Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “UT”

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 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
NASA NTRS2009

Five-Year (2004-2009)Observations of Upper Tropospheric Water Vapor and Cloud Ice from MLS and Comparisons with GEOS-5 Analyses

This paper gives an overview of August 2004 through July 2009 upper tropospheric (UT) water vapor (H2O) and ice water content (IWC) from the Aura Microwave Limb Sounder (MLS) and comparisons with outputs from the NASA Goddard Earth Observing System Version 5 (GEOS-5) data assimilation system. Both MLS and GEOS-5 show that high values of H2O and IWC at 215 to 147 hPa are associated with areas of deep convection. They exhibit good (within approximately 15%) agreement in IWC at these altitudes, but GEOS-5 H2O is approximately 50% (215 hPa) to approximately 30% (147 hPa) larger than MLS, possibility due to its higher temperatures at these altitudes. GOES-5 produces a weaker intertropical convergence zone than MLS, while a seasonally-migrating band of tropical deep convection is clearly evident in both the MLS and GEOS-5 UT H2O and IWC. MLS and GEOS-5 both show spatial anti-correlation between IWC and H2O at 100 hPa, where less H2O is associated with low temperatures in regions of tropical convection. At 100 hPa, GEOS-5 produces 50% less IWC and 15% less H2O in the tropics, and approximately 20% more H2O in the extra-tropics, than does MLS. Behavior of the 100 hPa H2O, which exhibits a quasi-biennial oscillation, appears consistent with it being controlled by temperature. The seasonal cycle in the vertical transport of tropical mean H2O from approximately 147 hPa to approximately 10 hPa appears much stronger in MLS than in GEOS-5. The UT IWC and H2O interannual variations, from both MLS and GEOS-5, show clear imprints of the El Nino-Southern Oscillation.

Jiang, Jonathan H.↗
NASA NTRS2013

Observations of Comet 2P/Encke During the Fall 2013 Apparition

We will present preliminary results from our observational campaign of Comet 2P/Encke during its 2013 perihelion passage. At optical wavelengths Encke is an extremely dust poor comet that has in past perihelion passages emitted a gas jet in the form a sunward fan. We expect to characterize both the morphology and lightcurve of the comet. The low optical dust means that even near perihelion the nuclear signature can be obtained in lightcurve data taken with narrowband continuum filters which cut out the gas emission. The campaign will consist of both narrowband and broadband imaging as well as infrared spectroscopy. Imaging will be obtained from 8 nights on the KPNO 2.1m between Sept. 7 and 14 UT. Additionally, the Murillo Family Observatory, a 0.5m telescope on the CSUSB campus which is equipped with both broadband filters and a narrowband Hale-Bopp set of filters will be used to observe the comet every clear night the moon allows between late August and early October to obtain extensive lightcurve data. These data will overlap both the Kitt Peak observations and the infrared spectroscopy which will be obtained with the SpeX instrument at the IRTF on four nights between September 26 UT and October 2 UT.

Abell, Paul↗
NASA NTRS2014

Comet C2012 S1 (ISON)s Carbon-rich and Micron-size-dominated Coma Dust

Comet C/2012 S1 (ISON) was unique in that it was a dynamically new comet derived from the Nearly Isotropic Oort cloud reservoir of comets with a sun-grazing orbit. We present thermal models for comet ISON (rh approx.1.15 AU, 2013-Oct-25 11:30 UT) that reveal comet ISON's dust was carbon-rich and dominated by a narrow size distribution dominated by approx. micron-sized grains. We constrained the models by our SOFIA FORCAST photometry at 11.1, 19.7 and 31.5 microns and by a silicate feature strength of approx.1.1 and an 8-13microns continuum greybody color temperature of approx. 275-280 K (using Tbb ∝ r−0.5 h and Tbb approx. 260-265 K from Subaru COMICS, 2013-Oct-19 UT)[1,2]. N-band spectra of comet ISON with the BASS instrument on the NASA IRTF (2013-Nov-11-12 UT) show a silicate feature strength of approx. 1.1 and an 11.2microns forsterite peak.[3] Our thermal models yield constraints the dust composition as well as grain size distribution parameters: slope, peak grain size, porosity. Specifically, ISON's dust has a low silicate-to- amorphous carbon ratio (approx. 1:9), and the coma size distribution has a steep slope (N4.5) such that the coma is dominated by micron-sized, moderately porous, carbon-rich dust grains. The N-band continuum color temperature implies submicronto micron-size grains and the steep fall off of the SOFIA far-IR photometry requires the size distribution to have fewer relative numbers of larger and cooler grains compared to smaller and hotter grains. A proxy for the dust production rate is fρ approx.1500 cm, akin to Afρ. ISON has a moderate-to-low dust-to-gas ratio. Comet ISON's dust grain size distribution does not appear similar to the few well-studied long-period Nearly Isotropic Comets (NICs), namely C/1995 O1 (Hale-Bopp) and C/2001 Q4 (NEAT) that had smaller and/or more highly porous grains and larger sizes, or C/2007 N4 (Lulin) and C/2006 P1 (McNaught) that had large and/or compact grains. Radial transport to comet-forming disk distances (≥ 20 AU) is easier for smaller grains (≤1 micron) than for larger grains (approx. 20 microns like Stardust terminal particles). The presence of predominantly micron-sized and smaller grains suggests comet ISON may have formed either earlier in disk evolution whereby larger grains did not have the time to be transported to distances beyond Neptune, or the comet formed so far out in the disk that larger grains did not traverse such large radial distances. The high carbon-content of ISON's refractory dust appears to be complimented by the presence of limitedlifetime organic (CHON-like) grain materials: preliminary analyses of near-IR and high-resolution optical spectra indicate that gas-phase daughter molecules C2, CN, and CH were more abundant than their parent molecules (C2H2, C2H6, measured in the near- IR). Dust composition as well as grain size distribution parameters (slope, peak grain size, and porosity) give clues to comet origins.

NASA/DLR Stratospheric Observatory↗
NASA NTRS2016

Investigating Ozone Sources in California Using AJAX Airborne Measurements and Models: Implications for Stratospheric Intrusion and Long Range Transport

High ozone concentrations at low altitudes near the surface were detected from airborne Alpha Jet Atmospheric eXperiment (AJAX) measurements on May 30, 2012. We investigate the causes of the elevated ozone concentrations using the airborne measurements and various models. GEOSchem and WRF-STILT model simulations show that the contribution from local sources is small. From MERRA reanalysis, it is found that high potential vorticity (PV) is observed at low altitudes. This high PV appears to be only partially coming through the stratospheric intrusions because the air inside the high PV region is moist, which shows that mixing appears to be enhanced in the low altitudes. Considering that diabatic heating can also produce high PV in the lower troposphere, high ozone is partially coming through stratospheric intrusion, but this cannot explain the whole ozone concentration in the target areas of the western U.S. A back-trajectory model is utilized to see where the air masses originated. The air masses of the target areas came from the lower stratosphere (LS), upper (UT), mid- (MT), and lower troposphere (LT). The relative number of trajectories coming from LS and UT is low (7.7% and 7.6%, respectively) compared to that from LT (64.1%), but the relative ozone concentration coming from LS and UT is high (38.4% and 20.95%, respectively) compared to that from LT (17.7%). The air mass coming from LT appears to be mostly coming from Asia. Q diagnostics show that there is sufficient mixing along the trajectory to indicate that ozone from the different origins is mixed and transported to the western U.S. This study shows that high ozone concentrations can be detected by airborne measurements, which can be analyzed by integrated platforms such as models, reanalysis, and satellite data.

gas composition↗
NASA NTRS2015

Investigation of Ozone Sources in California Using AJAX Airborne Measurements and Models: Implications for Stratospheric Intrusion and Long Range Transport

High ozone concentrations at low altitudes near the surface were detected from airborne Alpha Jet Atmospheric eXperiment (AJAX) measurements on May 30, 2012. We investigate the causes of the elevated ozone concentrations using the airborne measurements and various models. GEOS-chem and WRF-STILT model simulations show that the contribution from local sources is small. From MERRA reanalysis, it is found that high potential vorticity (PV) is observed at low altitudes. This high PV appears to be only partially coming through the stratospheric intrusions because the air inside the high PV region is moist, which shows that mixing appears to be enhanced in the low altitudes. Considering that diabatic heating can also produce high PV in the lower troposphere, high ozone is partially coming through stratospheric intrusion, but this cannot explain the whole ozone concentration in the target areas of the western U.S. A back-trajectory model is utilized to see where the air masses originated. The air masses of the target areas came from the lower stratosphere (LS), upper (UT), mid- (MT), and lower troposphere (LT). The relative number of trajectories coming from LS and UT is low (7.7 and 7.6, respectively) compared to that from LT (64.1), but the relative ozone concentration coming from LS and UT is high (38.4 and 20.95, respectively) compared to that from LT (17.7). The air mass coming from LT appears to be mostly coming from Asia. Q diagnostics show that there is sufficient mixing along the trajectory to indicate that ozone from the different origins is mixed and transported to the western U.S. This study shows that high ozone concentrations can be detected by airborne measurements, which can be analyzed by integrated platforms such as models, reanalysis, and satellite data.

low altitude↗
NASA NTRS2017

Observations of Deep Ionospheric F-Region Density Depletions with FPMU Instrumentation and their Relationship with the Global Dynamics of the June 22-23, 2015 Geomagnetic Storm

The magnetic storm that commenced on June 22, 2015 was one of the largest storms in the current solar cycle, resulting from an active region on the Sun that produced numerous coronal mass ejections (CMEs) and associated interplanetary shock waves. On June 22 at 18:36 UT the magnetosphere was impacted by the leading-edge shock wave and a sheath carrying a large and highly variable interplanetary magnetic field (IMF) Bz with values ranging from +25 to -40 nT. During the subsequent interval from 0000 to 0800 UT, there was a second intensification of the geomagnetic storm resulting from the impact of the CME. We present dramatic responses of simultaneous particle measurements from the high-altitude Magnetospheric Multiscale Mission (MMS) at high altitudes in the magnetosphere (approx. 9-12 Re) and from the low-altitude (F-region) Floating Potential Measurement Unit (FPMU) on board the International Space Station (ISS). We analyze potential causes of these dramatic particle flux dropouts by putting them in the context of storm-time electrodynamics, and support our results with numerical simulations of the global magnetosphere and ionosphere. During the sheath phase of the storm, the MMS spacecraft in the near-earth equatorial plane observed a rapid reconfiguration of the magnetic field near 1923 UT. Initially in the warm plasmasheet, particle flux dropouts were observed as they tracked the plasma-sheet to lobe transitions with the stretching and thinning of the plasmasheet. Anti-sunward flowing O+ ions of ionospheric origin were also measured during this period, confirming that the MMS spacecraft temporarily was in a lobe.

Coffey, Victoria↗
NASA NTRS2017

Observations of Deep Ionospheric F-Region Density Depletions with FPMU Instrumentation and Their Relationship with the Global Dynamics of the June 22-23, 2015 Geomagnetic Storm

The magnetic storm that commenced on June 22, 2015 was one of the largest storms in the current solar cycle, resulting from an active region on the Sun that produced numerous coronal mass ejections (CMEs) and associated interplanetary shock waves. On June 22 at 18:36 UT the magnetosphere was impacted by the leading-edge shock wave and a sheath carrying a large and highly variable interplanetary magnetic field (IMF) Bz with values ranging from +25 to -40 nT. During the subsequent interval from 0000 to 0800 UT, there was a second intensification of the geomagnetic storm resulting from the impact of the CME. We present dramatic responses of simultaneous particle measurements from the high-altitude Magnetospheric Multiscale Mission (MMS) at high altitudes in the magnetosphere (approx. 9-12 Re) and from the low-altitude (F-region) Floating Potential Measurement Unit (FPMU) on board the International Space Station (ISS). We analyze potential causes of these dramatic particle flux dropouts by putting them in the context of storm-time electrodynamics, and support our results with numerical simulations of the global magnetosphere and ionosphere. During the sheath phase of the storm, the MMS spacecraft in the near-earth equatorial plane observed a rapid reconfiguration of the magnetic field near 1923 UT. Initially in the warm plasma sheet, particle flux dropouts were observed as they tracked the plasma-sheet to lobe transitions with the stretching and thinning of the plasma sheet. Anti-sunward flowing O+ ions of ionospheric origin were also measured during this period, confirming that the MMS spacecraft temporarily was in a lobe.

Coffey, Victoria↗
NASA NTRS2017

Properties, Propagation, and Excitation of EMIC Waves Properties, Propagation, and Excitation of EMIC Waves

Electromagnetic ion cyclotron (EMIC) waves (0.1-5 Hz) play an important role in particle dynamics in the Earth's magnetosphere. EMIC waves are preferentially excited in regions where hot anisotropic ions and cold dense plasma populations spatially overlap. While the generation region of EMIC waves is usually on or near the magnetic equatorial plane in the inner magnetosphere, EMIC waves have both equatorial and off-equator source regions on the dayside in the compressed outer magnetosphere. Using field and plasma measurements from the Magnetospheric Multiscale (MMS) mission, we perform a case study of EMIC waves and associated local plasma conditions observed on 19 October 2015. From 0315 to 0810 UT, before crossing the magnetopause into the magnetosheath, all four MMS spacecraft detected long-lasting He(exp +)-band EMIC wave emissions around local noon (MLT = 12.7 - 14.0) at high L-shells (L = 8.8 - 15.2) and low magnetic latitudes (MLAT = -21.8deg - -30.3deg). Energetic (greater than 1 keV) and anisotropic ions were present throughout this event that was in the recovery phase of a weak geomagnetic storm (min. Dst = -48 nT at 1000 UT on 18 October 2015). The testing of linear theory suggests that the EMIC waves were excited locally. Although the wave event is dominated by small normal angles, its polarization is mixed with right- and left-handedness and its propagation is bi-directional with regard to the background magnetic field. The short inter-spacecraft distances (as low as ~15 km) of the MMS mission make it possible to accurately determine the k vector of the waves using the phase difference technique. Preliminary analysis finds that the k vector magnitude, phase speed, and wavelength of the 0.3-Hz wave packet at 0453:55 UT are 0.005 km(exp -1), 372.9 km/s, and 1242.9 km, respectively.

Zhang, Jichun↗
NASA NTRS2012

The Formation Heritage of Jupiter Family Comet 10P/Tempel 2 as Revealed by Infrared Spectroscopy

We present spectral and spatial information for major volatile species in Comet 10P/Tempel 2, based on high-dispersion infrared spectra acquired on UT 2010 July 26 (heliocentric distance R h = 1.44 AU) and September 18 (R h = 1.62 AU), following the comet’s perihelion passage on UT 2010 July 04. The total production rate for water on July 26 was (1.90 ± 0.12) × 10 28 molecules s −1 , and abundances of six trace gases (relative to water) were: CH 3 OH (1.58% ± 0.23%), C 2 H 6 (0.39% ± 0.04%), NH 3 (0.83% ± 0.20%), and HCN (0.13% ± 0.02%). A detailed analysis of intensities for water emission lines provided a rotational temperature of 35 ± 3 K. The mean OPR is consistent with nuclear spin populations in statistical equilibrium (OPR = 3.01 ± 0.18), and the (1σ) lower bound corresponds to a spin temperature >38 K. Our measurements were contemporaneous with a jet-like feature observed at optical wavelengths. The spatial profiles of four primary volatiles display strong enhancements in the jet direction, which favors release from a localized vent on the nucleus. The measured IR continuum is much more sharply peaked and is consistent with a dominant contribution from the nucleus itself. The peak intensities for H 2 O, CH 3 OH, and C 2 H 6 are offset by ∼200 km in the jet direction, suggesting the possible existence of a distributed source, such as the release of icy grains that subsequently sublimed in the coma. On UT September 18, no obvious emission lines were present in our spectra, nevertheless we obtained a 3σ upper limit Q(H 2 O) < 2.86 × 10 27 molecules s −1 .

Organic Chemistry↗
NASA NTRS2018

An Overview of the XRT Observations for the September 10 2017 X Flare

The September 10 2017 X8 flare was one of the biggest of the current solar cycle, and it was also incredibly well observed. This presentation will review the available XRT data for this flare. XRT observed the initial flux rope eruption, which was also well observed by AIA and SUVI. XRT data is missing for some of the impulsive phase of the flare, but the late phase shows some very interesting features. There is some nice XRT imaging of a plasma sheet above cusp-shaped loops between 16:47 and 18:39 UT. After 18:40 UT, a double loop structure becomes apparent, with a more rounded loop nested within a larger, cusp-shaped structure. The flare loops continue to grow, and at around 1 UT on September 11, supra-arcade down flows and shrinking loops become visible. There is also interesting data in the late phase of this flare from RHESSI and the Expanded Owens Valley Solar Array (EOVSA).

Reeves, Kathy↗
NASA NTRS2021

Rapidly Varying Anisotropic Methanol (CH3OH) Production in the Inner Coma of Comet 46P/Wirtanen as Revealed by the ALMA Atacama Compact Array

We report the detection of CH3OH emission in comet 46P/Wirtanen on UT 2018 December 8 and9 using the Atacama Compact Array (ACA), part of the Atacama Large Millimeter/Submillimeter Array (ALMA). These interferometric measurements of CH3OH along with continuum emission from dust probed the inner coma(<2000 km from the nucleus)of 46P/Wirtanen approximately oneweek before its closest approach to Earth(∆= 0.089 – 0.092 au), revealing rapidly varying and anisotropic CH3OH outgassing during five separate ACA executions between UT 23:57 December 7 and UT 04:55 December 9, with clear progression in the spectral line profilesover a timescaleofminutes. We present spectrally integrated flux maps, production rates, rotational temperatures, andspectral line profiles of CH3OH during each ACA execution. The variations in CH3OH outgassing are consistent with Wirtanen’s 9 hour nucleus rotational period derived from optical and millimeterwavelength measurements, and thus are likely coupled to the changing illumination of active sites onthe nucleus. The consistent blue offset of the line center indicates enhanced CH3OH sublimationfromthe sunward hemisphere of the comet, perhaps from icy grains. These results demonstrate theexceptional capabilities of the ACA for time-resolved measurements of comets such as 46P/Wirtanen.

Nathan X Roth↗
DOE OSTI2026

Attention-based 3D – convolutional neural network model for mechanical property predictions using visible light images in metal additive manufacturing

Additive manufacturing (AM), while commonly used for rapid prototyping and creating components with complex geometries, has not been widely adopted for critical applications across the aerospace, automotive, defense, energy, and medical industries. This is, in part, due to the challenges of controlling flaws and uncertainty in the mechanical behavior of additively manufactured components. In recent years, there has been an increase in research aimed at predicting the final mechanical properties of additively manufactured components during the printing process. To address these issues, a 3D-CNN model was trained using low-cost in situ visible-light camera data, anomaly classifications, and the chosen process parameters to predict the ultimate tensile strength (UTS), yield strength (YS), total elongation (TE), and uniform elongation (UE). The 3D-CNN layers of the model employed attention mechanisms to prioritize features in the data, thereby improving prediction accuracy. Furthermore, the effect of each process parameter and anomaly class is investigated using attention-based dynamic sigmoid weighted gates to interpret the influence each class has on the final prediction. Different combinations of the in situ data were fed into the 3D-CNN, with varying amounts of image layers, to determine the ideal combination for predicting mechanical properties in situ. Here, the 3D-CNN model achieved mean absolute percentage errors (MAPE) below 5% for both UTS and YS while using only a single camera input and under half of the available image layers.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Examining infrared thermography based approaches to rapid fatigue characterization of additively manufactured compression molded short fiber thermoplastic composites

A novel additive manufacturing (AM) methodology combined with a compression molding (CM) process has been developed to optimize the microstructure of short fiber thermoplastic composites (SFTs)with higher fiber alignment and lower porosity, yielding superior stiffness, strength, and structural integrity. Here, the current work examines the efficacy of the ‘passive’ infrared thermography (IRT) techniques for rapid fatigue characterization of SFTs that use the surface temperature evolution during cyclic loading due to self-heating as a fatigue indicator. A comparison of fatigue limits obtained from traditional stress-life (SN) (≈53.1%σ uts ) and IRT (≈54.1%σ uts ) shows a close match. However, the SN curve required 18 specimens and two weeks of continuous cyclic testing, while IRT used three specimens with 5 hours of testing. Thus, the IRT approach provides an accelerated testing framework for rapidly estimating the fatigue limit. Additionally, existing phenomenological approaches to IRT fatigue characterization have been examined.

42 ENGINEERING↗
DOE OSTI2025

High temperature tensile properties of H13 tool steel fabricated via binder jet additive manufacturing

H13 is one of the most used steels for hot work tooling applications like die casting. Binder jet additive manufacturing (BJAM) offers a cost-effective solution to fabricate tools with performance enhancing features like internal cooling channels. While full densification of BJAM H13 has been demonstrated in the literature, the room and high temperature tensile properties of the material, and the effect of post-process treatments like hot isostatic pressing (HIP), and heat treatment (HT) are unknown. Here we use HIP+HT to obtain a combination of ultimate tensile strength (UTS) and elongation to failure that is superior to H13 fabricated via laser powder bed fusion and wire arc additive manufacturing. HIP+HT BJAM H13 has a superior UTS (1836 MPa) compared with conventionally processed (1671 MPa) H13 but has lower plastic elongation. We report that as-sintered samples have an extensive network of continuous grain boundary (GB) carbides that contribute to brittle fracture. HIP+HT results in partial dissolution of GB carbides and formation of a fine martensitic microstructure resulting in superior tensile strength and elongation compared with as-sintered H13. We use thermo-kinetic calculations to rationalize the effects of HIP+HT on microstructure and properties.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Correlating energy density induced residual stress, porosity, and mechanical property variations in directed energy deposition using neutron diffraction and imaging techniques

In a directed energy deposition (DED) process, the energy density deposited into the materials drives the structure and properties. The energy density is influenced by several printing parameters, such as laser power, hatch spacing, scanning speed, etc. In this study, 316L stainless steels samples were fabricated with varying hatch spacing aiming to induce variations in sample microstructure and properties. A combination of techniques including engineering neutron diffraction, neutron computed tomography, and tensile testing, were employed to correlate the structure and macroscopic properties. Porosity was predominantly observed at the bottom of the deposited materials and was effectively reduced by decreased hatch spacing. Correlatively, compressive residual stress was observed at the bottom of the specimens, while internal stress is largely determined by the hatch spacing. The residual stress in general decreases as the hatch spacing decreases. The maximum ultimate tensile strength (UTS) was found to increase from 517.4 to 528.6 MPa with a strain increased from 0.847 to 0.897 as the hatch spacing decreased from 0.45 to 0.42 mm. However, further increase in energy density by reducing the hatch spacing resulted in a significant decrease in UTS (as low as 452 MPa). This work provides new insights from bulk non-destructive techniques into correlating energy density induced residual stress, porosity, and mechanical property variations in directed energy deposition.

316L stainless steel↗
DOE OSTI2024

First Observation of Harmonics of Magnetosonic Waves in Martian Magnetosheath Region

Abstract The present study provides an evidence for the generation of harmonics of magnetosonic waves in the Martian magnetosheath region. The wave signatures are manifested in the magnetic field measurements recorded by the fluxgate magnetometer instrument onboard the Mars Atmosphere and Volatile Evolution missioN (MAVEN) spacecraft in the dawn sector around 5–10 LT at an altitude of 4,000–6,000 kms. The wave that is observed continuously from 19.1 to 20.7 UT below the proton cyclotron frequency (f ci ≈ 46 mHz) is identified as fundamental mode of the magnetosonic wave. Whereas harmonics of the magnetosonic wave are observed during 19.7–20.3 UT at frequencies that are multiple off ci . The ambient solar wind proton density and plasma flow velocity are found to vary with a fundamental mode frequency of 46 mHz. It is noticed that the fundamental mode is mainly associated with the left‐hand (LH), and higher frequency harmonics are associated with the right‐hand (RH) circular polarizations. A clear difference in the polarization and ellipticity is noticed during the time of occurrence of harmonics. The magnetosonic wave harmonics are found to propagate in the quasi‐perpendicular directions to the ambient magnetic field. The results of linear theory and Particle‐In‐Cell simulation performed here are in agreement with the observations. The present study provides a conclusive evidence for the occurrence of harmonics of magnetosonic wave in the close vicinity of the magnetosheath region of the unmagnetized planet Mars.

Astronomy & Astrophysics↗
DOE OSTI2024

Mixture-of-Experts for Multi-Domain Defect Identification in Non-Destructive Inspection

Composite materials are widely used in aircraft structures because of their superior mechanical properties. However, their complex failure modes require sophisticated inspection methods to ensure structural integrity. Ultrasonic testing (UT) is a common non-destructive inspection (NDI) technique for aircraft composites that can detect internal and external defects with high resolution and accuracy. Despite their effectiveness, traditional UT methods rely on the manual interpretation of ultrasonic signals, which is time-consuming, labor-intensive, and subjective. Furthermore, processing such large-scale data, particularly across materials of varying thicknesses, significantly increases the computational demands of deep learning model optimization. To overcome these challenges, we propose an efficient sparse mixture-of-experts (MoE) model with a multi-level loss function and introduce four novel training objectives to improve computational efficiency and accuracy in identifying surface defects in composite aircraft materials. Here, we evaluated our approach on material with multiple thicknesses or domains comprising various defects. Our experimental results demonstrate higher accuracy and F1-Score, with only 10% training epochs compared to baseline MoE.

composite materials↗