Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Image Processing”

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 505 records · Page 28
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

An efficient instance segmentation approach for studying fission gas bubbles in irradiated metallic nuclear fuel

Abstract Gaseous fission products from nuclear fission reactions tend to form fission gas bubbles of various shapes and sizes inside nuclear fuel. The behavior of fission gas bubbles dictates nuclear fuel performances, such as fission gas release, grain growth, swelling, and fuel cladding mechanical interaction. Although mechanical understanding of the overall evolution behavior of fission gas bubbles is well known, lacking the quantitative data and high-level correlation between burnup/temperature and microstructure evolution blocks the development of predictive models and reduces the possibility of accelerating the qualification for new fuel forms. Historical characterization of fission gas bubbles in irradiated nuclear fuel relied on a simple threshold method working on low-resolution optical microscopy images. Advanced characterization of fission gas bubbles using scanning electron microscopic images reveals unprecedented details and extensive morphological data, which strains the effectiveness of conventional methods. This paper proposes a hybrid framework, based on digital image processing and deep learning models, to efficiently detect and classify fission gas bubbles from scanning electron microscopic images. The developed bubble annotation tool used a multitask deep learning network that integrates U-Net and ResNet to accomplish instance-level bubble segmentation. With limited annotated data, the model achieves a recall ratio of more than 90%, a leap forward compared to the threshold method. The model has the capability to identify fission gas bubbles with and without lanthanides to better understand the movement of lanthanide fission products and fuel cladding chemical interaction. Lastly, the deep learning model is versatile and applicable to the micro-structure segmentation of similar materials.

36 MATERIALS SCIENCE↗
DOE OSTI2016

Advanced Characterization of Rare Earth Elements in Coal Utilization Byproducts

Rare earth elements (REE) in various forms (e.g., crystalline mineral phases; adsorbed/absorbed state on and into organic macerals, neoformed glass from flyash or bottom ash) from domestic feedstocks such as coal deposits to coal utilization byproducts (CUB) have the potential to reduce foreign REE dependence and increase domestic resource security. Characterization is critical for understanding environmental risks related to their fate and transport as well as determining the most practical and economical techniques for concentrating the REE and converting them into chemical stocks for manufacturing. Several complementary electron microscopy (SEM-EDS, EPMA-WDS, FIB-SEM, cathodoluminescence, and XRD) and post image processing techniques were used to understand REE transition from coal to CUB. Sites of interest were identified and imaged and respective elemental x-ray maps acquired and montaged. Pixel classification of SEM imagers was completed using image analysis techniques to quantify the distribution of REE associated features. Quantitative elemental analysis of phases were completed using EMPA-WDS followed by FIB-SEM. The FIB-SEM results were reconstructed into 3D volumes and features of interest (e.g. monazite) were analyzed to determine the structure and volumetric estimation of REEs and thus predict detrital REE phases to ICP-MS results. Trace minerals were identified as pyrite, zircon, REE-phosphates' (monazite, xenotime), and barite within the coal tailings. In CUB, amorphous aluminosilicates, iron oxide cenospheres, and calcium oxides were present; monazite appear to be unaltered and unaffected by the combustion process in these samples. Thermal decomposition may have occurred due to presence of detrital zircon and xenotime and subsequent thin Ca-oxide coating enriched in trace REEs.

Verba, Circe↗
DOE OSTI2017

Characterization of Rare Earth Element Minerals in Coal Utilization Byproducts

The United States currently produces over 100 million tons of coal utilization byproducts (CUB) per year in the form of fly ash, bottom ash, slag, and flue gas (American Coal Ash Association (ACCA), 2015). But this “waste material” also contains potentially useful levels of rare earth elements (REE). Rare earth elements are crucial for many existing and emerging technologies, but the U.S. lacks a domestic, sustainable REE source. Our project explored the possibility of developing a supply of REEs for U.S. technologies by extracting REEs from CUBs. This work offers the potential to reduce our dependence on other countries for supply of these critical elements (NETL, REE 2016 Project Portfolio). Geologic and diagenetic history, industrial preparation methods, and the specific combustion process all play major roles in the composition of CUB. During combustion, inorganic mineral phases of coal particles are fluidized at temperatures higher than 1400oC, so inorganic mineral materials are oxidized, fused, disintegrated, or agglomerated into larger spherical and amorphous (non-crystalline) particles. The original mineralogy of the coal-containing rock and heating/cooling of the material significantly affects the composition and morphology of the particles in the combustion byproduct (Kutchko and Kim, 2006). Thus, different types of coal/refuse/ash must be characterized to better understand mineral evolution during the combustion process. Our research focused on developing a working model to address how REE minerals behave during the combustion process: this research should help determine the most effective engineering methods for extracting REEs from CUBs. We used multimodal imaging and image processing techniques to characterize six rock and ash samples from different coal power plants with respect to morphology, grain size, presence of mineral phases, and elemental composition. The results of these characterization activities provided thresholds for realizing the occurrence of REE mineral phases in CUB and allowed us to calculate structural and volumetric estimates of REE. Collectively, the rock and coal ash samples contained minerals such as quartz, kaolinite, muscovite/illite, iron oxide (as hematite or magnetite), mullite, and clinochlore. Trace minerals included pyrite, zircon, siderite, rutile, diopside, foresterite, gypsum, and barite. We identified REE phosphate minerals monazite (Ce,La,Nd,Th)(PO 4 ,SiO 4 ), xenotime (YPO 4 ,SiO 4 ), and apatite (Ca 5 (PO 4 ) 3 (F,Cl,OH) via SEM and electron microprobe analysis: these materials generally occurred as 1-10 μm-long crystals in the rock and ash samples. As has been shown in other studies, amorphous material-aluminosilicate glass or iron oxyhydroxide-are the major components of coal fly and bottom ash. Trace amounts of amorphous calcium oxide and mixed element (e.g., Al-Si-Ca-Fe) slag are also present. Quartz, mullite, hematite, and magnetite are the crystalline phases present. We found that REEs are present as monomineralic grains dispersed within the ash, as well as fused to or encapsulated by amorphous aluminosilicate glass particles. Monazite and xenotime have relatively high melting points (>1800 °C) compared to typical combustion temperatures; our observations indicate that the REE-phosphates, which presumably contribute a large percentage of REE to the bulk ash REE pool, as measured by mass spectroscopy, are largely unaltered by the combustion. Our study shows that conventional coal combustion processes sequester REE minerals into aluminosilicate glass phases, which presents a new engineering challenge for extracting REE from coal ash. The characterization work summarized in this report provides a semi-quantitative assessments of REE in coal-containing rock and CUB. The data we obtained from 2- and 3-D imaging, elemental mapping, volumetric estimates, and advanced high-resolution pixel classification successfully identified the different mineral phases present in CUB. Further, our characterization results can guide techniques for extracting REEs from CUB, or other geologic and engineered materials. Whilst, interpretations will inform future REE separation and extraction techniques and technologies practical for commercial utilization of combustion byproducts generated by power plants.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2024

Highly sensitive 2D X-ray absorption spectroscopy via physics informed machine learning

Abstract Improving the spatial and spectral resolution of 2D X-ray near-edge absorption structure (XANES) has been a decade-long pursuit to probe local chemical reactions at the nanoscale. However, the poor signal-to-noise ratio in the measured images poses significant challenges in quantitative analysis, especially when the element of interest is at a low concentration. In this work, we developed a post-imaging processing method using deep neural network to reliably improve the signal-to-noise ratio in the XANES images. The proposed neural network model could be trained to adapt to new datasets by incorporating the physical features inherent in the latent space of the XANES images and self-supervised to detect new features in the images and achieve self-consistency. Two examples are presented in this work to illustrate the model’s robustness in determining the valence states of Ni and Co in the LiNi x Mn y Co 1-x-y O 2 systems with high confidence.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Image registration for accurate electrode deformation analysis in operando microscopy of battery materials

Operando imaging techniques have become increasingly valuable in both battery research and manufacturing. However, the reliability of these methods can be compromised by instabilities in the imaging setup and operando cells, particularly when utilizing high-resolution imaging systems. The acquired imaging data often include features arising from both undesirable system vibrations and drift, as well as the scientifically relevant deformations occurring in the battery sample during cell operation. For meaningful analysis, it is crucial to distinguish and separately evaluate these two factors. To address these challenges, we employ a suite of advanced image-processing techniques. These include fast Fourier transform analysis in the frequency domain, power spectrum-based assessments for image quality, as well as rigid and non-rigid image-registration methods. These techniques allow us to identify and exclude blurred images, correct for displacements caused by motor vibrations and sample holder drift and, thus, prevent unwanted image artifacts from affecting subsequent analyses and interpretations. Additionally, we apply optical flow analysis to track the dynamic deformation of battery electrode materials during electrochemical cycling. This enables us to observe and quantify the evolving mechanical responses of the electrodes, offering deeper insights into battery degradation. Together, these methods ensure more accurate image analysis and enhance our understanding of the chemomechanical interplay in battery performance and longevity.

Sun, Tianxiao↗
DOE OSTI2016

SU-F-J-183: Interior Region-Of-Interest Tomography by Using Inverse Geometry System

Purpose: The inverse geometry computed tomography (IGCT) composed of multiple source and small size detector has several merits such as reduction of scatter effect and large volumetric imaging within one rotation without cone-beam artifact, compared to conventional cone-beam computed tomography (CBCT). By using this multi-source characteristics, we intend to present a selective and multiple interior region-of-interest (ROI) imaging method by using a designed source on-off sequence of IGCT. Methods: All of the IGCT sources are operated one by one sequentially, and each projection in the shape of narrow cone-beam covers its own partial volume of full field of view (FOV) determined from system geometry. Thus, through controlling multi source operation, limited irradiation within ROI is possible and selective radon space data for ROI imaging can be acquired without additional X-ray filtration. With this feature, we designed a source on-off sequence for multi ROI-IGCT imaging, and projections of ROI-IGCT were generated by using the on-off sequence. Multi ROI-IGCT images were reconstructed by using filtered back-projection algorithm. All these imaging process of our study has been performed by utilizing digital phantom and patient CT data. ROI-IGCT images of the phantom were compared to CBCT image and the phantom data for the image quality evaluation. Results: Image quality of ROI-IGCT was comparable to that of CBCT. However, the distal axial-plane from the FOV center, large cone-angle region, ROI-IGCT showed uniform image quality without significant cone-beam artifact contrary to CBCT. Conclusion: ROI-IGCT showed comparable image quality and has the capability to provide multi ROI image within a rotation. Projection of ROI-IGCT is performed by selective irradiation, hence unnecessary imaging dose to non-interest region can be reduced. In this regard, it seems to be useful for diagnostic or image guidance purpose in radiotherapy such as low dose target localization and patient alignment. This research was supported by the Mid-career Researcher Program through NRF funded by the Ministry of Science, ICT & Future Planning of Korea (NRF-2014R1A2A1A10050270) and by the Radiation Technology R&D program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (No. 2013M2A2A7038291)

60 APPLIED LIFE SCIENCES↗
DOE OSTI2020

Influence of Vegetation Canopy Architecture on Dynamics of Spatiotemporal Light Distribution and its Effect on Land-Biosphere-Atmosphere Fluxes of Energy, Carbon, and Water

The dynamic spatiotemporal distribution of light and micro climatic variables (wind speed, atmospheric CO 2 concentration, humidity, pressure, and air temperature) within the plant canopy as determined by the canopy architecture play a critical role in land-biosphere-atmosphere (LBA) photosynthetic carbon uptake, transpiratory water loss and energy partitioning. Vegetation models that include effects of sun and shade flecks, leaf temperature and energy balance partitioning between sensible and latent heat fluxes, and reduced turbulent mixing in dense vegetation canopies can accurately predict canopy performance, particularly its shaded leaves, which can constitute 80% of leaf area and contribute up to 50% of carbon update and 65% of water loss. However, current vegetation models with their bulk characterization of shaded leaves poorly represent the complex three dimensional (3D) canopy architecture are insufficient to capture the spatiotemporal dynamics of shaded leaves. Here, we propose to employ an x-ray computed tomography (XCT) imaging technique in combination with an automated, high throughput, image processing pipeline (3DArch, to be developed) to obtain an explicit 3D triangulated surface mesh of a plant canopy (Fig 1). 3DArch would provide a much needed technological jump in 3D LBA modeling where computer generated artificial 3D models suffer from lack of high resolution detailed canopy architecture parameters. The 3D canopy surface mesh output from 3DArch is coupled with an explicit 3D forward ray tracing model (fastTracer) to obtain a high resolution representation of the dynamic spatiotemporal light environment within the canopy over the course of a day (Fig 1). The results from the light attenuation model are then coupled with a process based multi-scale bio-physical canopy and leaf model of photosynthesis (PlantGro) to get fluxes of carbon, water and energy at the leaf and canopy scale, in addition to estimating whole plant carbon uptake, and water loss. Our proposed model framework would be the first of its kind to provide a breakthrough capability in the field of process-based plant modeling that bridges the key technological gap in shaded leaf modeling through explicit representation of 3D canopy architecture. Computer generated artificial 3D model canopy simulations on grasses show that the canopy light distribution does not follow an exponential behavior as assumed in most 1D vegetation models, but rather exhibits a heterogeneous behavior, with a small but significant portion of interior canopy leaves experiencing peak light levels due to sun flecks while the rest of the canopy receives substantially lower light levels. These results indicate a bias in traditional models that overestimates shaded leaf contribution towards whole canopy photosynthesis by up to 15% and underestimate transpiration by up to 12% with an associated increase in sensible heat fluxes since photosynthesis, transpiration and energy partitioning are tightly coupled processes. However, these test results need to be validated using the 3D architecture obtained from a real plant imaged in 3D using 3DArch, rather than an artificial model plant. Our proposed model framework would allow us to identify inefficiencies in canopy light distribution and prescribe optimal canopy architectures (leaf optical properties, leaf area, leaf angles, row spacing, planting density, etc.) to traditional and genetic engineering based crop breeders, for maximizing target functions such as light use efficiency, water use efficiency, drought tolerance, crop yield, crop resilience etc., with important applications in biofuel crops such as sorghum, and switchgrass. Our model can also be applied for designing optimal canopy architecture that increase surface albedo as a mitigation strategy for global warming and projections of crop yield for future food and water security under climate change.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Linking Transient Voltage to Spatially-Resolved Luminescence Imaging to Understand Reliability of Perovskite Photovoltaics: Preprint

In this work, we present a methodology to separate effects of perovskite device metastability from irreversible degradation, using stress/rest cycling under constant current bias while collecting a series of electroluminescence images and continuously monitoring voltage. We develop a simulation model and procedures for image processing to better understand the effects of ion parameters on the transient nature of voltage and evolving electroluminescence images.

41 EE - Solar Energy Technologies Office (EE-4S)↗
DOE OSTI2018

Automated data processing architecture for the Gemini Planet Imager Exoplanet Survey

The Gemini Planet Imager Exoplanet Survey (GPIES) is a multiyear direct imaging survey of 600 stars to discover and characterize young Jovian exoplanets and their environments. We have developed an automated data architecture to process and index all data related to the survey uniformly. An automated and flexible data processing framework, which we term the Data Cruncher, combines multiple data reduction pipelines (DRPs) together to process all spectroscopic, polarimetric, and calibration data taken with GPIES. With no human intervention, fully reduced and calibrated data products are available less than an hour after the data are taken to expedite follow up on potential objects of interest. The Data Cruncher can run on a supercomputer to reprocess all GPIES data in a single day as improvements are made to our DRPs. A backend MySQL database indexes all files, which are synced to the cloud, and a front-end web server allows for easy browsing of all files associated with GPIES. To help observers, quicklook displays show reduced data as they are processed in real time, and chatbots on Slack post observing information as well as reduced data products. Together, the GPIES automated data processing architecture reduces our workload, provides real-time data reduction, optimizes our observing strategy, and maintains a homogeneously reduced dataset to study planet occurrence and instrument performance.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

Automated TEM Reveals Intercrystalline Correlations of Conjugated Polymers

Transmission electron microscopy (TEM) continues to transform polymer science by revealing key aspects of chain packing, phase separation and nanoscale structure. The development of instrumentation and data analyses tools is driving the field forward and enabling new experiments. Here, we use automated high-resolution TEM (HRTEM) and image processing to identify the structure of a conjugated polymer used in organic electronics. Analysis of more than 600 HRTEM images reveals lattice parameters and orientation correlations between crystals, including the preferred alignment of neighboring crystals along the same crystallographic direction that is likely the result of liquid crystalline order.

Fair, Ryan A. [Pennsylvania State University, Univ↗
DOE OSTI2022

BFSVBF (BatFinder Smart Video BioFilter) [SWR-22-87] and Multi-class BatFinder Smart Video BioFilter Keras

Bats are notoriously difficult to study, therefore, identifying specific behavioral trends and the precise environmental conditions at the time of collision requires a monitoring solution that can reliably collect relevant data. To date, thermal infrared video surveillance has been extensively applied to study bats and has proven to be a powerful yet cumbersome tool. Current analytical approaches are time consuming because data processing data has not been fully automated. In the past, steps have been taken to record avian and bat activity in conjunction with complicated image processing techniques that separate species from other moving objects within the field of view (i.e. clouds and portions of the wind turbine). Once the videos are collected, the post-processing does not allow real time monitoring and identification, leading to a delay in both studying the behavior of these species and determining the effectiveness of any impact reduction strategy being studied. Moreover, object identification capability is lacking, thus limiting the usefulness of video data. To resolve these issues, we are using open source computer vision and machine learning techniques allowing for automatic detection of objects in real-time with the ability to correlate these objects with environmental variables and recording the flight paths of each object. The code has gone through five rounds of development with images used to train the models. This advancement allows for automated real-time data collection, identification, and tracking, thereby eliminating the need for long and tedious post-analysis processing of the videos. We will discuss the two open source and publicly available machine learning models developed within this scope of this work: 1) a binary model with a 97.5% accuracy in identifying the difference between an object and an empty scene, including wind turbine and clouds; and 2) a multiple classification model with the capability of identifying the type of object detected: bats (90% accuracy), birds (83% accuracy), insects (69% accuracy) and non-biological (99% accuracy).

Yarbrough, John↗
DOE OSTI2025

Imaging of dynamic processes in materials with a laser-wakefield accelerator

Betatron x rays generated from laser-wakefield accelerators are a promising source for imaging dynamic processes in materials. Here, we present proof-of-concept imaging of microstructural evolution in a hypermonotectic Al-Bi alloy, which consists of liquid Bi particles in a solid Al matrix. The images capture the elongation and fragmentation of Bi particles upon isothermal annealing. Because of the femtosecond time scale of the betatron source, the images are not subject to motion blur, whereas the accessibility of the source allows for studies of long-term processes such as annealing. The high-resolution data reveal that the evolution of Bi particles is mediated by an interplay of grain-boundary wetting and morphological instability, in stark contrast to an earlier proposal for (inverse) coarsening.

Alloys↗
DOE OSTI2015

Efficient Algorithms for Convolutional Sparse Representations

When applying sparse representation techniques to images, the standard approach is to independently compute the representations for a set of overlapping image patches. This method performs very well in a variety of applications, but results in a representation that is multi-valued and not optimized with respect to the entire image. Additionally, an alternative representation structure is provided by a convolutional sparse representation, in which a sparse representation of an entire image is computed by replacing the linear combination of a set of dictionary vectors by the sum of a set of convolutions with dictionary filters. The resulting representation is both single-valued and jointly optimized over the entire image. While this form of a sparse representation has been applied to a variety of problems in signal and image processing and computer vision, the computational expense of the corresponding optimization problems has restricted application to relatively small signals and images. Here, this paper presents new, efficient algorithms that substantially improve on the performance of other recent methods, contributing to the development of this type of representation as a practical tool for a wider range of problems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Coded-mask-based X-ray phase-contrast and dark-field imaging

Phase contrast and dark-field X-ray imaging enable imaging of objects that absorb or reflect very little X-ray light. Disclosed is a method and systems for performing coded-mask-based multi-contrast imaging (CMMI). The method includes providing radiation to a coded mask that has a known phase and absorption profile according to a pre-determined pattern. The radiation is then impingent upon a sample, and the radiation is detected to perform phase-reconstruction and image processing. The method and associated systems allow for the use of maximum-likelihood and machine learning methods for reconstruction images of the sample from the detected radiation.

Shi, Xianbo↗
DOE OSTI2016

Direct observations of damage during unconfined brittle failure of Carrara marble

Abstract To observe and quantify the production of microfracturing from initial yield to failure, we deformed Carrara marble samples in uniaxial compression at 20, 105, and 180°C and continuously observed a region of about 1 mm 2 on an exposed face with a long‐working distance microscope. Using image processing and microscale strain‐mapping techniques, we measured local strains over a length scale of tens of micrometers. By treating the images with various filters, we identified linear damage features, as well as the magnitude of localized strain and the mode of deformation, i.e., shear versus normal deformation. In general, shear deformation is more prevalent after initial yielding, while tensile deformation dominates closer to peak stress. Independent measurements of both stress and microcrack density at different stages of each experiment provide a unique opportunity to explicitly compare the data with damage models. The model of Ashby and Sammis (1990) significantly underestimated the damage that the rock could sustain before peak stress, perhaps owing to the influence of weak grain boundaries on the damage production. In these samples, microcracks tended to form near boundaries before yield stress. During strain hardening, the damage parameters increased rapidly as longer microcracks grew along the boundaries and finally transected grains as loading neared peak stress. The microcrack density can be empirically related to the reduction of Young's modulus; stiffness ratios decay exponentially with increasing microcrack density for T ≤ 105°C.

Tal, Yuval↗
DOE OSTI2024

Comparison of the spatial statistics of random and defined-sequence photoresist films

The resolution-line edge roughness-sensitivity tradeoff has motivated the exploration of potential improvements using defined sequence polymers and polymer-bound photoacid generators and quenchers. We characterize the internal structures of positive tone photoresist polymer films formed from defined sequence polymers and compare them with random copolymers of the same composition. We model their imaging to connect initially to developable film structures. We use a polymer packing algorithm to simulate films of diverse compositions and locations of photoacid generators and quenchers, using the composition of an ESCAP photoresist. We use a simple extreme ultraviolet exposure-deprotection algorithm to model developable image formation within them. In all cases, the spatial distribution of chemical moieties in the film for defined sequence polymers is nearly indistinguishable from random copolymers. We evaluate several exposure-deprotection scenarios and find that a defined sequence copolymer has a distinctive developable image under certain circumstances. The use of defined sequence polymers within a photoresist layer does not automatically result in improved imaging; however, they do have some characteristics different from random polymers of the same composition. Further study of these characteristics may provide a route to improved control over the nanoscale imaging process.

36 MATERIALS SCIENCE↗
DOE OSTI2025

FPGA-accelerated SpeckleNN with SNL for real-time X-ray single-particle imaging

We present the implementation of a specialized version of our previously published unified embedding model, SpeckleNN, for real-time speckle pattern classification in X-ray Single-Particle Imaging (SPI), using the SLAC Neural Network Library (SNL) on an FPGA platform. This hardware realization transitions SpeckleNN from a prototypic model into a practical edge solution, optimized for running inference near the detector in high-throughput X-ray free-electron laser (XFEL) facilities, such as those found at the Linac Coherent Light Source (LCLS). To address the resource constraints inherent in FPGAs, we developed a more specialized version of SpeckleNN. The original model, which was designed for broader classification across multiple biological samples, comprised ~5.6 million parameters. The new implementation, while reducing the parameter count to 64.6K (a 98.8% reduction), focuses on maintaining the model's essential functionality for real-time operation, achieving an accuracy of 90%. Furthermore, we compressed the latent space from 128 to 50 dimensions. This implementation was demonstrated on the KCU1500 FPGA board, utilizing 71% of available DSPs, 75% of LUTs, and 48% of FFs, with an average power consumption of 9.4W according to the Vivado post-implementation report. The FPGA performed inference on a single image with a latency of 45.015 microseconds at a 200 MHz clock rate. In comparison, running the same inference on an NVIDIA A100 GPU resulted in an average power consumption of ~73W and an image processing latency of around 400 microseconds. Our FPGA-accelerated version of SpeckleNN demonstrated significant improvements, achieving an 8.9 × speedup and a 7.8 × reduction in power consumption compared to the GPU implementation. Key advancements include model specialization and dynamic weight loading through SNL, which eliminates the need for time-consuming FPGA design re-synthesis, allowing fast and continuous deployment of models (re)trained online. These innovations enable real-time adaptive classification and efficient vetoing of speckle patterns, making SpeckleNN more suited for deployment in XFEL facilities. This implementation has the potential to significantly accelerate SPI experiments and enhance adaptability to evolving experimental conditions.

47 OTHER INSTRUMENTATION↗
DOE OSTI2025

Automated Scoring of Morphological Changes in Images of Pentaerythritol Tetranitrate

Recent advances in characterization techniques that generate large datasets of material microstructure images require robust, automated image-processing. We applied an unsupervised anomaly detection method called feature anomaly detection system (FADS) to automatically detect and quantify microstructure changes in images of the explosive pentaerythritol tetranitrate (PETN) aged at various temperatures. We demonstrated the FADS approach on two-dimensional images extracted from computed tomography scans, but the same technique can be readily applied to other imaging modalities. FADS calculates anomaly scores on the basis of differences in filter activations of nominal and test data in pretrained convolutional neural networks. The FADS scores successfully differentiated between pristine PETN and PETN aged at a temperature where material coarsening occurred. Morphological metric analysis of segmented images verified observed trends in FADS scores as a function of aging temperature and aging time, specifically by calculating volume fractions, specific boundary lengths, two-point correlation functions, and local thicknesses. Here, the FADS technique has two important advantages compared to traditional morphological analysis: First, it uses grayscale images as input, rather than images that are segmented to separate the appropriate phases; and second, FADS scores capture any type of changes among image sets, rather than requiring prior knowledge or selection of a relevant set of metrics.

Accelerated aging↗