminate mass imbalances and related errors in estimated concentrations. Furthermore, the results presented in this paper should be of value to those who perform water-quality simulations using EPANET or use the results of such simulations, including utility managers and engineers.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2015

Cyber-Informed Engineering: The Need for a New Risk Informed and Design Methodology

Current engineering and risk management methodologies do not contain the foundational assumptions required to address the intelligent adversary’s capabilities in malevolent cyber attacks. Current methodologies focus on equipment failures or human error as initiating events for a hazard, while cyber attacks use the functionality of a trusted system to perform operations outside of the intended design and without the operator’s knowledge. These threats can by-pass or manipulate traditionally engineered safety barriers and present false information, invalidating the fundamental basis of a safety analysis. Cyber threats must be fundamentally analyzed from a completely new perspective where neither equipment nor human operation can be fully trusted. A new risk analysis and design methodology needs to be developed to address this rapidly evolving threatscape.

CCE↗
DOE OSTI2026

Explainable Graph Learning for Particle Accelerator Operations

Particle accelerators are vital tools in physics, medicine, and industry, requiring precise tuning to ensure optimal beam performance. However, real-world deviations from idealized simulations make beam tuning a time-consuming and error-prone process. In this work, we propose an explanation-driven framework for providing actionable insight into beamline operations, with a focus on the injector beamline at the Continuous Electron Beam Accelerator Facility (CEBAF). We represent beamline configurations as heterogeneous graphs, where setting nodes represent elements that human operators can actively adjust during beam tuning, and reading nodes passively provide diagnostic feedback. To identify the most influential setting nodes responsible for differences between any two beamline configurations, our approach first predicts the resulting changes in reading nodes caused by variations in settings, and then learns importance scores that capture the joint influence of multiple setting nodes. Experimental results on real-world CEBAF injector data demonstrate the framework’s ability to generate interpretable insights that can assist human operators in beamline tuning and reduce operational overhead.

Wang, Song [Univ. of Virginia, Charlottesville, VA↗
DOE OSTI2026

Automated Classification of Vehicle Movements at Signalized Intersections Using Vehicle Trajectories

Accurate vehicle movement classification through signalized intersections is of paramount importance to the analysis of intersection performance and the optimization of traffic control strategies. Conventional techniques for tracking vehicle turning movements depend on infrastructure-based strategies like human counts, loop detectors, and video analytics, all of which are costly, prone to errors, and spatially constrained. High-frequency trajectory data can be utilized to determine vehicle movement patterns in a scalable and infrastructure-independent method due to the adoption of connected vehicles (CVs). In recent years, several studies have utilized connected vehicle data to generate performance measures. Most of the trajectory-based performance measures approaches, however, require map matching-i.e., extracting geospatial references from maps to identify the movements that individual vehicles make at a signalized intersection. These approaches are often time-consuming and hinder scalability since geographic features need to be provided for an analysis to be conducted. Map matching methods are prone to errors as different map versions change these geographic features. This research presents a novel automatic classification pipeline that uses CV trajectory data to classify vehicle movements at signalized crossings, specifically pass-through left-turn and right-turn maneuvers. The process starts by filtering trips that cross a spatial bounding box that has been defined at the target intersection. Approach and departure headings for each trajectory crossing the boundary are computed and are clustered together to identify dominant movements. The proposed algorithm is used to classify the movement of vehicles at 10 intersections in the state of California, and the results indicate that the algorithm can classify movements at these intersections with varying traffic volumes and road network configurations, all in a map-less framework with no need for conflation of vehicle trajectories to a digital base map.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2019

Machine-learning error models for approximate solutions to parameterized systems of nonlinear equations

Here, this work proposes a machine-learning framework for constructing statistical models of errors incurred by approximate solutions to parameterized systems of nonlinear equations. These approximate solutions may arise from early termination of an iterative method, a lower-fidelity model, or a projection-based reduced-order model, for example. The proposed statistical model comprises the sum of a deterministic regression-function model and a stochastic noise model. The method constructs the regression-function model by applying regression techniques from machine learning (e.g., support vector regression, artificial neural networks) to map features (i.e., error indicators such as sampled elements of the residual) to a prediction of the approximate-solution error. The method constructs the noise model as a mean-zero Gaussian random variable whose variance is computed as the sample variance of the approximate-solution error on a test set; this variance can be interpreted as the epistemic uncertainty introduced by the approximate solution. This work considers a wide range of feature-engineering methods, data-set-construction techniques, and regression techniques that aim to ensure that (1) the features are cheaply computable, (2) the noise model exhibits low variance (i.e., low epistemic uncertainty introduced), and (3) the regression model generalizes to independent test data. Finally, numerical experiments performed on several computational-mechanics problems and types of approximate solutions demonstrate the ability of the method to generate statistical models of the error that satisfy these criteria and significantly outperform more commonly adopted approaches for error modeling.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Human performance benefits gained by dynamic instructions compared to smart PDFs

Human performance benefits gained by dynamic instructions compared to smart PDFs In recent years multiple nuclear utilities have deployed an electronic work management solution to replace the traditional paper-based processes. There are many benefits of electronic work management solutions compared to paper processes, such as a more streamlined review and approval process of work packages and a better way to track and resolve any potential bottlenecks causing delays in the review process. These benefits have been recognized where electronic work management solutions have been successfully deployed. However the utilities also hoped the new solution would help the field worker conduct his/her job more efficiently, correctly, and safely. Unfortunately, the industry is now realizing that the deployed solutions did not fully help them reach that goal. The electronic work management solutions present the work instructions as an electronic copy (i.e., a portable document format (PDF)) of the paper instruction. The PDF versions used in these electronic work management solutions have some ability for data input, notes, and mark-ups, but are unable to provide many additional benefits compared to using paper. In fact, conducting work using a PDF on a tablet can be more cumbersome than with a paper instruction. For example, flipping back and forth in a paper instruction can be time consuming, but it is not at all as frustrating as scrolling back and forth on a tablet or continuously having to navigate between tabs. The researchers at Idaho National Laboratory (INL) approached the next step beyond smart PDFs from a human factors perspective. In close collaboration with multiple utilities they investigated how to present instruction content and how to design the interaction with the instruction to improve the workers performance and efficiency. Their solution is a dynamic presentation of the instruction that guides the worker through the correct path as well as ensuring the worker is on the correct component. The solution also automatically validates input and provides guidance if unexpected conditions are identified. A well-designed solution has the opportunity to vastly reduce the risk of human errors in the field, such as preventing actions on incorrect component or unintentionally skipping steps. The human factors researchers at INL leveraged their expertise, existing human performance tools, and a great amount of involvement from field workers in multiple utilities to develop and evaluate design concepts that will provide the human performance and efficiency benefits that the electronic work management solutions fall short to deliver. This paper will describe in detail how these new design concepts improve human performance.

47 OTHER INSTRUMENTATION↗
DOE OSTI2016

Implementation and Comparison of Different Algorithms for on the fly Doppler Broadening in RMC Code

Accurate cross sections are of great importance to Monte Carlo particle transport calculation codes which use pointwise cross sections. It is well known that cross sections have much dependence on temperature on account of Doppler effect which is particularly vital for resonance absorption cross sections. In general, pre-prepared continuous-energy libraries are used in Monte Carlo codes, which is feasible in many applications. However, in multi-physics calculations where neutronics and thermal hydraulics are coupled, cross sections at a fine interval of temperatures are desired, which brings a great challenge to computer memory. Consequently, different methods about on the fly Doppler broadening for cross sections are proposed, which can generate cross sections at the specific temperature points during neutron simulations. These methods can be roughly classified into two categories: one is to calculate the required cross sections by different algorithms, such as interpolation using a high order function expansion, multipole representation, and different integration algorithms based on Doppler broadening equation. The other, for instance, target motion sampling (TMS), does not calculate the Doppler broadening cross sections but takes into account Doppler effect by rejection technique. In this paper, several on the fly Doppler broadening algorithms based on Doppler broadening equation are implemented in the Reactor Monte Carlo code RMC, and then the accuracy and efficiency are discussed and compared. In this paper, several algorithms for on the fly Doppler broadening based on Doppler broadening equation have been implemented in RMC and the focus falls on the comparison of the Gauss Legendre quadrature and Gauss Hermite quadrature. Gauss Legendre quadrature is more accurate than Gauss Hermite quadrature but also more time consuming. Base temperature has a significant impact on Gauss Hermite quadrature while increasing base temperature to room temperature (300 K) can give satisfactory results. In addition, the proportion of overhead of Doppler broadening treatment in total computational time tends to decrease with an increase number of nuclides in fuel. For depletion calculation, both algorithms are in good agreement with reference and the relative errors of K{sub inf} are within 5 times of standard deviation. As for total computational time, the ratio between Gauss Legendre quadrature or Gauss Hermite quadrature and reference decreases and reaches 9.6 or 3.4 at last, respectively. Four important isotopes are chosen to compare the atom density and all relative errors are within 0.25%. After comparisons of different algorithms and different base temperatures, it can be found that Gauss Hermite quadrature broadening from 300 K can perform on the fly cross section Doppler broadening with least total computational time and acceptable precision. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

The many-body expansion for aqueous systems revisited: 1. Water-water interactions

We revisit the Many-Body Expansion (MBE) for water-water interactions by examining the effects of the basis set, including those resulting from the Basis Set Superposition Error (BSSE) correction, on the various terms for selected sizes of water clusters up to n = 21. The analysis is performed at the second order Møller-Plesset (MP2) perturbation theory with the family of augmented correlation consistent basis sets up to five zeta quality for the (H2O)n, n = 7, 10, 13, 16 and 21 clusters, for which we report either the complete MBE (n = 7, 10) or the one through the 6-body (n = 13) and the 5-body terms (n = 16, 21). Our results suggest that any sizeable contributions to the total cluster binding energy arising from the 5-body and larger terms are solely an artifact of the size of the basis set. Indeed, all terms above the 4-body converge monotonically to practically zero at the Complete Basis Set (CBS) limit. The magnitudes of the 5-body and above terms are accurately reproduced even with the smaller basis set of the series (aug-cc-pVDZ) once the BSSE correction is taken into account. The same level of theory (MP2/aug-cc-pVDZ, BSSEcorrected) also accurately reproduces the MP2/CBS values of the 3- and 4-body terms. The contribution of electron correlation to the 3- and 4-body terms is quite small, so that neglecting the correlation contribution in all terms above the 3-body results in an error of ??(0.1%). The BSSE correction to the largest 2-body term in the MBE was accurately estimated from the function ?? [1 + erf(-?? ??)], which is proportional to the common (overlapping) area between two Gaussian distributions whose centers are separated by R, with the constants ??, ?? fitted to the calculated BSSE corrections for the individual 2-body terms of the clusters with each basis set. Our results unquestionably demonstrate that the MBE for water-water interactions converges to the 4-body term, since any finite terms above the 4-body are artifacts of the size of the basis set and can be safely truncated at the 4-body term when either a very large basis set is used or BSSE corrections are taken into account even with the aug-cc-pVDZ basis set. We expect these findings to have important consequences in the pursuit of accurate many-body molecular dynamics simulations for aqueous systems.

Heindel, Joseph↗
DOE OSTI2019

The impact of atmospheric dispersion in the performance of high-resolution spectrographs

ABSTRACT Differential atmospheric dispersion is a wavelength-dependent effect introduced by the atmosphere. It is one of the instrumental errors that can affect the position of the target as perceived on the sky and its flux distribution. This effect will affect the results of astronomical observations if not corrected by an atmospheric dispersion corrector (ADC). In high-resolution spectrographs, in order to reach a radial velocity (RV) precision of 10 cm s−1, an ADC is expected to return residuals at only a few tens of milliarcseconds (mas). In fact, current state-of-the-art spectrograph conservatively require this level of residuals, although no work has been done to quantify the impact of atmospheric dispersion. In this work, we test the effect of atmospheric dispersion on astronomical observations in general, and in particular on RV precision degradation and flux losses. Our scientific objective was to quantify the amount of residuals needed to fulfil the requirements set on an ADC during the design phase. We found that up to a dispersion of 100 mas, the effect on the RV is negligible. However, on the flux losses, such a dispersion can create a loss of ∼2 per cent at 380 nm, a significant value when efficiency is critical. The requirements set on ADC residuals should take into consideration the atmospheric conditions where the ADC will function, and also all the aspects related with not only the RV precision requirements but also the guiding camera used, the tolerances on the flux loss, and the different melt data of the chosen glasses.

Wehbe, B.↗
DOE OSTI2019

Physics-Informed Machine Learning with Application to Solid Oxide Fuel Cell System Modeling and Optimization

Traditional data-driven machine learning (ML) techniques were combined with the physics-based SOFC-MP model toward improving SOFC system-level performance prediction. Four different physics-informed ML methods were proposed to demonstrate better prediction accuracy compared with traditional ML methods. First, it was preferred that the existing Kriging-based surrogate model not be entirely abandoned, so errors between Kriging regression prediction and true solution are learned using ML. The goal was to find the pattern of the error distribution as a function of model input so discrepancies between the Kriging prediction and true solution could be reduced by adding the additional ML-trained error term to the existing prediction. Second, deep neural networks (DNN) coupled with the mass balance model (MBM) have been developed to significantly decrease the reduced order model (ROM) prediction error with fewer training data compared to the traditional Kriging-based ROM and traditional DNN regression approaches. Third, DNN regression coupled with a neural networks (NN) classifier and MBM was developed. It provides superior performance on classifying the physically operational conditions for natural gas fuel cell (NGFC) than traditional classification approaches. Finally, two kinds of DNN transferring were studied and tested. One was transferring DNN for low-fidelity model data to a high-fidelity model. The second one was transferring DNN from state-of-the-art NGFC to advanced NGFC. These physic-informed ML frameworks can serve as a solid foundation for estimating U.S. Department of Energy-targeted fuel cell performance and exhibit great potential to be employed to other engineering applications.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2016

Online 3D EPID-based dose verification: Proof of concept

Purpose: Delivery errors during radiotherapy may lead to medical harm and reduced life expectancy for patients. Such serious incidents can be avoided by performing dose verification online, i.e., while the patient is being irradiated, creating the possibility of halting the linac in case of a large overdosage or underdosage. The offline EPID-based 3D in vivo dosimetry system clinically employed at our institute is in principle suited for online treatment verification, provided the system is able to complete 3D dose reconstruction and verification within 420 ms, the present acquisition time of a single EPID frame. It is the aim of this study to show that our EPID-based dosimetry system can be made fast enough to achieve online 3D in vivo dose verification. Methods: The current dose verification system was sped up in two ways. First, a new software package was developed to perform all computations that are not dependent on portal image acquisition separately, thus removing the need for doing these calculations in real time. Second, the 3D dose reconstruction algorithm was sped up via a new, multithreaded implementation. Dose verification was implemented by comparing planned with reconstructed 3D dose distributions delivered to two regions in a patient: the target volume and the nontarget volume receiving at least 10 cGy. In both volumes, the mean dose is compared, while in the nontarget volume, the near-maximum dose (D2) is compared as well. The real-time dosimetry system was tested by irradiating an anthropomorphic phantom with three VMAT plans: a 6 MV head-and-neck treatment plan, a 10 MV rectum treatment plan, and a 10 MV prostate treatment plan. In all plans, two types of serious delivery errors were introduced. The functionality of automatically halting the linac was also implemented and tested. Results: The precomputation time per treatment was ∼180 s/treatment arc, depending on gantry angle resolution. The complete processing of a single portal frame, including dose verification, took 266 ± 11 ms on a dual octocore Intel Xeon E5-2630 CPU running at 2.40 GHz. The introduced delivery errors were detected after 5–10 s irradiation time. Conclusions: A prototype online 3D dose verification tool using portal imaging has been developed and successfully tested for two different kinds of gross delivery errors. Thus, online 3D dose verification has been technologically achieved.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2023

Single-Point Extrapolation to the Complete Basis Set Limit through Deep Learning

Machine learning (ML) offers an attractive method for making predictions about molecular systems while circumventing the need to run expensive electronic structure calculations. Once trained on ab initio data, the promise of ML is to deliver accurate predictions of molecular properties that were previously computationally infeasible. In this work, we develop and train a graph neural network model to correct the basis set incompleteness error (BSIE) between a small and large basis set at the RHF and B3LYP levels of theory. Our results show that, when compared to fitting to the total potential, an ML model fitted to correct the BSIE is better at generalizing to systems not seen during training. We test this ability by training on single molecules while evaluating on molecular complexes. We also show that ensemble models yield better behaved potentials in situations where the training data is insufficient. However, even when only fitting to the BSIE, acceptable performance is only achieved when the training data sufficiently resemble the systems one wants to make predictions on. As a result, the test error of the final model trained to predict the difference between the cc-pVDZ and cc-pV5Z potential is 0.184 kcal/mol for the B3LYP density functional, and the ensemble model accurately reproduces the large basis set interaction energy curves on the S66x8 dataset.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Sensor Incipient Fault Impacts on Building Energy Performance: A Case Study on a Multi-Zone Commercial Building

Existing studies show sensor faults/error could double building energy consumption and carbon emissions compared with the baseline. Those studies assume that the sensor error is fixed or constant. However, sensor faults are incipient in real conditions and there were extremely limited studies investigating the incipient sensor fault impacts systematically. This study filled in this research gap by studying time-developing sensor fault impacts to rule-based controls on a 10-zone office building. The control sequences for variable air volume boxes (VAV) with an air handling unit (AHU) system were selected based on ASHRAE Guideline 36-2018: High-Performance Sequences of Operation for HVAC Systems. Large-scale simulations on cloud were conducted (3600 cases) through stochastic approach. Results show (1) The site energy differences could go –3.3% lower or 18.1% higher, compared with baseline. (2) The heating energy differences could go –66.5% lower or 314.4% higher, compared with baseline. (3) The cooling energy differences could go –11.5% lower or 65.0% higher, compared with baseline. (4) The fan energy differences could go 0.15% lower or 6.9% higher, compared with baseline.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

Large deformation and brittle failure calculated using the dual-domain material point method

The dual domain material point (DDMP) method is explored as a candidate to be implemented in a general purpose code to perform simulations of materials with complex geometry that undergo large history-dependent deformation and failure. To test its candidacy, we study its mesh convergence, its sensitivity to mesh orientation, and its ability to handle softening and failure of a material. Simulations of large deformation and simulations of mechanical failure are performed using both DDMP and the material point method (MPM). When cell-crossing of material points is not an issue and when there are a sufficient number of material points in each computation cell, the numerical error decreases with the square of the cell size as expected for both MPM and DDMP. DDMP has reduced error compared with MPM when there are many instances of material points crossing cell boundaries due to the continuous nature of the modified gradient of the shape functions. Simulations of a specimen under tension are also performed where the background mesh is aligned and misaligned with the tension direction. MPM displays a significant mesh-dependent stress field, DDMP shows negligible mesh dependency. Despite a mesh orientation-dependent stress field from MPM, the critical tension and failure mode from both MPM and DDMP calculations have negligible mesh dependency when using a non-local failure model. If only the failure mode is important (i.e., local stresses are unimportant), MPM with a non-local failure model is a suitable method for modeling failure with small deformations. However, if local stresses are also important or if there are large deformations with many cell-crossings before failure, DDMP should be the method that is used. A needed improvement for DDMP is identified from our numerical simulations.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Di-CNN: Domain-Knowledge-Informed Convolutional Neural Network for Manufacturing Quality Prediction

In manufacturing, convolutional neural networks (CNNs) are widely used on image sensor data for data-driven process monitoring and quality prediction. However, as purely data-driven models, CNNs do not integrate physical measures or practical considerations into the model structure or training procedure. Consequently, CNNs’ prediction accuracy can be limited, and model outputs may be hard to interpret practically. This study aims to leverage manufacturing domain knowledge to improve the accuracy and interpretability of CNNs in quality prediction. A novel CNN model, named Di-CNN, was developed that learns from both design-stage information (such as working condition and operational mode) and real-time sensor data, and adaptively weighs these data sources during model training. It exploits domain knowledge to guide model training, thus improving prediction accuracy and model interpretability. A case study on resistance spot welding, a popular lightweight metal-joining process for automotive manufacturing, compared the performance of (1) a Di-CNN with adaptive weights (the proposed model), (2) a Di-CNN without adaptive weights, and (3) a conventional CNN. The quality prediction results were measured with the mean squared error (MSE) over sixfold cross-validation. Model (1) achieved a mean MSE of 6.8866 and a median MSE of 6.1916, Model (2) achieved 13.6171 and 13.1343, and Model (3) achieved 27.2935 and 25.6117, demonstrating the superior performance of the proposed model.

47 OTHER INSTRUMENTATION↗
DOE OSTI2019

Quantitative capabilities of STXM to measure spatially resolved organic volume fractions of mixed organic/inorganic particles

Scanning transmission X-ray microscopy coupled with near-edge X-ray absorption and fine structure (STXM-NEXAFS) spectroscopy can be used to characterize the morphology and composition of aerosol particles. Here, two inorganic/organic systems are used to validate the calculation of organic volume fraction (OVF) and determine the level of associated error by using carbon K-edge STXM data at 278, 285.4, 288.6, and 320 eV. Using the mixture of sodium chloride and sucrose as one system and ammonium sulfate and sucrose as another, three solutions were made with 10:1, 1:1, and 1:10 mass ratios (inorganic to organic). The OVFs of the organic-rich aerosols of both systems deviated from the bulk OVF by less than 1%, while the inorganic-rich aerosols deviated by approximately 1 %. Aerosols from the equal mass mixture deviated more (about 4 %) due to thick inorganic regions exceeding the linear range of Beer's law. These calculations were performed after checking the data for poor image alignment, defocusing issues, and particles too thick to be analyzed. The potential for systematic error in the OVF calculation was also tested by assuming the incorrect composition. There is a small (about 0.5 %) OVF difference if the organic is erroneously assumed to be adipic acid rather than the known organic, sucrose. A much larger difference (up to 25 %) is seen if sodium chloride is assumed instead of ammonium sulfate. These results show that the OVF calculations are fairly insensitive to the organic while being much more sensitive to the choice of inorganic.

47 OTHER INSTRUMENTATION↗