Search NASA⌕ Search

SEARCH · Search NASA

Results for “Convolution”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,189 records · Page 66

Identifying common stored product insects using automated deep learning methods

Monitoring stored product insect pests is a common practice for post-harvest management of stored grain and grain-based commodities, which helps ensure product quality from harvest to final consumer. Current methods of sampling and monitoring can be time-consuming, labor-intensive, expensive and require expertise in insect identification. Therefore, this study aims to develop an image-based automated identification system for common stored product insect species using deep-learning methods. Top-down images of the common stored product adult insect species of Rhyzopertha dominica, Cryptolestes ferrugineus, Tribolium castaneum, Sitophilus oryzae, and Oryzaephilus surinamensis were acquired and analyzed. Deep learning-based, state-of-the-art Convolutional Neural Networks (CNN) models (ResNet-50, MobileNet-v2, DarkNet-53, and EfficientNet-b0) were fine-tuned with a transfer learning approach to classify the insect species. All models were able to correctly identify the insect species with at least 96% accuracy and with few misclassifications. One issue with trained CNNs is that they do not explain the reasoning for the classification and are often called a “black box”. Therefore, visualization methods called Gradient-weighted Class Activation Mapping (Grad-CAM) were implemented to explore the black box network. The Grad-CAM uses heat maps to highlight the major image features that the network focused on to make insect species predictions. The Grad-CAM verifies the network's prediction and also helps improve network performance. This study contributes to the overall goal of developing a camera-based system for monitoring stored grain insects. As a result, the developed system would empower warehouse, flour mills, and other food facilities with a tool to quickly and accurately identify insect species in stored product environments and could be implemented as part of a close to real-time monitoring system.

60 APPLIED LIFE SCIENCES↗

Grain structure, texture and mechanical property evolution of automotive aluminium sheet during high power ultrasonic welding

High power ultrasonic spot welding (HPUSW) is a joining technique which is performed within less than a second and provides a more energy-efficient alternative to friction stir spot welding (FSSW), which is considered a longer cycle manufacturing process for joining automotive alloys. To date, only a few reports exist on the deformation mechanisms that take place during high power ultrasonic spot welding. In this work, dynamic recrystallization and grain growth were examined using electron backscatter diffraction (EBSD). HPUSW causes extensive deformation within the weld zone where the temperature increases to 440 °C. An ultra-fine grain structure was observed in a thin band of flat weld interface within a short welding time of 0.10 s. With increasing welding time the interface was displaced and ‘folds’ or ‘crests’ appeared together with shear bands. The weld interface progressively changed from flat to sinusoidal and eventually to a convoluted wave-like pattern when the tool fully penetrated the workpiece, having a wavelength of ~ 1 mm after 0.40 s. Finally, the microstructure and texture varied significantly depending on the location within the weld. Although the texture near the weld interface was relatively weak, a shift was observed with increasing welding time from an initially Cube-dominated texture to one where the typical β-fibre Brass component prevailed. - Highlights: •Lap shear strength of ~2.9 kN was achieved in 0.30 sec welding time. •Temperature approached 440 °C along the weld centreline for the highest welding time. •The texture near the teeth was dominated by Brass, P and S components at optimum condition. •The weld interface showed typical β-fibre deformation texture at optimum condition.

36 MATERIALS SCIENCE↗

Characterizing dislocation configurations and their evolution during creep of a new 12% Cr steel

Highlights: • Applicability of different techniques to characterize particular dislocation configuration. • XRD models in combination with TEM and EBSD to measure the dislocation density. • Substructure evolution during creep of a new 12% Cr steel is discussed. - Abstract: In this research, different types of dislocations are characterized and quantified in a newly developed 12% Cr steel in as-received as well as in crept conditions. X-ray diffraction (XRD) patterns are analyzed using three different models. The dislocation densities are determined employing the Convolution Multiple Whole Profile (CMWP) fitting, the analysis of the asymptotic part of the X-ray line profile (Groma's method), and the Rietveld refinement coupling Popa's and Williamson-Smallman approaches. CMWP method is used to measure the internal dislocation density inside the subgrains consisting of mobile and dipole dislocations, and the results show good agreement with Transmission Electron Microscopy (TEM) measurements. Groma's method is used to evaluate the total dislocation density, with results comparable to those obtained from the combination of Electron Backscatter Diffraction (EBSD) and TEM. Rietveld method gives the internal dislocation density or mobile dislocation density depending on the model chosen. The experimentally determined results such as creep strain, dislocation densities and subgrain size are compared with theoretical predictions and they are found in good agreement.

36 MATERIALS SCIENCE↗

Austenitic parent grain reconstruction in martensitic steel using deep learning

In this work we develop a deep convolutional architecture to estimate the prior austenite structure from observed martensite electron backscatter diffraction micrographs. A novel data augmentation strategy randomizes the global reference coordinate system which makes it possible to train our model from only four micrographs. The model is much faster than algorithmic approaches and generalizes well when applied to micrographs of a different material. Empirical evidence suggests the efficacy of the model depends on the scale of the microstructure and receptive field of the vision model. Furthermore, this work demonstrates that modern computer vision approaches are well suited for capturing complex spatial-orientation patterns present in orientation imaging micrographs.

36 MATERIALS SCIENCE↗

Multi phenomena melt pool sensor data fusion for enhanced process monitoring of laser powder bed fusion additive manufacturing

Finding actionable trends in laser-based metal additive manufacturing process monitoring data is challenging owing to the diversity and complexity of the underlying physical interactions. A single monitoring solution that captures a particular process phenomenon, such as a photodiode that tracks melt pool intensity, is not alone capable of evaluating process stability or detecting flaw formation with sufficient precision for routine application in industry. In this work, to improve flaw detection performance, we adopted a data fusion approach that captures multiple process phenomena. To demonstrate this, we acquired data from laser powder bed fusion (LPBF) builds of cylindrical specimens produced with different laser spot sizes, emulating defocusing due to process faults such as thermal lensing. The resulting specimens had porosity of varying types and severity, quantified by post-build non-destructive X-ray computed tomography, Archimedes density measurements, and destructive metallographic characterization. During the build, the melt pool state was monitored with two coaxial high-speed video cameras and a temperature field imaging system. Physically intuitive low-level melt pool signatures, such as melt pool temperature, shape and size, and spatter intensity were extracted from this high-dimensional, image-based sensor data. These process signatures were subsequently used as input features in relatively simple machine learning models, such as a support vector machine, which were trained to detect laser defocusing, and in addition, predict porosity type and severity. The results show that the data fusion approach significantly enhanced system performance by reducing the overall false positive rate from ~ 0.1 to ~ 0.001 without sacrificing the true positive rate (~0.90). These results were at par with a black-box, deep machine learning approach (convolutional neural network).

36 MATERIALS SCIENCE↗

Graph neural networks for mechanical property prediction of 2D fiber composites

This work investigates the ability of graph neural networks (GNNs) to homogenize 2D fiber composite microstructures. We use different inhomogeneity and anisotropy indices to motivate and show that the Volume Elements (VEs) used in ML methods should ideally be far from their Representative Volume Element (RVE) size limit and, consequently, are notably anisotropic. Hence, training only the isotropic limit properties may not be acceptable. Another aspect is the need to normalize elastic stiffness values for ML, especially when high elastic contrast ratios are encountered between composite phases or in the material set. We introduce a normalization technique based on the mean-field method (MFM) to handle such high contrast ratios and train for the entire stiffness tensor. We show that the proposed GNN approaches exhibit high accuracy and efficiency compared to traditional methods and convolutional neural networks, utilizing unstructured graphs constructed from microstructure topology. Our model successfully predicts the stiffness tensor, peak strength under bulk damage, and brittle fracture initiation strength across diverse microstructure configurations while maintaining high accuracy even for extreme material contrasts and volume fractions. We also present a method to improve prediction accuracy for small dataset sizes using Voronoi partitioning.

Brittle strength↗

Super-resolution model for overlapping peak detection and improved spatial resolution in high-energy diffraction microscopy

Reconstruction quality in Far-field High-Energy Diffraction Microscopy (FF-HEDM) is limited by the spatial resolution of area detectors and the frequent occurrence of overlapping diffraction spots. To address these challenges, we developed a super-resolution (SR) framework using convolutional neural networks (CNNs) to recreate 2D diffraction peaks at up to ×8 resolution from raw detector data. A specialized simulation tool was created to generate synthetic training datasets with varying degrees of peak overlap. Integrated into the Microstructural Imaging using Diffraction Analysis Software (MIDAS), the SR model improves the spatial accuracy and precision of 3D grain reconstruction by an order of magnitude. This approach provides a robust solution for investigation of complex micromechanical states and material classes where the analysis is limited by the presence of overlapping peaks. Furthermore, the methodology developed here can potentially be extended to other techniques that require sub-pixel accuracy for high-fidelity data analysis.

High-energy diffraction microscopy↗

In situ neutron diffraction analyses of temperature and stresses during friction stir processing of Mg-3Al-1Zn magnesium alloy

In this paper, the evolution of temperature and thermal stresses were examined during friction stir processing of a magnesium alloy using in situ neutron diffraction. Measured total lattice strains were de-convoluted into the thermal and stress components. Finally, it shows the maximum temperature of 464 °C and compressive stress distributions up to -52 MPa underneath the tool via the thermo-mechanical deformation in terms of the Zener-Hollomon parameter of 4.51 × 10 10 s -1 .

36 MATERIALS SCIENCE↗

Resolving three-dimensional nanoscale heterogeneities in lithium metal batteries with cryoelectron tomography

Current direct observation of sensitive battery materials and interfaces primarily relies on two-dimensional (2D) imaging, leaving out their three-dimensional (3D) relationship. Here, in this study, we used cryoelectron tomography (cryo-ET) to visualize the lithium metal anode in 3D at nanometer resolution and cryoelectron microscopy (cryo-EM) to reveal atomic details in local regions. We imaged both freshly prepared and calendar-aged Li metal anodes to reveal the development of LiH in Li dendrites and the Li-LiH interface, as well as the development of the solid-electrolyte interphase (SEI). Using a convolutional neural network-based technique, the 3D arrangement of Li metal, along with nanoscale LiH and Cu heterogeneities in dendrites, was visualized and annotated. In longer-term calendar aging, we observed more substantial LiH growth accompanied by extended SEI growth. Our results show that the growth of LiH and the extended SEI during battery calendar aging are temporally and spatially separate processes.

LiH↗

Extracting physics through deep data analysis

In recent decades humankind has become very apt at generating and recording enormous amounts of data, ranging from tweets and selfies on social networks, to financial transactions in banks and stores. The scientific community has not shunned this popular trend and now routinely produces hundreds of petabytes of data per year [1]. This is because materials and phenomena in the world around us exist in an interweaved, entangled form, which gives rise to the complexity of the Universe and determines the size and complexity of the data that describes it. Science and technology endeavor to unravel this convolution and extract pure components from the mixtures, be it in ore mining and metal smelting or separation of thermal conductivity into the electronic and phononic contributions. Decomposition of complex behavior is the key to understanding manifestations of Nature. However, tools to carry out this task are not readily available, and therefore, intricate systems often remain well-characterized experimentally, but still not well understood due to intricacy of the collected data. Lastly, in materials science, understanding and ultimately designing new materials with complex properties will require the ability to integrate and analyze data from multiple instruments, including computational models, designed to probe complementary ranges of space, time, and energy.

36 MATERIALS SCIENCE↗

Impurity gas monitoring using ultrasonic sensing and neural networks: forward and inverse problems

Ultrasonic sensing is a non-invasive technique for monitoring impurity gas composition in various industrial applications where safety and regulatory compliance are crucial. In this study, ultrasonic sensing and neural networks were used to analyze impurity gases (i.e., air and argon) in helium. An experimental platform was established to acquire ultrasonic data. In the forward problem, an artificial neural network (ANN) model was used to forecast the response and time-of-flight (TOF) based on the excitation, and argon and air concentrations. The inverse problem was solved using a convolutional neural network (CNN) to predict the argon and air concentrations given the ultrasonic response and excitation. The results showed that the ANN accurately predicted the ultrasonic response and the change in TOF with concentration. As the air concentration was increased from 0 to 9.8%, the TOF sensitivity to detect argon decreased by 39.8% and 16.1% from ANN and sound speed theory, respectively. The CNN demonstrated high accuracy in predicting concentrations for inputs in the testing dataset. The application of the trained CNN indicated that it over-predicts air concentration while under-predicting the argon concentration. To improve accuracy, the predicted air and argon concentrations should be corrected by -0.992% and 1.027% bias, respectively.

47 OTHER INSTRUMENTATION↗

A machine-learning approach to measure 3D sample properties from 2D Transmission Electron Microscopy images

Transmission Electron Microscopy (TEM) is a powerful tool for the characterization of materials at the nanoscale; however, its inherent two-dimensional (2D) nature poses significant challenges to accurately measure three-dimensional (3D) properties. We introduce a supervised machine-learning model that predicts 3D structural information, such as sample thickness and curvature, from a series of conventional 2D TEM images. The model, a U-Net convolutional neural network, is trained on a large synthetic dataset generated from dynamical diffraction simulations that model TEM’s complex, nonlinear image formation, accounting for sample thickness and curvature. This physically realistic framework enables exploration of a broad parameter space impractical to sample experimentally. We demonstrate that the trained model has accurate predictions for experimental single-crystal silicon samples, achieving performance comparable to established measurement techniques. This work highlights the critical role of robust, simulation-based training in overcoming the limitations of real-world imaging artifacts and inconsistent sample geometries. By integrating machine learning with numerical simulations, we offer an efficient and scalable framework for quantitative TEM analysis, paving the way for more sophisticated 3D characterization of complex materials.

Dynamical diffraction↗

Tissue Heterogeneity in IMRT Dose Calculation for Lung Cancer

The aim of this study was to evaluate the differences in accuracy of dose calculation between 3 commonly used algorithms, the Pencil Beam algorithm (PB), the Anisotropic Analytical Algorithm (AAA), and the Collapsed Cone Convolution Superposition (CCCS) for intensity-modulated radiation therapy (IMRT). The 2D dose distributions obtained with the 3 algorithms were compared on each CT slice pixel by pixel, using the MATLAB code (The MathWorks, Natick, MA) and the agreement was assessed with the {gamma} function. The effect of the differences on dose-volume histograms (DVHs), tumor control, and normal tissue complication probability (TCP and NTCP) were also evaluated, and its significance was quantified by using a nonparametric test. In general PB generates regions of over-dosage both in the lung and in the tumor area. These differences are not always in DVH of the lung, although the Wilcoxon test indicated significant differences in 2 of 4 patients. Disagreement in the lung region was also found when the {Gamma} analysis was performed. The effect on TCP is less important than for NTCP because of the slope of the curve at the level of the dose of interest. The effect of dose calculation inaccuracy is patient-dependent and strongly related to beam geometry and to the localization of the tumor. When multiple intensity-modulated beams are used, the effect of the presence of the heterogeneity on dose distribution may not always be easily predictable.

61 RADIATION PROTECTION AND DOSIMETRY↗

The Accuracy of Inhomogeneity Corrections in Intensity Modulated Radiation Therapy Planning in Philips Pinnacle System

The degree of accuracy of inhomogeneity corrections in a treatment planning system is dependent on the algorithm used by the system. The choice of field size, however, could have an effect on the calculation accuracy as well. There have been several evaluation studies on the accuracy of inhomogeneity corrections used by different algorithms. Most of these studies, however, focus on evaluating the dose in phantom using simplified geometry and open/static fields. This work focuses on evaluating the degree of dose accuracy in calculations involving intensity-modulated radiation therapy (IMRT) fields incident on a phantom containing both lung- and bone-equivalent heterogeneities using 6 and 10 MV beams. IMRT treatment plans were generated using the Philips Pinnacle treatment planning system and delivered to a phantom containing 55 thermoluminescent dosimeter (TLD) locations within the lung and bone and near the lung and bone interfaces with solid water. The TLD readings were compared with the dose predicted by the planning system. We find satisfactory agreement between planned and delivered doses, with an overall absolute average difference between measurement and calculation of 1.2% for the 6 MV and 3.1% for the 10 MV beam with larger variations observed near the interfaces and in areas of high-dose gradient. The results presented here demonstrate that the convolution algorithm used in the Pinnacle treatment planning system produces accurate results in IMRT plans calculated and delivered to inhomogeneous media, even in regions that potentially lack electronic equilibrium.

61 RADIATION PROTECTION AND DOSIMETRY↗

Determination of peripheral underdosage at the lung-tumor interface using Monte Carlo radiation transport calculations

Prediction of dose distributions in close proximity to interfaces is difficult. In the context of radiotherapy of lung tumors, this may affect the minimum dose received by lesions and is particularly important when prescribing dose to covering isodoses. The objective of this work is to quantify underdosage in key regions around a hypothetical target using Monte Carlo dose calculation methods, and to develop a factor for clinical estimation of such underdosage. A systematic set of calculations are undertaken using 2 Monte Carlo radiation transport codes (EGSnrc and GEANT4). Discrepancies in dose are determined for a number of parameters, including beam energy, tumor size, field size, and distance from chest wall. Calculations were performed for 1-mm{sup 3} regions at proximal, distal, and lateral aspects of a spherical tumor, determined for a 6-MV and a 15-MV photon beam. The simulations indicate regions of tumor underdose at the tumor-lung interface. Results are presented as ratios of the dose at key peripheral regions to the dose at the center of the tumor, a point at which the treatment planning system (TPS) predicts the dose more reliably. Comparison with TPS data (pencil-beam convolution) indicates such underdosage would not have been predicted accurately in the clinic. We define a dose reduction factor (DRF) as the average of the dose in the periphery in the 6 cardinal directions divided by the central dose in the target, the mean of which is 0.97 and 0.95 for a 6-MV and 15-MV beam, respectively. The DRF can assist clinicians in the estimation of the magnitude of potential discrepancies between prescribed and delivered dose distributions as a function of tumor size and location. Calculation for a systematic set of 'generic' tumors allows application to many classes of patient case, and is particularly useful for interpreting clinical trial data.

61 RADIATION PROTECTION AND DOSIMETRY↗

Correlation of dosimetric parameters obtained with the analytical anisotropic algorithm and toxicity of chest chemoradiation in lung carcinoma

The purpose of this study was to analyze and revisit toxicity related to chest chemoradiotherapy and to correlate these side effects with dosimetric parameters obtained using analytical anisotropic algorithm (AAA) in locally unresectable advanced lung cancer. We retrospectively analyzed data from 47 lung cancer patients between 2005 and 2008. All received conformal 3D radiotherapy using high-energy linear accelerator plus concomitant chemotherapy. All treatment planning data were transferred into Eclipse 8.05 (Varian Medical Systems, Palo Alto, CA) and dosimetric calculations were performed using AAA. Thirty-three patients (70.2%) developed acute pneumopathy after radiotherapy (grades 1 and 2). One patient (2.1%) presented with grade 3 pneumopathy. Thirty-one (66%) presented with grades 1-2 lung fibrosis, and 1 patient presented with grade 3 lung fibrosis. Thirty-four patients (72.3%) developed grade 1-2 acute oesophagic toxicity. Four patients (8.5%) presented with grades 3 and 4 dysphagia, necessitating prolonged parenteral nutrition. Median prescribed dose was 64 Gy (range 50-74) with conventional fractionation (2 Gy per fraction). Dose-volume constraints were respected with a median V20 of 23.5% (maximum 34%) and a median V30 of 17% (maximum 25%). The median dose delivered to healthy contralateral lung was 13.1 Gy (maximum 18.1 Gy). At univariate analysis, larger planning target volume and V20 were significantly associated with the probability of grade {>=}2 radiation-induced pneumopathy (p = 0.022 and p = 0.017, respectively). No relation between oesophagic toxicity and clinical/dosimetric parameters could be established. Using AAA, the present results confirm the predictive value of the V20 for lung toxicity as already demonstrated with the conventional pencil beam convolution approach.

61 RADIATION PROTECTION AND DOSIMETRY↗

Quantitative assessment of the accuracy of dose calculation using pencil beam and Monte Carlo algorithms and requirements for clinical quality assurance

To compare the doses calculated using the BrainLAB pencil beam (PB) and Monte Carlo (MC) algorithms for tumors located in various sites including the lung and evaluate quality assurance procedures required for the verification of the accuracy of dose calculation. The dose-calculation accuracy of PB and MC was also assessed quantitatively with measurement using ionization chamber and Gafchromic films placed in solid water and heterogeneous phantoms. The dose was calculated using PB convolution and MC algorithms in the iPlan treatment planning system from BrainLAB. The dose calculation was performed on the patient's computed tomography images with lesions in various treatment sites including 5 lungs, 5 prostates, 4 brains, 2 head and necks, and 2 paraspinal tissues. A combination of conventional, conformal, and intensity-modulated radiation therapy plans was used in dose calculation. The leaf sequence from intensity-modulated radiation therapy plans or beam shapes from conformal plans and monitor units and other planning parameters calculated by the PB were identical for calculating dose with MC. Heterogeneity correction was considered in both PB and MC dose calculations. Dose-volume parameters such as V95 (volume covered by 95% of prescription dose), dose distributions, and gamma analysis were used to evaluate the calculated dose by PB and MC. The measured doses by ionization chamber and EBT GAFCHROMIC film in solid water and heterogeneous phantoms were used to quantitatively asses the accuracy of dose calculated by PB and MC. The dose-volume histograms and dose distributions calculated by PB and MC in the brain, prostate, paraspinal, and head and neck were in good agreement with one another (within 5%) and provided acceptable planning target volume coverage. However, dose distributions of the patients with lung cancer had large discrepancies. For a plan optimized with PB, the dose coverage was shown as clinically acceptable, whereas in reality, the MC showed a systematic lack of dose coverage. The dose calculated by PB for lung tumors was overestimated by up to 40%. An interesting feature that was observed is that despite large discrepancies in dose-volume histogram coverage of the planning target volume between PB and MC, the point doses at the isocenter (center of the lesions) calculated by both algorithms were within 7% even for lung cases. The dose distributions measured with EBT GAFCHROMIC films in heterogeneous phantoms showed large discrepancies of nearly 15% lower than PB at interfaces between heterogeneous media, where these lower doses measured by the film were in agreement with those by MC. The doses (V95) calculated by MC and PB agreed within 5% for treatment sites with small tissue heterogeneities such as the prostate, brain, head and neck, and paraspinal tumors. Considerable discrepancies, up to 40%, were observed in the dose-volume coverage between MC and PB in lung tumors, which may affect clinical outcomes. The discrepancies between MC and PB increased for 15 MV compared with 6 MV indicating the importance of implementation of accurate clinical treatment planning such as MC. The comparison of point doses is not representative of the discrepancies in dose coverage and might be misleading in evaluating the accuracy of dose calculation between PB and MC. Thus, the clinical quality assurance procedures required to verify the accuracy of dose calculation using PB and MC need to consider measurements of 2- and 3-dimensional dose distributions rather than a single point measurement using heterogeneous phantoms instead of homogenous water-equivalent phantoms.

61 RADIATION PROTECTION AND DOSIMETRY↗

Experimental and Monte Carlo evaluation of Eclipse treatment planning system for effects on dose distribution of the hip prostheses

The present study aimed to investigate the effects of titanium, titanium alloy, and stainless steel hip prostheses on dose distribution based on the Monte Carlo simulation method, as well as the accuracy of the Eclipse treatment planning system (TPS) at 6 and 18 MV photon energies. In the present study the pencil beam convolution (PBC) method implemented in the Eclipse TPS was compared to the Monte Carlo method and ionization chamber measurements. The present findings show that if high-Z material is used in prosthesis, large dose changes can occur due to scattering. The variance in dose observed in the present study was dependent on material type, density, and atomic number, as well as photon energy; as photon energy increased back scattering decreased. The dose perturbation effect of hip prostheses was significant and could not be predicted accurately by the PBC method for hip prostheses. The findings show that for accurate dose calculation the Monte Carlo-based TPS should be used in patients with hip prostheses.

61 RADIATION PROTECTION AND DOSIMETRY↗