SEARCH · Search NASA

Results for “multiple input images”

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.

SourceResearch connectionFrom yearThrough year

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

At least 55 records · Page 3

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Solving Seismic Wave Equations on Variable Velocity Models With Fourier Neural Operator

Here, in the study of subsurface seismic imaging, solving the acoustic wave equation is a pivotal component in existing models. The advancement of deep learning (DL) enables solving partial differential equations (PDEs), including wave equations, by applying neural networks to identify the mapping between the inputs and the solution. This approach can be faster than traditional numerical methods when numerous instances are to be solved. Previous works that concentrate on solving the wave equation by neural networks consider either a single velocity model or multiple simple velocity models, which is restricted in practice. Instead, inspired by the idea of operator learning, this work leverages the Fourier neural operator (FNO) to effectively learn the frequency domain seismic wavefields under the context of variable velocity models. We also propose a new framework paralleled FNO (PFNO) for efficiently training the FNO-based solver given multiple source locations and frequencies. Numerical experiments demonstrate the high accuracy of both FNO and PFNO with complicated velocity models in the OpenFWI datasets. Furthermore, the cross-dataset generalization test verifies that PFNO adapts to out-of-distribution velocity models. Finally, PFNO admits higher computational efficiency on large-scale testing datasets than the traditional finite-difference method. The aforementioned advantages endow the FNO-based solver with the potential to build powerful models for research on seismic waves.

58 GEOSCIENCES↗

DOE OSTI2026

Surface and Buried Thermal, and RGB Unexploded Ordnance Data Collection

This document provides a description of a data collection campaign of unexploded ordnance (UXOI) set. The dataset captures a controlled UAV imaging campaign designed to support detection of UXO across varied environmental conditions. Data were collected during three campaigns in Norris and Northeast Knoxville, Tennessee, using RGB, and thermal sensors mounted on Parrot UKR. In total, the dataset contains 9925 images, 26 full-motion video, and approximately 81.99 GB of data, collected across late spring/summer conditions, every hour during sunlight, and multiple surface contexts, including tall grass, short grass, gravel, as well as buried in sand, and other gravel mixtures. The collection was designed to capture thermal and visual variability relevant to UXO detection in agricultural land, bare earth, and subsurface. Review of the imagery showed that ordnance was most detectable during periods of changing solar input, especially approximately 10-60 minutes after sunrise, approximately 20-60 minutes after sunset, and 2-3 min after cloud cover interrupted prolonged solar heating. These conditions increased thermal contrast because many ordnance items retained or released heat differently than the surrounding vegetation and ground surface. This dataset provides a useful resource for developing and evaluating airborne UXO detection methods under realistic field conditions. All ordnance used in the study was inert, and thermal behavior may differ from that of live ordnance. In addition, variation in ordnance type, composition, and placement introduced differences in thermal response that should be considered when interpreting results.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

DOE OSTI2021

Reconstructing magnetic deflections from sets of proton images using differential evolution

Proton imaging is a powerful technique for imaging electromagnetic fields within an experimental volume, in which spatial variations in proton fluence are a result of deflections to proton trajectories due to interaction with the fields. When deflections are large, proton trajectories can overlap, and this nonlinearity creates regions of greatly increased proton fluence on the image, known as caustics. The formation of caustics has been a persistent barrier to reconstructing the underlying fields from proton images. We have developed a new method for reconstructing the path-integrated magnetic fields, which begins to address the problem posed by caustics. Our method uses multiple proton images of the same object, each image at a different energy, to fill in the information gaps and provide some uniqueness when reconstructing caustic features. We use a differential evolution algorithm to iteratively estimate the underlying deflection function, which accurately reproduces the observed proton fluence at multiple proton energies simultaneously. We test this reconstruction method using synthetic proton images generated for three different, cylindrically symmetric field geometries at various field amplitudes and levels of proton statistics and present reconstruction results from a set of experimental images. Here, the method we propose requires no assumption of deflection linearity and can reliably solve for fields underlying linear, nonlinear, and caustic proton image features for the selected geometries and is shown to be fairly robust to noise in the input proton intensity.

47 OTHER INSTRUMENTATION↗

DOE OSTI2021

Low-Cost Radiant Heater for Rapid Response, High-Temperature Heating

High-temperature processing has an irreplaceable role in many research and industrial applications. Despite remarkable development spanning over a century, the pursuit of even higher thermal flux density and more rapid thermal transients has not slowed down. As part of the ongoing energy evolution, many industrial applications are transitioning from direct combustion of fossil fuels as primary energy sources to increasing electrification, capable of adapting to renewable power grids. Thus, there is an emerging need for electrical heaters that can replace burners and supply the heat demand, especially at the highest temperatures. In this study, we report on a radiant heater design that can achieve cyclic heating/cooling rates of up to 400 K min –1 and a temperature range in excess of 1,800 K, comparable to those of commercial infrared gold image furnaces, at high surface and volumetric power densities. The heater consists of a modular unit of incandescent tungsten filament and is enclosed in an evacuated ceramic envelope, chemically inert, tolerant of thermal shock, and impervious to gasses. The material and manufacture cost of such heaters, which is estimated at ~$0.05/W, is less than 0.03% of that for infrared gold image furnaces, which is at >$2/W. Tests of more than 10,000 demanding cycles (high temperature and high heating/cooling rate) over 350 h of total operational time and in different temperature ranges confirm the robust performance of radiant heater prototypes. The design is widely applicable to high-temperature reactor and furnace designs. In thermochemistry research and practice, these radiant heaters could offer multiple benefits compared to solar simulators, lasers, infrared gold furnaces, ceramic heaters, or direct concentration of solar input.

99 GENERAL AND MISCELLANEOUS↗

DOE OSTI2022

Applying Infrared Thermography as a Method for Online Monitoring of Turbine Blade Coolant Flow

As gas turbine engine manufacturers strive to implement condition-based operation and maintenance, there is a need for blade monitoring strategies capable of early fault detection and root-cause determination. Given the importance of blade cooling flows to turbine blade health and longevity, there is a distinct lack of methodologies for coolant flowrate monitoring. The present study addresses this identified opportunity by applying an infrared thermography system on an engine-representative research turbine to generate data-driven models for prediction of blade coolant flowrate. Thermal images were used as inputs to a linear regression and regularization algorithm to relate blade surface temperature distribution with blade coolant flowrate. Additionally, this study investigates how coolant flowrate prediction accuracy is influenced by the number and breadth of diagnostic measurements. Here, the results of this study indicate that a source of high-fidelity training data can be used to predict blade coolant flowrate within about six percent error. Furthermore, identification of prioritized sensor placement supports application of this technique across multiple sensor technologies capable of measuring blade surface temperature in operating gas turbine engines, including spatially resolved and point-based measurement techniques.

42 ENGINEERING↗

DOE OSTI2021

Dark Energy Survey Y3 results: blending shear and redshift biases in image simulations

ABSTRACT As the statistical power of galaxy weak lensing reaches per cent level precision, large, realistic, and robust simulations are required to calibrate observational systematics, especially given the increased importance of object blending as survey depths increase. To capture the coupled effects of blending in both shear and photometric redshift calibration, we define the effective redshift distribution for lensing, nγ(z), and describe how to estimate it using image simulations. We use an extensive suite of tailored image simulations to characterize the performance of the shear estimation pipeline applied to the Dark Energy Survey (DES) Year 3 data set. We describe the multiband, multi-epoch simulations, and demonstrate their high level of realism through comparisons to the real DES data. We isolate the effects that generate shear calibration biases by running variations on our fiducial simulation, and find that blending-related effects are the dominant contribution to the mean multiplicative bias of approximately $-2{{\ \rm per\ cent}}$. By generating simulations with input shear signals that vary with redshift, we calibrate biases in our estimation of the effective redshift distribution, and demonstrate the importance of this approach when blending is present. We provide corrected effective redshift distributions that incorporate statistical and systematic uncertainties, ready for use in DES Year 3 weak lensing analyses.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2023

Gumby: Quantifying multi-modal model resiliency

With the rise of cheap data and sensors, more use cases are emerging for multi-input models. Research has shown that including multiple data modalities can improve performance, suggesting that deep learning models can successfully learn to leverage complementary information from different modalities. However, this improved predictive power comes with unanticipated costs: additional inputs change model resiliency and expand the threat space for adversarial attacks. We first provide theoretical underpinnings for how adversarial success scales with input dimension. We then characterize the performance of a suite of multispectral deep learning models with different fusion approaches, quantify their relative reliance on different input bands, and evaluate their robustness to naturalistic and adversarial image corruptions.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Comparative Assessment of U-Net-Based Deep Learning Models for Segmenting Microfractures and Pore Spaces in Digital Rocks

Segmentation of high-resolution X-ray microcomputed tomography (µCT) images is crucial in digital rock physics (DRP), affecting the characterization and analysis of microscale phenomena in the porous media. The complexity of geological structures and nonideal scanning conditions pose significant challenges to conventional image segmentation approaches. Motivated by the recent increasing popularity of deep learning (DL) techniques in image processing, this work undertakes a comparative study of DL models, specifically U-Net and its variants, for segmenting multiple targets with distinguished features in digital rocks, including discrete fracture networks (DFNs), pore spaces, and solid rock. Particularly, DFNs have a smaller volumetric fraction over others, bringing in a substantial challenge of imbalanced segmentation. The primary focus is to evaluate the architecture and feature enhancement strategies of various DL models, including U-Net, attention U-Net, residual U-Net, U-Net++, and residual U-Net++. The models were designed as 2.5D, utilizing a central 2D image and its two adjacent upper and lower 2D images as input to provide a pseudo-3D context. In addition, because the ground truth of segmentation was unknown for real-world digital rocks, we created a benchmark data set following the inverse operations of segmentation. The data synthesis started from the label images (i.e., solid rock, pore spaces, and DFNs), followed by simulating partial volume blurring, adding random background noise, and introducing ring artifacts to mimic real raw X-ray µCT images. The data set, which included various rock types (i.e., sandstone and artificial data), scanning resolution, and magnitudes of noise and artifacts, was divided into training and testing data sets with a 90% and 10% ratio, respectively. Moreover, in addition to the conventional pixel-wise evaluation metrics, the physics-based metric of the lattice-Boltzmann method (LBM) simulated permeability provided more comprehensive assessments. The results demonstrated that the residual connections, nested architectures, and redesigned skip connections contribute to the model performance and give the residual U-Net++ the highest accuracy. The improvements were mainly on the boundaries and small targets, especially the DFNs, which dominate the interconnectivity and therefore affect the permeability greatly. This study also rigorously evaluated the efficiency and generalization of each model, demonstrating that the sophisticated architectures achieved excellent practicability and maintained robust performance on completely unseen data, ensuring their suitability for diverse and challenging DRP applications.

58 GEOSCIENCES↗

DOE OSTI2020

Sparse Matrix-Based HPC Tomography

Tomographic imaging has benefited from advances in X-ray sources, detectors and optics to enable novel observations in science, engineering and medicine. These advances have come with a dramatic increase of input data in the form of faster frame rates, larger fields of view or higher resolution, so high performance solutions are currently widely used for analysis. Tomographic instruments can vary significantly from one to another, including the hardware employed for reconstruction: from single CPU workstations to large scale hybrid CPU/GPU supercomputers. Furthermore, flexibility on the software interfaces and reconstruction engines are also highly valued to allow for easy development and prototyping. This paper presents a novel software framework for tomographic analysis that tackles all aforementioned requirements. The proposed solution capitalizes on the increased performance of sparse matrix-vector multiplication and exploits multi-CPU and GPU reconstruction over MPI. Furthermore, the solution is implemented in Python and relies on CuPy for fast GPU operators and CUDA kernel integration, and on SciPy for CPU sparse matrix computation. As opposed to previous tomography solutions that are tailor-made for specific use cases or hardware, the proposed software is designed to provide flexible, portable and high-performance operators that can be used for continuous integration at different production environments, but also for prototyping new experimental settings or for algorithmic development. The experimental results demonstrate how our implementation can even outperform state-of-the-art software packages used at advanced X-ray sources worldwide.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Characterizing natural fractures and sub-seismic faults for well completion of Marcellus shale in the MSEEL Consortium project, West Virginia, USA

The Middle Devonian Marcellus shale play has emerged as a major world-class hydrocarbon accumulation and represents one of the largest and most prolific shale plays in the world. According to many outcrop studies in the region, natural fractures are well developed in the Marcellus Shale. However, evaluating fractures in the subsurface is often a significant challenge due to a lack of sufficient data. Therefore, in the Marcellus Shale Energy and Environment Laboratory (MSEEL) consortium project, significant efforts have been made to acquire high-quality image logs in the Marcellus laterals. The project provided tremendous opportunities to characterize the natural fractures and sub-seismic faults and to evaluate their impact on well stimulation. In this study, about 70,000 ft of acquired high-resolution logging while drilling (LWD) acoustic images from five long laterals located in Monongalia County, West Virginia, were processed and interpreted. In addition, the study used high-quality micro-resistivity images from a pilot well, allowing the evaluation of natural fractures in the entire Marcellus vertical sequence. Based on the available acoustic images, the natural fractures were classified into three basic categories: high-amplitude fractures, low-amplitude fractures, and faults. Further, larger open fractures can also be determined when a low-amplitude fracture is evident on caliper images. The fractures in the Marcellus usually have a medium to high angle dip; however, multiple fracture sets in terms of strike orientation were clearly observed in all the laterals. The fracture set with a strike at NE-SW (or 60-240 deg) seems to be the predominant one in all the wells. A few other sets, including those with N-S, NWW-SEE, and E-W strikes, were also observed. Several sub-seismic faults, with mostly a low dip angle and a NE-SW strike, have also been seen in two of the laterals. The fracture density is variable across all the laterals, ranging from very low (or none) to very high (up to 5 fractures per ft). The average fracture density for all the laterals is about 1 fracture per 10 ft. In the vertical sequence, the natural fracture development showed a clear preference for shale or shaly facies over carbonate-rich or thin limestone layers. The interpreted fracture and fault data were used as input data for the stimulation design with the purpose of better understanding the fractures’ impact on well stimulation. Production data from the laterals were also used to evaluate the natural fractures’ influence on well performance. The quality image database and the consistent interpretation results for the entire project enabled a systematic approach to characterizing fractures and, more importantly, to evaluating the impact of fractures on well stimulation and production.

03 NATURAL GAS↗

DOE OSTI2023

A synthetic Roman Space Telescope High-Latitude Time-Domain Survey: supernovae in the deep field

ABSTRACT NASA will launch the Nancy Grace Roman Space Telescope (hereafter Roman) in the second half of this decade, which will allow for a generation-defining measurement of dark energy through multiple probes, including Type Ia supernovae (SNe Ia). To improve decisions on survey strategy, we have created the first simulations of realistic Roman images that include artificial SNe Ia injected as point sources in the images. Our analysis combines work done on Roman simulations for weak gravitational lensing studies as well as catalogue-level simulations of SN Ia samples. We have created a time series of images over 2 yr containing ∼1050 SNe Ia, covering a 1 deg2 subarea of a planned 5 deg2 deep survey. We have released these images publicly for community use along with input catalogues of all injected sources. We create secondary products from these images by generating coadded images and demonstrating recovery of transient sources using image subtraction. We perform first-use analyses on these images in order to measure galaxy detection efficiency, point source detection efficiency, and host-galaxy association biases. The simulated images can be found here at https://roman.ipac.caltech.edu/sims/SN_Survey_Image_sim.html.

Astronomy & Astrophysics↗

DOE OSTI2024

Efficient distributed continual learning for steering experiments in real-time

Deep learning has emerged as a powerful method for extracting valuable information from large volumes of data. However, when new training data arrives continuously (i.e., is not fully available from the beginning), incremental training suffers from catastrophic forgetting (i.e., new patterns are reinforced at the expense of previously acquired knowledge). Training from scratch each time new training data becomes available would result in extremely long training times and massive data accumulation. Rehearsal-based continual learning has shown promise for addressing the catastrophic forgetting challenge, but research to date has not addressed performance and scalability. To fill this gap, we propose an approach based on a distributed rehearsal buffer that efficiently complements data-parallel training on multiple GPUs to achieve high accuracy, short runtime, and scalability. It leverages a set of buffers (local to each GPU) and uses several asynchronous techniques for updating these local buffers in an embarrassingly parallel fashion, all while handling the communication overheads necessary to augment input minibatches using unbiased, global sampling. We further propose a generalization of rehearsal buffers to support both classification and generative learning tasks, as well as more advanced rehearsal strategies (notably Dark Experience Replay, leveraging knowledge distillation). We illustrate this approach with a real-life HPC streaming application from the domain of ptychographic image reconstruction. Furthermore, we run extensive experiments on up to 128 GPUs of the ThetaGPU supercomputer to compare our approach with baselines representative of training-from-scratch (the upper bound in terms of accuracy) and incremental training (the lower bound). Results show that rehearsal-based continual learning achieves a top-5 validation accuracy close to the upper bound, while simultaneously exhibiting a runtime close to the lower bound.

Asynchronous data management↗

DOE OSTI2021

Community RNA-Seq: multi-kingdom responses to living versus decaying roots in soil

Abstract Roots are a primary source of organic carbon input in most soils. The consumption of living and detrital root inputs involves multi-trophic processes and multiple kingdoms of microbial life, but typical microbial ecology studies focus on only one or two major lineages. We used Illumina shotgun RNA sequencing to conduct PCR-independent SSU rRNA community analysis (“community RNA-Seq”) and simultaneously assess the bacteria, archaea, fungi, and microfauna surrounding both living and decomposing roots of the annual grass, Avena fatua. Plants were grown in 13CO2-labeled microcosms amended with 15N-root litter to identify the preferences of rhizosphere organisms for root exudates (13C) versus decaying root biomass (15N) using NanoSIMS microarray imaging (Chip-SIP). When litter was available, rhizosphere and bulk soil had significantly more Amoebozoa, which are potentially important yet often overlooked top-down drivers of detritusphere community dynamics and nutrient cycling. Bulk soil containing litter was depleted in Actinobacteria but had significantly more Bacteroidetes and Proteobacteria. While Actinobacteria were abundant in the rhizosphere, Chip-SIP showed Actinobacteria preferentially incorporated litter relative to root exudates, indicating this group’s more prominent role in detritus elemental cycling in the rhizosphere. Our results emphasize that decomposition is a multi-trophic process involving complex interactions, and our methodology can be used to track the trajectory of carbon through multi-kingdom soil food webs.

Nuccio, Erin E. (ORCID:000000030189183X)↗

DOE OSTI2026

The 3D clustering of Lyman Alpha Emitters measured with DESI

We present a clustering analysis of Lyman-$α$ emitters (LAEs) using spectroscopic observations from the Dark Energy Spectroscopic Instrument (DESI) of candidates selected from the Blanco/DECam Intermediate-Band Imaging Survey (IBIS). We measure the two-point correlation function and the power spectrum, including cross-correlations with DESI quasars. Using both analytical and halo occupation distribution (HOD) simulation-based modeling, we find a linear bias of $b \sim 2.31$--$2.62$ for LAEs over the redshift range $2.26 < z < 3.41$. The analytical modeling also provides constraints on the strength of radiative transfer effects, while the HOD analysis characterizes the LAE-halo connection across multiple models. Finally, we quantify the magnitude of non-perturbative clustering effects such as Fingers of God in the LAE population, providing essential input for the accurate modeling of LAE-based cosmological analyses in forthcoming high-redshift surveys such as DESI-II.

Ebina, H. [UC, Berkeley; LBL, Berkeley] (ORCID:000↗

DOE OSTI2024

Super Resolving Unrolled Neural Networks for Remote Sensing

In remote sensing systems, the capabilities of the system are constrained by the complex interactions between size, weight, and power (SWAP) of potential designs. In electro-optical (EO) systems, examples of these critical parameters include the system’s sensitivity and resolution. Those parameters can be increased by ever larger optical apertures and focal planes but at the cost of more SWAP. Multi-image super resolution (MISR) techniques allow resolution to be enhanced via computation rather than more sophisticated optical hardware. These algorithms combine multiple images together into a single, higher resolution image, trading temporal resolution and computation for spatial resolution. Fielded MISR techniques, such as Drizzle, can require several hundred images to create a single super resolved image, implying reduced temporal resolution, increased data acquisition load, and limiting mission applications. Iterative techniques, such as model-based image reconstruction and compressive sensing, have been shown to create super resolved images using fewer images than Drizzle. They do this by posing an optimization problem that balances accuracy between a highly accurate physical model and an image model. In the case of super resolution, the physical model is defined by the relation between low resolution input images and the desired high resolution output image. The image model encodes some assumptions about the super resolved image. These assumptions are meant to suppress reconstruction artifacts that arise due to deterministic physical model error, stochastic measurement noise, and potential undersampling. In practice, the performance of iterative methods are limited by imaging models compatible with optimization. Deep learning-based methods can effectively learn image models of arbitrary complexity, but lack the theoretical explainability and robustness of iterative techniques. Consensus equilibrium (CE) generalizes the iterative techniques beyond optimization, enabling blackbox algorithms such as traditional and neural image denoisers to be used as the image model. CE-based approaches retain much of the explainability and robustness of iterative techniques while allowing the expressiveness of machine learning image models to be used. Additionally, by unrolling iterations of CE with an embedded image denoiser, the image denoiser can be further trained and specialized to the specific application with potentially higher quality reconstructions. Under this project, we demonstrated the feasibility of training an unrolled neural network based upon CE. While we didn’t train one, we showed that the CE process is differentiable and its gradient can be tractably computed. We also explored the usage of a variants of CE akin to generative neural works. Most importantly, we applied the CE framework to a number of problems including non-blind deconvolution, upsampling, single-image super resolution, MISR, event-based sensing, and saturated deconvolution. Our MISR prototype creates high quality reconstructions with an order of magnitude fewer images than previous approaches and, critically, produces these reconstructions fast enough for practical usage.

47 OTHER INSTRUMENTATION↗

DOE OSTI2022

A low-input strategy for chromium removal from industrial stormwater using peat sorbent

Low-cost and low-input water treatment systems are important for industrial stormwater remediation. Here we examine a flow-through reactor treatment installation where water exceeds the allowable maximum concentration for drinking water in multiple metals (e.g., chromium [Cr], cadmium [Cd], zinc [Zn]) prior to treatment. Specifically, we seek to understand why Cr attenuated in the reactors is not leachable by identifying the specific chemical form of Cr and dominant mechanisms promoting sequestration in the reactors. Total solid-phase Cr concentration in the peat media ranged from 50 to 150 mg/kg after 1 yr of exposure to stormwater to 300 to 900 mg/kg after 3.5 yr. X-ray fluorescence mapping images show Cr, iron (Fe), and Zn spatially correlated over a scale of 10 μm to 5 mm. Chromium rinds form on the edges of peat particles as Cr accumulates. Chromium and Fe K-edge X-ray absorption near edge structure spectroscopy reveal chromium predominately in the 3+ oxidation state with lesser amounts of elemental Cr. We propose the primary means of chromium attenuation in the reactors is precipitation as Cr-Fe hydroxides combined with trivalent Cr adsorption onto peat surfaces.

36 MATERIALS SCIENCE↗

DOE OSTI2022

Downscaling atmospheric chemistry simulations with physically consistent deep learning

Abstract. Recent advances in deep convolutional neural network (CNN)-based super resolution can be used to downscale atmospheric chemistry simulations with substantially higher accuracy than conventional downscaling methods. This work both demonstrates the downscaling capabilities of modern CNN-based single image super resolution and video super-resolution schemes and develops modifications to these schemes to ensure they are appropriate for use with physical science data. The CNN-based video super-resolution schemes in particular incur only 39 % to 54 % of the grid-cell-level error of interpolation schemes and generate outputs with extremely realistic small-scale variability based on multiple perceptual quality metrics while performing a large (8×10) increase in resolution in the spatial dimensions. Methods are introduced to strictly enforce physical conservation laws within CNNs, perform large and asymmetric resolution changes between common model grid resolutions, account for non-uniform grid-cell areas, super-resolve lognormally distributed datasets, and leverage additional inputs such as high-resolution climatologies and model state variables. High-resolution chemistry simulations are critical for modeling regional air quality and for understanding future climate, and CNN-based downscaling has the potential to generate these high-resolution simulations and ensembles at a fraction of the computational cost.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2022

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↗