Search NASA⌕ Search

SEARCH · Search NASA

Results for “random testing”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Random forest models accurately classify synthetic opioids using high-dimensionality mass spectrometry datasets

Detection of novel threat agents presents several challenges, a principle one being the development of untargeted methods to screen an increasing number of threat chemicals whose exact structures are unknown. With the use of Machine Learning (ML) tools, we can guide the development of analytical methods for broad-spectrum detection of unbounded threat chemical families in complex mixtures. Toward this goal, we used nominal mass and high-resolution mass spectrometry data for hundreds of synthetic opioids and non-opioid compounds. We tested two ML techniques, logistic regression and random forest, to develop models towards a practical, implementable method for opioid detection. We found that of these tested ML methods, random forest models resulted in the highest validation accuracy (95+%) for both nominal mass and high-resolution classification of opioids versus non-opioids, with low false positive and false negative rates. The RF models were then used to successfully predict the classification of 10 compounds—five opioids and five non-opioids not part of the training and validation analysis. This application of ML is a critical step towards the development of field-deployable nominal mass spectrometers with ML-driven analyses for classification of emergent threats.

Arasteh, Kourosh [Lawrence Livermore National Labo↗
DOE OSTI2024

Macroscopic Traffic Modeling Using Probe Vehicle Data: A Machine Learning Approach

Abstract The macroscopic fundamental diagram (MFD) captures an orderly relationship among traffic flow, density, and speed at the network level. It is a simple yet powerful tool for modeling traffic dynamics in large urban networks with broad application in traffic control and management. However, empirically derived MFDs in urban regions require high-resolution traffic data from the network. Having the network flow and vehicular density estimated at the (granular) census tract level using vehicle probe data, we apply machine learning methods to predict the MFDs across U.S. urban areas and capture the impacts of location-specific input features on the network flow–density relationships at a large scale. The results show that, among the four tested machine learning approaches (Random Forest, XGBoost, Support Vector Machine, and Neural Network), XGBoost delivers the best performance in predicting network traffic flow based on vehicular density and location attributes. Using interaction Shapley Additive explanation (SHAP) values and partial correlation analysis, we examine the factors influencing MFD shapes across different locations. Our empirical findings reveal that across U.S. urban areas, network topology, transportation infrastructure, and land use are primary factors shaping MFD curves, while demand and trip-related factors play a lesser role. Specifically, higher ranking roads, centrality, and development levels correlate positively with network capacity and critical density, whereas negative associations are observed for network connectivity, mixed-use development, and road roughness levels.

Jin, Ling↗
DOE OSTI2025

DROP DURABILITY ASSESSMENT OF ELECTRONIC ASSEMBLIES UNDER OFF-AXIS LOADING WITH SKEWED FIXTURES

This thesis studies drop durability of electronic assemblies when the acceleration vector is oriented at 45° to the out-of-plane direction of the circuit card. The off-axis drop tests are accomplished with a skewed fixture and are conducted as a proxy for multiaxial drop testing. Advanced shock testing and vibration test methods have been developed over the last few decades to better represent real-world field environments during ground-based laboratory testing. However, many of these test methods require expensive and specialized equipment not available in most laboratories. An alternative approach for approximating simultaneous loading along multiple axes on conventional equipment utilizes skewed fixtures which have seen use in off-axis random vibration and drop impact testing. These methods generally rely on the conversion of a uniaxial input load from the test equipment (using a uniaxial drop tower or shaker) into a multiaxial load when resolved in the reference frame of the test article (mounted on a skewed fixture). Skewed fixture design is presented and recommendations for conducting skewed angle drop testing are introduced based on local measurements along the skewed face of the fixture to accurately monitor the impact event. Characterization tests were performed with a skewed fixture, at simultaneous acceleration loads from 500 to 3,000 g in two (in-plane and out-of-plane) directions, while meeting standard time domain tolerances. Upon experimental characterization, drop shock durability tests were conducted on a printed circuit assembly (PCA). Mean drops-to-failure were measured and quantified with Weibull statistics. Dominant solder joint failure modes were identified via failure analysis. Prior work on inclined angle impact testing is limited, and the majority of solder joint interconnect level fatigue studies are conducted considering perpendicular loading normal the circuit card. Low-cycle fatigue curves are generated based on plastic strain and plastic work density within the solder joint. A multiscale nonlinear finite element model is used to relate board-level flexure to solder joint interconnect level plastic strain. A high strain rate solder constitutive model allows for accurate modeling of solder plasticity resulting from high-impact drop shock. Fatigue parameters are computed from the Coffin-Manson relation and Palmgren-Miner damage accumulation. This work serves to apply established low-cycle fatigue methods for conventional drop shock loading (impact normal to circuit card) to non-perpendicular loading with a skewed fixture.

Hower, Jonathan [Kansas City National Security Cam↗
DOE OSTI2026

Active Learning‐Driven Inkless Additive Nanomanufacturing for Printed Electronics

Inkless additive nanomanufacturing for printed electronics promises broad material and substrate versatility, yet the high-dimensional print parameter space makes tuning print parameters time-intensive. We present a Bayesian optimization study that constructs a digital twin from printed-silver data to benchmark surrogate models, acquisition functions, and batch sizes head-to-head to achieve user-specified target resistance. Tested surrogate models included Gaussian process, random forest, and Bayesian neural network surrogates with expected improvement and confidence bound acquisition functions. In total, we evaluate 48 unique model configurations alongside a random sampling baseline for comparison. For printed silver, the Bayesian neural network with a batch size of one achieved the lowest average cumulative regret, approximately four times more efficient on average than random sampling. To balance performance and substrate space, a random forest model with expected improvement and a batch size of four was chosen as the model for validation testing. Applying this chosen configuration to copper with an additional print parameter, the model achieved a resistance within 0.15 Ω of a 1 Ω target in fewer than 30 printed lines across five validation sets. Altogether, the workflow yields a tuned and validated model that efficiently guides experiments toward the target while simultaneously learning the parameter space.

Bevel, Colton [Auburn University, AL (United State↗
DOE OSTI2025

Performance and automatic calibration scheme of the waveform sampler in the ETROC2 ASIC chip

The waveform sampler in the CMS ETROC2 chip for LGAD gain aging monitoring is a 2.56-GS/s 12-bit 8x-Interleaved ADC that consists of a coarse SAR stage, and a fine stage. This architecture delivers high performance on a relatively modest 65 nm process, while requires finding up to 24 calibration constants through calibration. We developed an automatic calibration method using charge injection test data. After calibration, the baseline random error is reduced by a factor of 2.5–3 compared to the default calibration, and a 5% charge measurement precision is achieved in 15 fC charge injection tests.

Fu, Tao [Unlisted, US]↗
DOE OSTI2026

Validation and Verification of Python based Neutron Spectrum Unfolding Software

To validate and verify the python-based code (PySL), designed to replicate the programs used by STAYSL for Beam Correction Factor (BCF) and Self-Shielding Factor (SHIELD), a series of tests were performed. To test BCF a python script was written to generate a random flux history file and both versions of the code processed the data. The test verified matching values up to at least one decimal place, approximately 10,000 tests where run and each one passed. Isotopes began to fail the tests once neutron saturation was reached. To verify this the total time of exposure was varied the isotopes that failed were compared to a list of their half-lives. The test process for SHIELD was very similar but, in this case, the code began by producing an input file with varying thickness and device type/environment for the SHIELD input. The failure condition for this test was if any of the data points for an isotope had a difference above 3%. Approximately 40 of these tests were run and there were only 3 isotopes that had reoccurring failures but only 2% of their points were above the 3% difference. A visual comparison was conducted by plotting the results from both programs. Although the test failed, the differences between their values were minuscule, and the self-shielding factor’s shape was preserved when plotted. Next steps for this project will be validating and verifying the python-based SigPhi code and then reproducing and testing the least squares unfolding performed by STAYSL.

73 - NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Constraining Galaxy-Halo connection using machine learning

We investigate the potential of machine learning (ML) methods to model small-scale galaxy clustering for constraining Halo Occupation Distribution (HOD) parameters. Our analysis reveals that while many ML algorithms report good statistical fits, they often yield likelihood contours that are significantly biased in both mean values and variances relative to the true model parameters. This highlights the importance of careful data processing and algorithm selection in ML applications for galaxy clustering, as even seemingly robust methods can lead to biased results if not applied correctly. ML tools offer a promising approach to exploring the HOD parameter space with significantly reduced computational costs compared to traditional brute-force methods if their robustness is established. Using our ANN-based pipeline, we successfully recreate some standard results from recent literature. Properly restricting the HOD parameter space, transforming the training data, and carefully selecting ML algorithms are essential for achieving unbiased and robust predictions. Among the methods tested, artificial neural networks (ANNs) outperform random forests (RF) and ridge regression in predicting clustering statistics, when the HOD prior space is appropriately restricted. We demonstrate these findings using the projected two-point correlation function (w p (r p )), angular multipoles of the correlation function (ξ ℓ (r)), and the void probability function (VPF) of Luminous Red Galaxies from Dark Energy Spectroscopic Instrument mocks. Our results show that while combining w p (r p ) and VPF improves parameter constraints, adding the multipoles ξ 0 , ξ 2 , and ξ 4 to w p (r p ) does not significantly improve the constraints.

cosmology↗
DOE OSTI2025

Monte Carlo Dropout Uncertainty Quantification of Long Short-Term Memory Autoencoder Anomaly Detection in a Liquid Sodium Cold Trap

Advanced high-temperature fluid reactors, such as sodium-cooled fast reactors (SFRs) and molten salt–cooled reactors (MSCRs), require coolant purification systems to prevent fluid contamination and local freezing that can lead to plugging. Liquid sodium purification can be achieved with a cold trap, where the sodium temperature is reduced to a near-freezing point to precipitate out impurities. Automation of monitoring of the cold trap performance with machine learning algorithms can aid in early detection of incipient anomalies. An efficient approach to loss-of-coolant–type anomaly detection in a cold trap monitored with more than two dozen thermal-hydraulic sensors consists of a long short-term memory (LSTM) autoencoder. This work develops the uncertainty quantification of the LSTM autoencoder performance for cold trap anomaly detection using the Monte Carlo (MC) dropout method. The MC dropout methodology creates a distribution of sister distributions that all slightly differ from each other because of random neurons being turned off for testing. The variances of the sister network distributions are used to make an uncertainty interval. Our analysis shows that the uncertainty in the autoencoder performance is largest near the peak of the anomaly signal. Using the MC dropout method, we investigate the uncertainty in the anomaly detection with missing sensor inputs. This capability allows the reactor operator to evaluate resilience of the anomaly detection system and to make informed decisions about continuity of operation in the event of sensor failure.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Green solvent-processable poly(3-hexylthiophene) derivative maintains high hole mobility in diodes

The development of environmentally sustainable fabrication methods for organic solar cells (OSCs) is critical to enable their large-scale adoption. Conventional solution-processed OSCs often rely on halogenated, toxic solvents that pose health and environmental risks, limiting their scalability. This is because π-conjugated polymers tend to have a low solubility in non-halogenated solvents. A common strategy to enhance solubility in alternative solvents is through the incorporation of polar solubilizing groups, often on every repeat unit of at least one monomer. However, too many polar side chains tend to disrupt favorable morphology and decrease charge transport properties. In this study, we demonstrate that a minimal degree of side chain functionalization can improve green solvent processability while preserving electrical performance. We tested this hypothesis using our previously reported random copolymer derivative of poly(3-hexylthiophene) (P3HT) where ∼10 mol% of the side chains are 6-pentanoatehexyl side chains. We find that this low-level functionalization significantly broadens the solvents that can be used to process P3HT to less toxic solvents such as o-xylene and anisole. Furthermore, the copolymer processed from greener solvents showed high hole mobilities (∼10⁻ 3 cm 2 /Vs by the Space-Charge Limiting Current method in diodes), comparable to P3HT films cast from toxic chloroform. These findings suggest that minimal side-chain modification is a viable strategy for expanding green solvent compatibility while preserving electrical performance, paving the way toward more sustainable organic semiconductor designs.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Machine learning approaches for influenza A virus risk assessment identifies predictive correlates using ferret model in vivo data

In vivo assessments of influenza A virus (IAV) pathogenicity and transmissibility in ferrets represent a crucial component of many pandemic risk assessment rubrics, but few systematic efforts to identify which data from in vivo experimentation are most useful for predicting pathogenesis and transmission outcomes have been conducted. To this aim, we aggregated viral and molecular data from 125 contemporary IAV (H1, H2, H3, H5, H7, and H9 subtypes) evaluated in ferrets under a consistent protocol. Three overarching predictive classification outcomes (lethality, morbidity, transmissibility) were constructed using machine learning (ML) techniques, employing datasets emphasizing virological and clinical parameters from inoculated ferrets, limited to viral sequence-based information, or combining both data types. Among 11 different ML algorithms tested and assessed, gradient boosting machines and random forest algorithms yielded the highest performance, with models for lethality and transmission consistently better performing than models predicting morbidity. Comparisons of feature selection among models was performed, and highest performing models were validated with results from external risk assessment studies. Our findings show that ML algorithms can be used to summarize complex in vivo experimental work into succinct summaries that inform and enhance risk assessment criteria for pandemic preparedness that take in vivo data into account.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Direct pulse-level compilation of arbitrary quantum logic gates on superconducting qutrits

Advanced simulations and calculations on quantum computers require high-fidelity implementations of quantum operations. The universal gateset approach builds complex unitaries from a small set of primitive gates, often resulting in a long gate sequence, which is typically a leading factor in the total accumulated error. Compiling a complex unitary for processors with higher-dimensional logical elements, such as qutrits, exacerbates the accumulated error per unitary, since an even longer gate sequence is required. Optimal control methods promise time- and resource-efficient compact gate sequences and, therefore, higher fidelity. These methods generate pulses that can directly implement any complex unitary on a quantum device. In this work, we demonstrate that any arbitrary qubit and qutrit gate can be realized with high fidelity, which can significantly reduce the length of a gate sequence. We generate and test pulses for a large set of randomly selected arbitrary unitaries on several quantum processing units (QPUs): the Lawrence Livermore National Laboratory Quantum Device and Integration Testbed’s (QuDIT’s) standard QPU and three of Rigetti’s QPUs: Ankaa-2, Ankaa-9Q-1, and Aspen-M-3. On the QuDIT platform’s standard QPU, the average fidelity of random qutrit gates is 97.9 ± 0.5% measured with conventional QPT and 98.8 ± 0.6% from QPT with gate folding. Rigetti’s Ankaa-2 achieves random qubit gates with an average fidelity of 98.4 ± 0.5% (conventional QPT) and 99.7 ± 0.1% (QPT with gate folding). On Ankaa-9Q-1 and Aspen-M-3, the average fidelities with conventional qubit QPT measurements were higher than 99% (see Appendix). Here we show that optimal control gates are robust to drift for at least 3 h and that the same calibration parameters can be used for all implemented gates. Our work promises that the calibration overheads for optimal control gates can be made small enough to enable efficient quantum circuits based on this technique.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2024

The Evaluation of Machine Learning Techniques for Isotope Identification Contextualized by Training and Testing Spectral Similarity

Precise gamma-ray spectral analysis is crucial in high-stakes applications, such as nuclear security. Research efforts toward implementing machine learning (ML) approaches for accurate analysis are limited by the resemblance of the training data to the testing scenarios. The underlying spectral shape of synthetic data may not perfectly reflect measured configurations, and measurement campaigns may be limited by resource constraints. Consequently, ML algorithms for isotope identification must maintain accurate classification performance under domain shifts between the training and testing data. To this end, four different classifiers (Ridge, Random Forest, Extreme Gradient Boosting, and Multilayer Perceptron) were trained on the same dataset and evaluated on twelve other datasets with varying standoff distances, shielding, and background configurations. A tailored statistical approach was introduced to quantify the similarity between the training and testing configurations, which was then related to the predictive performance. Wilcoxon signed-rank tests revealed that the OVR-wrapped XGB significantly outperformed the other algorithms, with confidence levels of 99.0% or above for the 133Ba, 60Co, 137Cs, and 152Eu sources. The findings from this work are significant as they outline techniques to promote the development of robust ML-based approaches for isotope identification.

domain adaptation↗
DOE OSTI2024

Evaluation of the Planetary Boundary Layer Height From ERA5 Reanalysis With MOSAiC Observations Over the Arctic Ocean

The planetary boundary layer height (PBLH) is a crucial indicator reflecting the region of the atmosphere characterized by continuous turbulence. Here, we use radiosonde and surface meteorological observations (4–7 times per day, year-round measurements) during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition to derive the PBLH (PBLH MOSAiC ), and further evaluate the PBLH from the ERA5 reanalysis (PBLH ERA5 ). Comparisons between PBLH MOSAiC and PBLH ERA5 from different perspectives reveal that: (a) The overestimation of PBLH ERA5 when the sea ice concentration is >90% is significant with the centered root mean squared error reaching up to 201 m; (b) The difference between the two products is notably pronounced in cold seasons, while it is comparatively diminished in warm seasons; (c) In neutral boundary layers, differences in PBLH ERA5 are larger compared with stable and convective boundary layers. In addition, the analysis of error sources indicates that the bias of PBLH ERA5 is sensitive to the bias of vertical thermal structure and wind speed profiles in ERA5 data sets in all conditions. Finally, we find a Random Forest model effectively reduces the bias of PBLH ERA5 with the index of agreement reaching up to 0.71 in the test data set, while a multiple linear regression demonstrates comparable performance to the Random Forest model.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

An improved dataset for predicting mammal infecting viruses from genetic sequence information

There have been several attempts to develop machine learning (ML) models to identify human infecting viruses from their genomic sequences, with varying degrees of success. Direct comparison between models is problematic, because these models are typically trained and evaluated on different datasets with alternative data splitting schemes, features, and model performance metrics. In this paper we present a standardized dataset of mammal infecting and non-infecting viral pathogens, refined from the previous work of Mollentze et al. to include the latest literature evidence, roughly doubling the number of curated host-virus records available to the community, and new host target labels, primate and mammal. The new host labels were included for several reasons, including previous reports that classification performance is better at broader taxonomic ranks and the idea that there may be more data for primate infection that might serve as a suitable proxy for zoonotic potential and avoidance of false positives for human infection due to absence of evidence. On this dataset, we report the performance of eight machine learning models for predicting mammal-infecting viruses from their genomic sequences. We find that randomly assigning cases in our improved dataset to training/testing sets, when compared to the original assignments into training/testing in Mollentze et al., increases the overall average ROC AUC of prediction of human infection from 0.663 ± 0.070 to 0.784 ± 0.013, consistent with the reduction in phylogenetic distance between train and test sets (relative entropy change from 3.00 to 0.08). The broadest host category of mammal infection can be predicted most reliably at 0.850 ± 0.020. We share our improved dataset and code to enable standardized comparisons of machine learning methods to predict human host infections. Overall, we have presented preliminary evidence that classification of virus host infection is more tractable at higher taxonomic ranks, that unsurprisingly reducing the phylogenetic distance between training and test sets can improve predictive performance, that peptide kmer features appear to be harmful to out of sample model performance, and we are left with the question of whether models for virus host prediction can reasonably be expected to perform well in out of sample scenarios given the likelihood that viruses do not share a common ancestor. Consistent with this concern, when the data is resampled such that there is no overlap between viral families in training and test sets (relative entropy > 24), models perform no better than random chance at prediction of human infection regardless of whether kmers are included (ROC AUC 0.50 ± 0.08) or not (ROC AUC 0.50 ± 0.04).

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

scPlantAnnotate: an accurate and robust transformer-based model for plant cell type annotation

Accurate cell type annotation remains a major bottleneck in plant single-cell RNA sequencing (scRNA-seq), where existing tools are often adapted from animal studies and perform sub-optimally on plant data. The lack of plant-specific computational frameworks limits the construction of plant cell atlases and downstream biological discovery. We develop and evaluate scPlantAnnotate, a Transformer-based reference annotation framework tailored for plant scRNA-seq data, and benchmark it against state-of-the-art deep learning and conventional methods across multiple plant species. Species-specific scPlantAnnotate models were trained using curated datasets from Arabidopsis thaliana, Zea mays, Oryza sativa, and Glycine max. We compared scPlantAnnotate with leading baselines under both standard random-split evaluation and a more stringent leave-one-dataset-out setting, which tests robustness to completely unseen datasets and tissue types. scPlantAnnotate consistently outperforms existing approaches across all four species under random-split evaluation. In the leave-one-dataset-out setting for A. thaliana, where performance drops markedly for all methods due to strong batch effects and dataset heterogeneity, scPlantAnnotate nonetheless achieves the highest Accuracy, Macro-F1, Balanced Accuracy, and Macro-AUROC on average and ranks first on most held-out datasets. These results demonstrate improved robustness to dataset shifts, a critical yet underexplored challenge in plant scRNA-seq analysis. A freely accessible web server enables users to annotate their own datasets using pretrained models. scPlantAnnotate provides a plant-specific, Transformer-based framework for single-cell annotation that delivers state-of-the-art performance and enhanced robustness to unseen datasets. By addressing limitations of existing tools and enabling scalable reference-based annotation, scPlantAnnotate supports the development of comprehensive plant cell atlases and facilitates broader use of single-cell genomics in plant biology.

Bioinformatics↗
DOE OSTI2024

Binary pseudo-random array standards for calibration of 3D optical surface profilers used for metrology with significantly curved x-ray optics

High-accuracy metrology is vitally important in manufacturing ultra-high-quality free-form mirrors designed to manipulate X-ray light with nanometer-scale wavelengths. The current capabilities and possibility for improvements in X-ray mirror manufacturing are limited by inherent imperfections of the integrated metrology tools. In the case of Fizeau interferometry, metrology tools are currently calibrated with super-polished flat test-standard/reference mirrors. This is acceptable for measuring slightly curved X-ray optics. However, for even moderately curved aspherical X-ray mirrors the flat-reference calibration is not sufficiently accurate and stitching Fizeau interferometer-based surface metrology is used to mitigate the problem. But still, the retrace and aberration errors, as well as the limited spatial resolution, described with the instrument transfer function (ITF), can be transferred into the optical surface topography of X-ray mirrors obtained in stitching metrology. For ITF calibration, we have developed an original technique, based on test standards structured as two-dimensional (2D) highly-randomized (HR) binary pseudo-random arrays (BPRAs). The technique employs the unique properties of the HR BPRA patterns in the spatial frequency domain., i.e. the inherent 2D power spectral density of the HR BPRA pattern has a deterministic white-noise-like character that allows direct determination of the ITF with uniform sensitivity over the entire spatial frequency range and field-of-view of an instrument. Here, we explore technological, metrological, and analytical aspects essential for calibration of the retrace and aberration errors of Fizeau interferometers using different types of tilted test samples, including a super-polished reference mirror for the re-trace calibration and the uniformly redundant array (URA) BPRA standards for the geometrical distortion (aberration) calibration. While the first method was previously demonstrated by researchers at DIAMOND Light Source, a method based on the URA BPRA is described and demonstrated here for the first time. We outline the design and fabrication process used in fabrication of URA BPRA test standards, and present the results of application of the URA BPRA standards demonstrating the high efficacy of our approach to geometrical distortion calibration of Fizeau interferometers. We also discuss the possible sources of unexpected peculiarities of the systematic errors, including an astigmatic character of the retrace error, observed with Fizeau interferometers at the Advanced Light Source X-Ray Optics Laboratory.

Munechika, Keiko↗
DOE OSTI2025

Comparative Analysis of Radial and Random Microstructures of Mesophase Pitch Carbon Fibers

Carbon fibers (CF) with radial and random microstructures are produced. Here, these fibers are subjected to identical treatment before being mechanically tested and analyzed with Weibull analysis, with the results revealing a statistically significant difference in tensile strengths of 2.23 GPa for random CF and 1.69 GPa for radial CF. Raman mapping probed the crystalline structure perpendicular to the fiber axis and found a uniform structure, while wide‐angle X‐ray diffraction showed a significant difference of 7.5 Å in the crystallites’ basal lengths parallel to the fiber. Small‐angle X‐ray scattering is completed parallel to the fiber for the first time. A cross‐section Guinier plot of the 1D azimuthal integration is generated assuming symmetric scattering, and the parallel scatterers are found to have a similar length scale to the crystallite's length, validating the testing method. Finally, transmission electron microscopy is completed on the longitudinal cross‐section of each fiber. The radial carbon fiber is found to have a core–shell structure, as evidenced further by fast Fourier transform images. Through all studies, it is shown that the structure developed during mesophase pitch spinning altered the microstructure, thus impacting the mechanical properties, confirming a direct relationship between processing, structure, and properties.

Scherschel, Alexander [Univ. of Virginia, Charlott↗
DOE OSTI2025

Testing convolutional neural network based deep learning systems: a statistical metamorphic approach

Machine learning technology spans many areas and today plays a significant role in addressing a wide range of problems in critical domains,i.e., healthcare, autonomous driving, finance, manufacturing, cybersecurity,etc. Metamorphic testing (MT) is considered a simple but very powerful approach in testing such computationally complex systems for which either an oracle is not available or is available but difficult to apply. Conventional metamorphic testing techniques have certain limitations in verifying deep learning-based models (i.e., convolutional neural networks (CNNs)) that have a stochastic nature (because of randomly initializing the network weights) in their training. In this article, we attempt to address this problem by using a statistical metamorphic testing (SMT) technique that does not require software testers to worry about fixing the random seeds (to get deterministic results) to verify the metamorphic relations (MRs). We propose seven MRs combined with different statistical methods to statistically verify whether the program under test adheres to the relation(s) specified in the MR(s). We further use mutation testing techniques to show the usefulness of the proposed approach in the healthcare space and test two CNN-based deep learning models (used for pneumonia detection among patients). The empirical results show that our proposed approach uncovers 85.71% of the implementation faults in the classifiers under test (CUT). Furthermore, we also propose an MRs minimization algorithm for the CUT, thus saving computational costs and organizational testing resources.

Computer Science↗