Synthesis and characterization of sintered H–Y zeolite-derived waste forms for dehalogenated electrorefiner salt
Not Available
Gardner, Levi D.↗
SEARCH · Search NASA
Results for “Advanced Baseline Imager”
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.
TopTemp: Parsing Precipitate Structure from Temper Topology
Technological advances are in part enabled by the development of novel manufacturing processes that give rise to new materials or material property improvements. Development and evaluation of new manufacturing methodologies is labor-, time-, and resource-intensive expensive due to complex, poorly defined relationships between advanced manufacturing process parameters and the resulting microstructures. In this work, we present a topological representation of temper (heat-treatment) dependent material micro-structure, as captured by scanning electron microscopy, called TopTemp. We show that this topological representation is able to support temper classification of microstructures in a data limited setting, generalizes well to previously unseen samples, is robust to image perturbations, and captures domain interpretable features. The presented work outperforms conventional deep learning baselines and is a first step towards improving understanding of process parameters and resulting material properties.
Kassab, Lara↗
Reducing the Energy Consumption of Magnetic Resonance Imaging and Computed Tomography Scanners: Integrating Ecodesign and Sustainable Operations
This review aims to provide valuable insights into how energy consumption in magnetic resonance imaging (MRI) and computed tomography (CT) scanners can be effectively monitored, managed, and reduced, thereby contributing to more sustainable medical imaging practices. Demand for advanced imaging technologies such as MRI and CT scanners continues to increase, and understanding the resultant impact on greenhouse gas emissions requires a thorough evaluation of their energy consumption. Here, this review examines the energy monitoring and consumption characteristics of MRI and CT scanners, highlighting potential approaches for energy savings. An overview of MRI and CT principles, hardware components, and their associated energy consumption is provided. After addressing the technical aspects, the hardware and software requirements essential for accurate energy metering are detailed. Baseline measurements of energy consumption data are then provided as a foundation to understand current usage patterns and identify areas for improvement. Ongoing efforts to reduce energy consumption are categorized into 3 main strategies: operations, scanner design enhancements, and active scanning techniques, including accelerated MRI protocols. Ultimately, we emphasize that achieving sustainability in medical imaging requires collaboration across disciplines. By incorporating eco-friendly design in new imaging equipment, we can reduce the environmental impact, promote sustainability, and set a health care industry standard for a healthier planet.
62 RADIOLOGY AND NUCLEAR MEDICINE↗
High-Fidelity Energy Deposition Ignition Model Coupled with Flame Propagation Models at Engine-like Flow Conditions
With the heightened pressure on car manufacturers to increase the efficiency and reduce the carbon emissions of their fleets, more challenging engine operation has become a viable option. Highly dilute, boosted, and stratified charge, among others, promise engine efficiency gains and emissions reductions. At such demanding engine conditions, the spark-ignition process is a key factor for the flame initiation propagation and the combustion event. From a computational standpoint, there exist multiple spark-ignition models that perform well under conventional conditions but are not truly predictive under strenuous engine operation modes, where the underlying physics needs to be expanded. In this paper, a hybrid Lagrangian-Eulerian spark-ignition (LESI) model is coupled with different turbulence models, grid sizes, and combustion models. The ignition model, previously developed, relies on coupling Eulerian energy deposition with a Lagrangian particle evolution of the spark channel, at every time-step. The spark channel is attached to the electrodes and allowed to elongate at a speed derived from the flow velocity. The LESI model is used to simulate spark ignition in a nonquiescent crossflow environment at engine-like conditions, using converge commercial computational fluid dynamics (CFD) solver. The results highlight the consistency, robustness, and versatility of the model in a range of engine-like setups, from typical with Reynolds-averaged Navier-Stokes (RANS) and a larger grid size to high fidelity with large-eddy simulation (LES) and a finer grid size. The flame kernel growth is then evaluated against Schlieren images from an optical constant volume ignition chamber with a focus on the performance of flame propagation models, such as G-equation and thickened flame model, versus the baseline well-stirred reactor model. Finally, future development details are discussed.
Advanced ignition modeling↗
X-ray Computed Tomography of Irradiated and Unirradiated AGR-3/4 Compacts
X-ray Computed Tomography (XCT) has been utilized to image and characterize compacts from the combined third and fourth irradiation of the Advanced Gas Reactor (AGR) Program, AGR-3/4, fuel. The experiment contained tristructural isotropic (TRISO)-coated fuel particles as well as designed-to-fail (DTF) fuel particles. Two irradiated compacts, representing the lower and higher range of AGR-3/4 burnup (4.85% and 14.92% fissions per initial heavy metal atom FIMA) were examined. These represent the first known highly irradiated TRISO fuel compacts to be examined via X-ray CT. Additionally, two unirradiated compacts from the same production batch as the examined irradiation compacts were also imaged for a baseline comparison. As XCT of irradiated TRISO compacts is not a commonly implemented characterization technique, a significant portion of the report focuses on developed methodology and imaging conditions. A specialized sample shielding device was developed and fabricated specifically to limit received dose to staff during sample preparation for XCT and to minimize excess gamma radiation dose to sensitive electronic components with the utilized X-ray system. Significant penetration through the uranium oxycarbide fuel kernels by significantly hardening the X-ray beam with specialized proprietary filters acquired from Carl Zeiss NTS Ltd. The filter utilized resulted in an average X ray photon energy of ~110 keV which approaches uranium’s K-edge (~115 keV), maximizing penetration for a microfocus X-ray source. The gamma-radiation emitted from the irradiated AGR-3/4 TRISO compacts, has the same properties and mechanisms for interaction with matter as X-rays, thus the detection of gamma-radiation by the utilized X-ray detectors was initially a concern. However, although ?-rays did produce an observable signal on the X-ray detector, its contribution to the overall imaging results appeared negligible upon 3D reconstruction. The neglibile impact on the resulting 3D reconstructed volumes were likely the result of: (1) a significantly lower detection efficiency for ?-rays relative to X-rays; (2) An X-ray flux at the detector several orders of magnitude higher than that of the impinging ?-rays from the irradiated compacts. These results suggest that irradiated compacts with significantly higher radiation fields can be examined in the future if an acceptable route for sample handling and preparation can be determined. Additionally, the 3D imaging results of XCT can provide a valuable means of assessing compacts. While in many ways complimentary to traditional post irradiation examination techniques such as optical ceramography, XCT can provide additional insight into compact features traditionally difficult to discern directly from cross-sectional imaging alone. Preliminary analyses on kernel size, morphology (aspect ratio and sphericity), and kernel orientation were presented. Sphericity, a simple morphological shape descriptor, was utilized to screen for kernel extrusions within the high burnup compact. The number of kernel extrusions identified via XCT represented an approximate two-fold increase from the quantity of extruded particles observed (via optical ceramography) in adjacent compacts from the same irradiation capsule. While numerical analysis of the compact datasets was highly preliminary, initial results show promise for providing complimentary metrics to current AGR-3/4 PIE and potentially additional insight into the processes driving TRISO fuel degradation during reactor operation. Additional analyses to be performed at a later date include a more detailed examination of kernel size, kernel sphericity (and observed kernel extrusions), and sphericity. Given all particles can be observed in a single data volume possible correlation of spatial position with observed kernel features will also be made at a later date.
11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
Vehicular Re-Identification from Uncontrolled Multiple Views
Vehicle re-identification (re-ID) across disparate sensing modalities remains a fundamental challenge for transportation research. In this work, we introduce a deep multi-view vehicle re-ID framework that leverages Siamese networks to compare pairs of vehicle images and produce matching scores, enabling robust association across drastically different viewpoints such as those from UAVs, surveillance cameras, and ground sensors. The model exploits convolutional neural networks to learn features that remain discriminative under changes in angle, distance, and illumination, supporting more generalizable re-ID performance. As part of this effort, we also developed an automated pipeline to synchronize roadside and UAV video streams, producing a multi-perspective dataset that complements preexisting real collections and a synthetic dataset generated in this study. Together, these contributions advance the capability to re-identify vehicles across wide viewing baselines; establish a foundation for scalable, reproducible research in vehicle re-ID; and open pathways for future applications, such as inferring routine behaviors, movement patterns, and daily habits of the individual associated with the vehicle.
convolutional neural networks↗
Active Learning Surrogates for Integrating Electron Microscopy and Computational Insights from Simulations in Autonomous Experiments
Artificial Intelligence (AI) combined with simulations and experiments has great potential to accelerate scientific discovery across technology and pharmaceuticals. However, the gap between simulations and experiments is challenging due to disparities in time and scale, making it difficult to estimate properties like energy and electronic states from experiments, and to provide feedback based on theoretical insights.Our research addresses the challenge by developing unique deep kernel based surrogate models that learns from microscopic images, mapping structural features to energy differences from defect formation. We start with full-training using simulated images to determine optimal settings, establishing a baseline for active learning. Using these settings from the baseline, active learning is trained, and predicts structures along simulation trajectories based on uncertainty and energetic stability, thus reducing data requirements, simulation time and computational costs. The results demonstrate that the model achieves a low average error margin of approximately 0.03 meV, indicating good performance. To enhance feature extraction and reconstruction capabilities, we developed an autoencoder-decoder as additional surrogate to create latent space to capture essential features, enabling precise comparisons between simulations and experiments. The results from this model achieved a reconstruction loss of around 0.2 and accurately reconstructed molecular structures.Overall, this work advances the steering of experiments through computational simulations by employing a surrogate models that actively predicts the trajectories of structural evolution, achieving time-to-solution comparable to experimental measurements.
Saranathan, Gayathri [Hewlett-Packard]↗
3013 Baseline Inner Container Report: Description, Imaging, and Analysis
This report presents consolidated findings from three microscopy studies of baseline Bagless Transfer Containers (BTCs), which are the inner containers used in DOE Standard 3013-compliant plutonium storage packages. The studies were conducted by Los Alamos National Laboratory (LANL), Savannah River National Laboratory (SRNL), and DNV, a subcontractor specializing in corrosion science. The baseline BTCs examined had not been exposed to corrosive environments, serving as control samples for comparison with destructively examined (DE) containers. Additionally, the Background section provides an overview of the BTC design, including manufacturing methods, material properties, and container configurations that may influence corrosion behavior. The primary objective of the LANL and SRNL studies was to establish reference characteristics of BTCs to distinguish manufacturing artifacts from corrosion-induced features observed in DE containers. Both laboratories used advanced microscopy techniques, including Wide Area Microscopy (WAMS), Laser Confocal Microscopy (LCM), Scanning Electron Microscopy (SEM), and Focused Ion Beam (FIB), to reveal that most surface anomalies were shallow and attributable to flow-forming processes, not corrosion. Subsurface impurities, identified as non-metallic inclusions, were also observed. DNV’s investigation focused on characterizing the mechanical and chemical properties of the BTC material to support crack growth modeling. Their findings indicate higher hardness and strength but reduced ductility—consistent with flow-formed 304L stainless steel and a higher-than-expected martensite content, in contrast to manufacturer claims. Ongoing and future studies, including Electron Backscatter Diffraction (EBSD), aim to further understand the microstructural factors influencing crack initiation and propagation. The combined studies provide a crucial baseline for distinguishing corrosion effects in DE containers, contributing to the safe and reliable long-term storage of plutonium-bearing materials.
36 MATERIALS SCIENCE↗
Bridging multimodal microscopy for advanced characterization on nuclear fuel using machine learning
Uranium dioxide (UO 2 ), widely used as driver fuel in light water reactors, experiences microstructure and property change by nuclear fission reactions. This paper bridges the characterization of fresh UO 2 fuel at different length scales, serving as a baseline for future post irradiation examination of irradiated UO 2 fuel. To characterize the microstructural change of nuclear fuel, modern approaches cover a wide range of length scales through different characterization techniques, such as mm scale for Synchrotron-based X-ray computed tomography (SXCT) and microscale for focused ion beam (FIB) and scanning electron microscopy (SEM). It is challenging to bridge the data and knowledge of the same sample in different length scales. This paper proposed a deep learning framework leveraging transfer learning to detect microstructural defects, trained from a sparse FIB, SEM, and SXCT images. The proposed model achieved superior performance in defect segmentation on multiscale microscopic data compared to four of the latest deep learning models.
36 MATERIALS SCIENCE↗
Coal to Carbon Fiber (C2CF) Continuous Processing for High Value Composites (Final Report)
Coal tar is a condensed and recovered by-product of the coking of metallurgical coal for steel production. The heaviest fraction of distilled coal tar is an isotropic pitch largely used as a binder in the manufacturing of carbonaceous electrodes for primary aluminum smelting and in electric arc furnaces. Coal tar pitch offers high carbon yield upon carbonization. In this project, a process to convert the domestically sourced isotropic coal tar pitch, containing very low particulates (QI = 0.32 wt.%), to form flow-domain mesophase pitch amenable for melt spinning into precursor (green) fibers for carbon fiber, was developed and optimized. The final reproducible processing developed is reviewed in this report, along with several characterizations of the mesophase pitch. A final definition of the characteristics of a ‘spinnable’ mesophase pitch is presented. With this mesophase pitch, reproducible and stable multifilament melt spinning was developed, producing green fiber tows with filament diameters of approximately 20 m. The multifilament melt spinning was the most challenging aspect of the project and required the most effort. The best practices learned from this project for melt spinning are reviewed in this report. Once spun, the green fibers were oxidatively stabilized, carbonized and graphitized under inert gas atmosphere to form the final carbon fibers. Given the relatively high softening point of the mesophase pitch, no issues of interfilament fusion were observed during batch oxidation, and subsequent batch carbonization and graphitization went smoothly in all cases. An ~ 80 wt.% conversion of the green fiber mass to final carbon fiber was achieved. After graphitization, the carbon fibers showed high tensile moduli (most were ~ 600 GPa, or 87 Msi) consistent with commercially-available high-performance pitch-based carbon fiber. However, tensile strength and strain to failure were comparatively low. Further work to reduce defects in and on the fiber surfaces would increase these properties. SEM imaging of the graphitic fiber textures is presented herein. Rudimentary composites were fabricated from the carbon fibers and characterized showing similar modulus to baseline composites fabricated with commercial carbon fiber. Finally, a basic economic analysis was done showing the potential to increase the value of the isotropic coal tar pitch by up to 13.6 to 136 times based on a carbon fiber value of $\$$5/lb to $\$$50/lb, respectively. Moreover, the site case study suggests that the coal tar from the single integrated steel mill could supply production of up to 16 kt/yr of carbon fiber. Finally, a technological gap analysis was done which shed light on remaining technical challenges. These challenges included: recovery and utilization of condensates from the mesophase pitch processing, further advancing and increasing stability of the multifilament melt spinning processing, optimization of the oxidation processing, defect reduction for increased carbon fiber strength, and the development of a weaving process towards carbon fiber fabrics. To maximize the coal value chain, the primary objectives of this project were to (a) develop and scale efficient processing technology for producing melt-spinnable mesophase pitch from isotropic coal tar pitch, (b) clarify and simplify tedious continuous fiber processing technologies (particularly multifilament melt spinning of mesophase pitch) towards the efficient production of high performance carbon fiber, and (c) demonstrate and characterize representative composite parts derived from the final carbon fiber. Immense progress was made on all 3 objectives and is detailed in this report.
01 COAL, LIGNITE, AND PEAT↗
AI-optimized detector design for the future Electron-Ion Collider: the dual-radiator RICH case
Advanced detector R&D requires performing computationally intensive and detailed simulations as part of the detector-design optimization process. Here, we propose a general approach to this process based on Bayesian optimization and machine learning that encodes detector requirements. As a case study, we focus on the design of the dual-radiator Ring Imaging Cherenkov (dRICH) detector under development as a potential component of the particle-identification system at the future Electron-Ion Collider (EIC). The EIC is a US-led frontier accelerator project for nuclear physics, which has been proposed to further explore the structure and interactions of nuclear matter at the scale of sea quarks and gluons. We show that the detector design obtained with our automated and highly parallelized framework outperforms the baseline dRICH design within the assumptions of the current model. Our technique can be applied to any detector R&D, provided that realistic simulations are available.
46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
Pre-training Vision Models for the Classification of Alerts from Wide-field Time-domain Surveys
Modern wide-field time-domain surveys facilitate the study of transient, variable and moving phenomena by conducting image differencing and relaying alerts to their communities. Machine learning tools have been used on data from these surveys and their precursors for more than a decade, and convolutional neural networks (CNNs), which make predictions directly from input images, saw particularly broad adoption through the 2010s. Since then, continually rapid advances in computer vision have transformed the standard practices around using such models. It is now commonplace to use standardized architectures pre-trained on large corpora of everyday images (e.g., ImageNet). In contrast, time-domain astronomy studies still typically design custom CNN architectures and train them from scratch. Here, we explore the effects of adopting various pre-training regimens and standardized model architectures on the performance of alert classification. We find that the resulting models match or outperform a custom, specialized CNN like what is typically used for filtering alerts. Moreover, our results show that pre-training on galaxy images from Galaxy Zoo tends to yield better performance than pre-training on ImageNet or training from scratch. We observe that the design of standardized architectures are much better optimized than the custom CNN baseline, requiring significantly less time and memory for inference despite having more trainable parameters. On the eve of the Legacy Survey of Space and Time and other image-differencing surveys, these findings advocate for a paradigm shift in the creation of vision models for alerts, demonstrating that greater performance and efficiency, in time and in data, can be achieved by adopting the latest practices from the computer vision field.
79 ASTRONOMY AND ASTROPHYSICS↗
Deep learning-based spatio-temporal fusion for high-fidelity ultra-high-speed X-ray radiography
Full-field ultra-high-speed (UHS) X-ray imaging experiments have been well established to characterize various processes and phenomena. However, the potential of UHS experiments through the joint acquisition of X-ray videos with distinct configurations has not been fully exploited. In this paper, we investigate the use of a deep learning-based spatio-temporal fusion (STF) framework to fuse two complementary sequences of X-ray images and reconstruct the target image sequence with high spatial resolution, high frame rate and high fidelity. We applied a transfer learning strategy to train the model and compared the peak signal-to-noise ratio (PSNR), average absolute difference (AAD) and structural similarity (SSIM) of the proposed framework on two independent X-ray data sets with those obtained from a baseline deep learning model, a Bayesian fusion framework and the bicubic interpolation method. The proposed framework outperformed the other methods with various configurations of the input frame separations and image noise levels. With three subsequent images from the low-resolution (LR) sequence of a four times lower spatial resolution and another two images from the high-resolution (HR) sequence of a 20 times lower frame rate, the proposed approach achieved average PSNRs of 37.57 dB and 35.15 dB, respectively. When coupled with the appropriate combination of high-speed cameras, the proposed approach will enhance the performance and therefore the scientific value of UHS X-ray imaging experiments.
deep learning↗
On-policy learning-based deep reinforcement learning assessment for building control efficiency and stability
Artificial intelligence technologies have emerged as a game changer not only in specific applications such as image recognition and machine translation but also in many scientific domains. In particular, as deep reinforcement learning (DRL) has shown great success in complex control problems, DRL-based control has been considered as a potential solution to efficiently control and manage building systems. However, broad assessment of DRL-based building control is still required to characterize their pros and cons in comparison with conventional building control methods (e.g., rule-based feedback controls). In this paper, we assessed DRL-based controls with on-policy learning-based algorithms and continuous control actions for cooling control of large office buildings in the summer season to minimize whole-building energy use and occupant discomfort. We compared DRL-based control methods with two baseline control methods: (1) a pre-determined schedule with supply temperature and static pressure setpoints, and (2) advanced reset method that adjusts setpoints based on heuristic rules, i.e., ASHRAE Guideline 36. We also tested the DRL algorithms to evaluate their performances in multiple climate locations. We found that DRL-based control methods outperformed the baseline control methods in terms of energy savings while maintaining a thermal comfort. DRL reduced energy use between ~4%–22% on average compared to the baseline methods, depending on climate location. We also evaluated DRL-based control in terms of control stability and showed that DRL-based methods should address the span of hardware lifetimes in practical operations.
control stability↗
Selective Amnesia using Contrastive Subnet Erasure for Class Level Unlearning in Vision Models
We study concept-level forgetting in pretrained vision models: removing an entire semantic category so the system no longer recognizes that object in unseen images and contexts, rather than merely forgetting specific training examples. Prior work either applies blunt global projections or fine-tunes parameters, which can introduce collateral damage to unrelated features, add compute, and become unstable as forgetting strength increases. We introduce Contrastive Subnet Erasure (CSE), a training-free, encoder-centric edit that targets a compact set of channels most responsible for the class and attenuates them in a calibrated manner. The modification is algebraically folded into the subsequent layer, yielding no inference-time overhead and leaving task heads unchanged. To evaluate whether forgetting generalizes beyond the data used to specify the class, we introduce a cross dataset protocol in which the class is defined on a source dataset and performance is measured on a disjoint target dataset drawn from a different distribution with no shared images. This setup tests whether the model still fails to recognize the object when it looks different or appears in new scenes, and it helps avoid overfitting to patterns in the source dataset. Across CIFAR 10, CIFAR 100, and ImageNet under this protocol, CSE achieves stronger forgetting of the target class while better preserving non target utility than existing baselines in both single class and multi class settings. Overall, CSE provides a simple, stable, and deployment-ready mechanism for class-level unlearning in vision.
Kotevska, Olivera [ORNL] (ORCID:0000000316772243)↗
Automated tape-target centering based on template matching for construction measurement tasks using robotic total station
Robotic total stations have transformed surveying and construction mapping through precise, efficient, and automated measurements. These instruments integrate a theodolite, which measures horizontal and vertical angles, with an electronic distance measurement (EDM) unit to determine distances, allowing accurate 3D measurement of observable points. Traditionally, users manually aimed the total station at a target. Recent advancements in robotic motor control and integration of cameras now enable automatic rotation and aiming, typically requiring only a single operator to position the target. Automated aiming is commonly performed using retroreflector prisms, which reflect light back to the source with minimal scattering, enabling high-precision measurements over long distances. However, retroreflectors, especially those designed for 360 degree, can be expensive and impractical for certain applications. This paper presents an algorithm for automating the center detection of low-cost disposable tape targets using the total station's onboard camera and a template matching algorithm. The algorithm identifies the target's center pixel in the image, and the instrument is commanded to aim at that location. We evaluated the performance of this template-matching-based centering method under various distances, angles, lighting conditions, and field environments, comparing its results to both manual aiming and conventional automated aiming of tape targets. Manual aiming was used as the baseline operational reference in the absence of an independent ground-truth measurement. The proposed approach achieves target centering results comparable to the manual baseline. This method provides a cost-effective alternative for high-precision applications where retroreflector targets are constrained by budget or logistics.
Harrington, Joshua [ORNL]↗
Self-Supervised Anomaly Detection via Neural Autoregressive Flows with Active Learning
Many self-supervised methods have been proposed with the target of image anomaly detection. These methods often rely on the paradigm of data augmentation with predefined transformations such as flipping, cropping, and rotations. However, it is not straightforward to apply these techniques for non-image data, such as time series or tabular data, while the performance of the existing deep approaches has been under our expectation on tasks beyond images. In this work, we propose a novel active learning (AL) scheme that relied on neural autoregressive flows (NAF) for self-supervised anomaly detection, specifically on small-scale data. Unlike other generative models such as GANs or VAEs, flow-based models allow to explicitly learn the probability density and thus can assign accurate likelihoods to normal data which makes it usable to detect anomalies. The proposed NAF-AL method is achieved by efficiently generating random samples from latent space and transforming them into feature space along with likelihoods via invertible mapping. The samples with lower likelihoods are selected and further checked by outlier detection using Mahalanobis distance. The augmented samples incorporating with normal samples are used for training a better detector so as to approach decision boundaries. Compared with random transformations, NAF-AL can be interpreted as a likelihood-oriented data augmentation that is more efficient and robust. Extensive experiments show that our approach outperforms existing baselines on multiple time series and tabular datasets, and a real-world application in advanced manufacturing, with significant improvement on anomaly detection accuracy and robustness over the state-of-the-art.
Zhang, Jiaxin↗
A unified large language model–based framework for heterogeneous PV image diagnosis
With advances in imaging technologies, modern photovoltaic (PV) systems generate large volumes of heterogeneous image data, including visible, electroluminescence (EL), and infrared (IR) images. Existing PV image analysis models, particularly deep learning approaches, are typically task-specific and lack cross-modality generalization. To address this limitation, this paper proposes an open-source large language model (LLM)–based unified framework for heterogeneous PV image diagnostics. Through task-aware diagnostic prompting, the framework enables analysis of visible, EL, and IR images within a single pipeline, supporting both zero-shot and few-shot inference and binary and multiclass classification. It is compatible with state-of-the-art multimodal LLMs, including ChatGPT, Gemini, Claude, Qwen, and CLIP. The framework is evaluated on PV module condition classification (clean, soiling, snow, hail, and bird droppings) using visible images, cell crack detection using EL images, and hotspot detection using IR images. GPT-5.1 in few-shot mode achieves the best performance, with classification accuracy exceeding 97.3%. Open-source models such as Qwen and CLIP also deliver competitive results on visible images (around 90% accuracy), though their performance is more limited on EL and IR modalities. On the full ELPV dataset, the framework achieves 83.5% zero-shot accuracy, within 2.8% of the supervised CNN baseline, confirming scalability to larger benchmarks. Practical aspects such as reproducibility, response latency, and confidence estimation are systematically analyzed. The framework operates across PV image modalities without modality- or task-specific training, making it well suited as a rapid pre-screening tool to support downstream detailed diagnostics. A benchmark dataset of diverse labeled PV images is also released.
Li, Baojie↗