Search NASA⌕ Search

SEARCH · Search NASA

Results for “higher accuracy”

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 757 records · Page 42
All sourcesNASA NTRSDOE OSTI
NASA NTRS1991

The control network of Mars: April 1991

The modern geodetic control network of Mars was first established based on Mariner 9 images with 1-2 km/pixel resolutions and covered almost the entire Martian surface. The introduction of higher resolution (10-200 meter/pixel) Viking Orbiter images greatly improved the accuracy and density of points in the control network. Analysis of the Viking Lander radio tracking data led to more accurate measurements of Mars' rotation period, spin axis direction, and the lander coordinates relative to the inertial reference frame. The prime meridian on Mars was defined by the Geodesy/Cartography Group of the Mariner 9 Television Team as the crater Airy-0, located about 5 degrees south of the equator. The Viking 1 Lander site was identified on a high resolution Viking frame. The control point measurements form the basis of a least squares solution determined by analytical triangulation after the pixel measurements are corrected for geometric distortions and converted to millimeter coordinates in the camera focal plane. Photogrammetric strips encircling Mars at the equator and at 60 degree north south were used to strengthen the overall net and improve the accuracy of the coordinates of points. In addition, photogrammetric strips along 0, 90, 180, and 270 degrees longitude to the Viking 1 Lander site have all significantly strengthened the control network. Most recently, photogrammetric strips were added to the net along 30 degrees north latitude between 0 and 180 degrees, and along 30 degrees between 180 and 360 degrees. The Viking 1 Lander site and Airy-0 are linked through photogrammetric strips occurring along the 0 degree meridian from Airy-0 to 65 degrees north, from that point through the Viking 1 Lander site to the equator, and along the equator to 180 degrees longitude. The Viking 1 lander site is thus a well calibrated area with coordinates of points accurate to approximately 200 meters relative to the J2000 inertial coordinate system. This will be a useful calibration region for upcoming missions. The current status of the control network calculations is presented.

Davies, Merton E.↗
NASA NTRS1993

Comparison of symbolic and numerical integration methods for an assumed-stress hybrid shell element

Hybrid shell elements have long been regarded with reserve by the commercial finite element developers despite the high degree of reliability and accuracy associated with such formulations. The fundamental reason is the inherent higher computational cost of the hybrid approach as compared to the displacement-based formulations. However, a noteworthy factor in favor of hybrid elements is that numerical integration to generate element matrices can be entirely avoided by the use of symbolic integration. In this paper, the use of the symbolic computational approach is presented for an assumed-stress hybrid shell element with drilling degrees of freedom and the significant time savings achieved is demonstrated through an example.

Rengarajan, Govind↗
NASA NTRS2002

Comparison of Mean Climate Trends in the Northern Hemisphere Between N.C.E.P. and Two Atmosphere-Ocean Model Forced Runs

Results are presented for two greenhouse gas experiments of the Goddard Institute for Space Studies Atmosphere-Ocean Model (AOM). The computed trends of surface pressure, surface temperature, 850, 500 and 200 mb geopotential heights and related temperatures of the model for the time frame 1960-2000 are compared to those obtained from the National Centers for Environmental Prediction observations. A spatial correlation analysis and mean value comparison are performed, showing good agreement. A brief general discussion about the statistics of trend detection is presented. The domain of interest is the Northern Hemisphere (NH) because of the higher reliability of both the model results and the observations. The accuracy that this AOM has in describing the observed regional and NH climate trends makes it reliable in forecasting future climate changes.

Lucarini, Valerio↗
NASA NTRS2011

Fast Approximate Broadband Phase Retrieval for Segmented Systems

Broadband phase retrieval needed when: a) Narrow spectral filters are unavailable. b) Dim sources. c) Low throughput due to misalignment. d) Short exposures times. i.e., Pointing instability (space); and Atmospheric instability (ground based AO). Traditional approach is computationally burdensome for extreme bandwidths. Approximate approach: a) Substitute monochromatic model. b) Blur model and data. Test case performance: a) approx.270x reduction in computational cost for FGS-like test case. b) Good accuracy for monolithic system. c) Acceptable accuracy for segmented systems. i.e., Reduced by diffraction and Reduced by higher order segment model.

Jurling, Alden S.↗
NASA NTRS2013

PHyTIR - A Prototype Thermal Infrared Radiometer

This paper describes the PHyTIR (Prototype HyspIRI Thermal Infrared Radiometer) instrument, which is the engineering model for the proposed HyspIRI (Hyperspectral Infrared Imager) earth observing instrument. The HyspIRI mission would be comprised of the HyspIRI TIR (Thermal Infrared Imager), and a VSWIR (Visible Short-Wave Infra-Red Imaging Spectrometer). Both instruments would be used to address key science questions related to the earth's carbon cycle, ecosystems, climate, and solid earth properties. Data gathering of volcanic activities, earthquakes, wildfires, water use and availability, urbanization, and land surface compositions and changes, would aid the predictions and evaluations of such events and the impact they create. Even though the proposed technology for the HyspIRI imager is mature, the PHyTIR prototype is needed to advance the technology levels for several of the instrument's key components, and to reduce risks, in particular to validate 1) the higher sensitivity, spatial resolution, and higher throughput required for this focal plane array, 2) the pointing accuracy, 2) the characteristics of several spectral channels, and 4) the use of ambient temperature optics. The PHyTIR telescope consists of the focal plane assembly that is housed within a cold housing located inside a vacuum enclosure; all mounted to a bulkhead, and an optical train that consists of 3 powered mirrors; extending to both sides of the bulkhead. A yoke connects the telescope to a scan mirror. The rotating mirror enables to scan- a large track on the ground. This structure is supported by kinematic mounts, linking the telescope assembly to a base plate that would also become the spacecraft interface for HyspIRI. The focal plane's cooling units are also mounted to the base plate, as is an overall enclosure that has two viewing ports with large exterior baffles, shielding the focal plane from incoming stray light. PHyTIR's electronics is distributed inside and near the vacuum enclosure, and in a nearby rack. The data acquisition technique would be to take measurements over a 51deg wide swath in the cross spacecraft velocity direction, which is brought into view through the rotating scan mirror. A landscape mosaic thus can be assembled by overlaying rows of measurements. The paper briefly outlines the proposed HyspIRI mission and its data acquisition technique; it then describes the prototype PHyTIR instrument.

Space Instrument↗
NASA NTRS2015

Numerical Predictions of Mode Reflections in an Open Circular Duct: Comparison with Theory

The NASA Broadband Aeroacoustic Stator Simulation code was used to compute the acoustic field for higher-order modes in a circular duct geometry. To test the accuracy of the results computed by the code, the duct was terminated by an open end with an infinite flange or no flange. Both open end conditions have a theoretical solution that was used to compare with the computed results. Excellent comparison for reflection matrix values was achieved after suitable refinement of the grid at the open end. The study also revealed issues with the level of the mode amplitude introduced into the acoustic held from the source boundary and the amount of reflection that occurred at the source boundary when a general nonreflecting boundary condition was applied.

Duct Modes↗
NASA NTRS2018

Optimal Estimation-Based Algorithm to Retrieve Aerosol Optical Properties for GEMS Measurements Over Asia

The Geostationary Environment Monitoring Spectrometer (GEMS) is scheduled to be in orbit in 2019 onboard the GEO-KOMPSAT 2B satellite and will continuously monitor air quality over Asia. The GEMS will make measurements in the UV spectrum (300-500 nm) with 0.6 nm resolution. In this study, an algorithm is developed to retrieve aerosol optical properties from UV-visible measurements for the future satellite instrument and is tested using 3 years of existing OMI L1B data. This algorithm provides aerosol optical depth (AOD), single scattering albedo (SSA) and aerosol layer height (ALH) using an optimized estimation method. The retrieved AOD shows good correlation with Aerosol Robotic Network (AERONET) AOD with correlation coefficients of 0.83, 0.73 and 0.80 for heavy-absorbing fine (HAF) particles, dust and non-absorbing (NA) particles, respectively. However, regression tests indicate underestimation and overestimation of HAF and NA AOD, respectively. In comparison with AOD from the OMI/Aura Near-UV Aerosol Optical Depth and Single Scattering Albedo 1-orbit L2 Swath 13 km x 24 km V003 (OMAERUV) algorithm, the retrieved AOD has a correlation coefficient of 0.86 and linear regression equation, AOD(sub GEMS) = 1.18AOD(sub OMAERUV) + 0.09. An uncertainty test based on a reference method, which estimates retrieval error by applying the algorithm to simulated radiance data, revealed that assumptions in the spectral dependency of aerosol absorptivity in the UV cause significant errors in aerosol property retrieval, particularly the SSA retrieval. Consequently, retrieved SSAs did not show good correlation with AERONET values. The ALH results were qualitatively compared with the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) products and were found to be well correlated for highly absorbing aerosols. The difference between the attenuated-backscatter-weighted height from CALIOP and retrieved ALH were mostly closed to zero when the retrieved AOD is higher than 0.8 and SSA is lower than 0.93. Although retrieval accuracy was not significantly improved, the simultaneous consistent retrieval of AOD, SSA and ALH alone demonstrates the value of this stand-alone algorithm, given their nature for error using other methods. The use of these properties as input parameters for the air mass factor calculation is expected to improve the retrieval of other trace gases over Asia.

Geostationary Environment Monitoring Spectrometer ↗
NASA NTRSDate not supplied

Scale-Resolving Simulations of a Supersonic Retro-Propulsion Concept For Mars Entry, Descent, and Landing

Supersonic retro-propulsion (SRP) is a deceleration technology that could enable largerpayloads and potentially humans to be brought from space to the surface of Mars safely.Accurate and reliable CFD predictions are needed to help design future Mars entry, descent,and landing (EDL) vehicles due to the cost and limitations of wind tunnel tests. TypicalCFD approaches struggle at points along the Mars EDL trajectory with high enough altitude(low enough ambient pressure) that the SRP motor plumes are over-expanded. This workdemonstrates how high-order scale-resolving simulations on Cartesian AMR grids can provideaccurate and reliable time-averaged aerodynamic loads on the vehicle for this challengingflow regime. Greater than second order convergence in the time-averaged integrated axialforce𝐶𝐴is obtained across the three finest mesh resolutions simulated, and the estimateduncertainty is smaller than the unsteady standard deviation – demonstrating the accuracy andreliability of the method. Although costly, this approach can yield higher confidence in𝐶𝐴thanpreviously obtained in a reasonable turnaround time and is within reach for a few points in thetrajectory (i.e. Mach 2.4, 2.52, and 2.78). Further computational performance and numericalmethods improvements are needed to reduce the cost enough to bring this method into the EDLengineering design cycle.

S&M↗
NASA NTRSDate not supplied

A Reynolds-Averaged Navier-Stokes Perspective for the High Lift Common Research Model Using the LAVA Framework

An in-depth assessment of the High Lift-Common Research Model (HL-CRM) aerodynamic performance is made using a Reynolds-Averaged Navier-Stokes (RANS) methodology. Results presented here were produced in support of the 4th AIAA High Lift Prediction Workshop(HLPW-4), with additional simulation results and analysis used to support findings from the workshop. Mesh resolution, RANS closure models, and tunnel effects were among the sources of simulation sensitivity addressed, all in the pursuit of evaluating best-practice simulation methods for complex high-lift configurations. In general, free-air simulations using RANS yield accurate predictions of integrated loads in the pre-stall regime for the HL-CRM, while strong sensitivity to simulation method is observed near CLmax. Additionally, a study assessing changes in aerodynamic performance due to flap deflection angle indicated strong sensitivity to turbulence model closure and mesh resolution, particularly at the highest deflection setting. Several available corrections for the Spalart-Allmaras (SA) turbulence model were tested and generally introduced more error relative to the experiment in comparison to the baseline SA closure model. The Standard Menter Baseline Two-Equation (k-ωBSL) turbulence model was also explored, which arguably yielded results highly consistent with experiment, but at a substantially higher computational cost. Therefore, as a compromise between computational cost and accuracy, the baseline SA closure model was selected for additional HL-CRM studies that assess alternative solution methods. These strategies include “warm start” approaches, which initialize the simulation from a previously converged solution at a nearby angle of attack, unsteady calculations, and simulations that model the experimental wind tunnel. Results from these studies would reinforce the shortcomings of RANS in predicting aircraft performance at high-lift conditions, namely at high-αand highly loaded flap conditions, while also answering questions about appropriate modeling techniques to minimize the errors associated with RANS where possible

TTT↗
DOE OSTI2023

Faster approximate subgraph counts with privacy

One of the most common problems studied in the context of differential privacy for graph data is counting the number of non-induced embeddings of a subgraph in a given graph. These counts have very high global sensitivity. Therefore, adding noise based on powerful alternative techniques, such as smooth sensitivity and higher-order local sensitivity have been shown to give significantly better accuracy. However, all these alternatives to global sensitivity become computationally very expensive, and to date efficient polynomial time algorithms are known only for few selected subgraphs, such as triangles, k-triangles, and k-stars. In this paper, we show that good approximations to these sensitivity metrics can be still used to get private algorithms. Using this approach, we much faster algorithms for privately counting the number of triangles in real-world social networks, which can be easily parallelized. We also give a private polynomial time algorithm for counting any constant size subgraph using less noise than the global sensitivity; we show this can be improved significantly for counting paths in special classes of graphs

Nguyen, Dung↗
DOE OSTI2024

SHOCK TUBE IGNITION STUDIES OF RENEWABLE DIESEL FUELS FOR MEDIUM AND HEAVY-DUTY TRANSPORTATION

Currently extensive research on alternative fuels is being conducted due to their increasing demand to reduce greenhouse emissions. One renewable fuel studied in this work is dimethyl ether (DME) blended with propane(C3H8) as a potential mixture for heavy-duty engines used in semi-trucks. The blend has the potential to drastically reduce particulate and greenhouse gas emissions compared to a conventional diesel engine operating under similar conditions. To develop the use of mixture, one must conduct detailed conceptual and simulation studies before progressing to detail studies in CFD, engine modifications, and live testing. For simulations, accurate high-fidelity chemical kinetic models are necessary. However, the validity of the chemical kinetic mechanism for operating conditions of a heavy-duty mixing-controlled compression (MCCI) engine was widely unknown until recent work presented here and published. In this work, we studied the ignition of DME and propane blends in a shock tube under MCCI engine conditions. Ignition delay time (IDT) gathered behind the reflected shock for DME-propane mixtures for heavy-duty compression ignition (CI) engine parameters. Testing was conducted for undiluted varieties spanning from temperatures of 700 to 1100 K at pressures ranging from 55 to 84 bar for various blends (100% CH3OCH3, 100% C3H8, 60% CH3OCH3/ 40% C3H8) of DME and propane were combusted in synthetic air (21% O2/ 79% N2). Several experiments were conducted at higher pressures (90-120 bar) to improve the model performance and accuracy. The ignition delay times (IDTs) were compared to recent mechanisms, including Aramco3.0, NUIG, and Dames et al. A common trend among the mechanisms was overpredicted experimental IDTs. Further studies were conducted by a sensitivity analysis using the Dames et al. model, and critical reactions sensitive to IDTs of DME-propane mixture near 60 bar are outlined. Chemical analysis was conducted on the NTC region to explain chemical kinetics which is critical for developing MCCI heavy duty engines.

Mohammed, Zuhayr Pasha↗
NASA NTRS2020

Object-Based Comparison of Data-Driven and Physics-Driven Satellite Estimates of Extreme Rainfall

The Global Precipitation Measurement (GPM) constellation of spaceborne sensors provides a variety of direct and indirect measurements of precipitation processes. Such observations can be employed to derive spatially and temporally consistent gridded precipitation estimates either via data-driven retrieval algorithms or by assimilation into physically based numerical weather models. We compare the data-driven Integrated Multisatellite Retrievals for GPM (IMERG) and the assimilation-enabled NASA-Unified Weather Research and Forecasting (NU-WRF) model against Stage IV reference precipitation for four major extreme rainfall events in the southeastern United States using an object-based analysis framework that decomposes gridded precipitation fields into storm objects. As an alternative to conventional ‘‘grid-by-grid analysis,’’ the object-based approach provides a promising way to diagnose spatial properties of storms, trace them through space and time, and connect their accuracy to storm types and input data sources. The evolution of two tropical cyclones are generally captured by IMERG and NU-WRF, while the less organized spatial patterns of two mesoscale convective systems pose challenges for both. NU-WRF rain rates are generally more accurate, while IMERG better captures storm location and shape. Both show higher skill in detecting large, intense storms compared to smaller, weaker storms. IMERG’s accuracy depends on the input microwave and infrared data sources; NU-WRF does not appear to exhibit this dependence. Findings highlight that an object-oriented view can provide deeper insights into satellite precipitation performance and that the satellite precipitation community should further explore the potential for ‘‘hybrid’’ data-driven and physics-driven estimates in order to make optimal usage of satellite observations.

extreme events↗
NASA NTRS1991

Precise time dissemination and applications development on the Bonneville Power Administration system

The Bonneville Power Administration (BPA) uses IRIG-B transmitted over microwave as its primary system time dissemination. Problems with accuracy and reliability have led to ongoing research into better methods. BPA has also developed and deployed a unique fault locator which uses precise clocks synchronized by a pulse over microwaves. It automatically transmits the data to a central computer for analysis. A proposed system could combine fault location timing and time dissemination into a Global Position System (GPS) timing receiver and close the verification loop through a master station at the Dittmer Control Center. Such a system would have many advantages, including lower cost, higher reliability, and wider industry support. Test results indicate the GPS has sufficient accuracy and reliability for this and other current timing requirements including synchronous phase angle measurements. A phasor measurement system which provides phase angle has recently been tested with excellent results. Phase angle is a key parameter in power system control applications including dynamic braking, DC modulation, remedial action schemes, and system state estimation. Further research is required to determine the applications which can most effectively use real-time phase angle measurements and the best method to apply them.

K. E. Martin↗
DOE OSTI2024

Experimental testing of additively manufactured embedded fiber optic smart devices for clean energy applications

Abstract An additively manufactured prototype smart device was created to investigate in-flow temperature distributions using embedded high-definition fiber optic sensors within a component for clean energy systems. The devices were created using Ultrasonic Additive Manufacturing to create a unique embedded pathway within a flow conditioner for the high-definition fiber optic sensors to be placed within. The fibers used allowed for temperature measurements to be taken every 0.65 mm along the fiber. The high-resolution fibers were thermally calibrated enable the 2D reconstruction of the temperature profile in the flow path of the structure. This is due to the temperature-related strain response of the material and strain measurements of the fibers. Hot airflow testing of these devices showed the ability to identify localized temperature differences in the flow. The observed strain response within the smart device had higher strain concentrations in the thicker web regions than in the thinner web regions. These higher strain regions resulted in higher uncertainties for the temperature responses. Further calibration is needed to improve the accuracy of the smart devices, specifically within the inner web structures of a flow straightening component. Further investigations of the devices within flow showed the temperature sensing to be independent of the effects of flow velocity. The devices were able to distinguish temperature differences within single and two-phase flow and showed local sensitivity to the temperature changes with the identification of hot and cold spots. The presented results showed the viability of the smart device for obtaining detailed temperature distributions using common industrial components. Eventually, the goal for these smart devices will be to withstand higher temperature and pressure environments such as those experienced in nuclear, fusion, and concentrated solar energy systems.

Donlan, Connor F. (ORCID:0000000223317882)↗
DOE OSTI2023

Photometric metallicity for 694 233 Galactic giant stars from Gaia DR3 synthetic Strömgren photometry: Metallicity distribution functions of halo substructures

We used two previous calibrations and the standardised synthetic photometry in the v, b, and y Strömgren passbands from Gaia DR3 BP/RP spectra to obtain photometric metallicities for a selected sample of 694 233 old Galactic giant stars having |b|> 20.0° and parallax uncertainties lower than 10%. The zero point of both sets of photometric metallicities has been shifted to ensure an optimal match with the spectroscopic [Fe/H] values for 44 785 stars in common with APOGEE DR17, focusing on the metallicity range where they provide the highest accuracy. The metallicities derived in this way from one calibration display a typical accuracy of ≲0.1 dex and 1σ precision ≲0.2 dex in the range -2.2 ≲ [Fe/H] ≲ - 0.4, while they show a systematic trend with [Fe/H] at higher metallicity, beyond the applicability range of the relation. Those derived from the other calibration display, in general, reduced precision, and lower accuracy in the metal-poor regime, but have a median accuracy < 0.05 dex for [Fe/H] ≥ - 0.8. These results are confirmed and, consequently, the metallicities are validated, by comparison with large sets of spectroscopic metallicities from various surveys. The newly obtained metallicities are used to derive metallicity distributions for several previously identified substructures in the Galactic halo with an unprecedented number of stars. The catalogue including the two sets of metallicities and the associated uncertainties is made publicly available.

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRS1993

Testing higher-order Lagrangian perturbation theory against numerical simulation. 1: Pancake models

We present results showing an improvement of the accuracy of perturbation theory as applied to cosmological structure formation for a useful range of quasi-linear scales. The Lagrangian theory of gravitational instability of an Einstein-de Sitter dust cosmogony investigated and solved up to the third order is compared with numerical simulations. In this paper we study the dynamics of pancake models as a first step. In previous work the accuracy of several analytical approximations for the modeling of large-scale structure in the mildly non-linear regime was analyzed in the same way, allowing for direct comparison of the accuracy of various approximations. In particular, the Zel'dovich approximation (hereafter ZA) as a subclass of the first-order Lagrangian perturbation solutions was found to provide an excellent approximation to the density field in the mildly non-linear regime (i.e. up to a linear r.m.s. density contrast of sigma is approximately 2). The performance of ZA in hierarchical clustering models can be greatly improved by truncating the initial power spectrum (smoothing the initial data). We here explore whether this approximation can be further improved with higher-order corrections in the displacement mapping from homogeneity. We study a single pancake model (truncated power-spectrum with power-spectrum with power-index n = -1) using cross-correlation statistics employed in previous work. We found that for all statistical methods used the higher-order corrections improve the results obtained for the first-order solution up to the stage when sigma (linear theory) is approximately 1. While this improvement can be seen for all spatial scales, later stages retain this feature only above a certain scale which is increasing with time. However, third-order is not much improvement over second-order at any stage. The total breakdown of the perturbation approach is observed at the stage, where sigma (linear theory) is approximately 2, which corresponds to the onset of hierarchical clustering. This success is found at a considerable higher non-linearity than is usual for perturbation theory. Whether a truncation of the initial power-spectrum in hierarchical models retains this improvement will be analyzed in a forthcoming work.

Buchert, T.↗
NASA NTRS1985

Optical multicolors - A poor person's z machine for galaxies

The feasibility, limits, accuracy, and uncertainties of using optical broadband colors in the form of two-color plots to measure the redshifts of faint galaxies are explored. Comparison of redshifts from slit spectra and 4-m photographic UBVI photometry of 100 faint galaxies confirms that colors do yield redshifts. These galaxies of 17-23 mag in B span a wide range of intrinsic color and are thus unlike previous samples of faint galaxies that were largely radio or cluster galaxies of early (red) type. For redshifts less than 0.35, two-thirds of the color redshifts agree with the spectroscopic values to within + or - 0.04 in z; for higher redshifts to z of 0.6, the errors increase to + or - 0.06. These accuracies were achieved for blue as well as red galaxies, and ensure that multicolors can serve as a poor person's redshift machine to study distant galaxies in clusters and the field.

Koo, D. C.↗
DOE OSTI2022

DeepAdversaries: examining the robustness of deep learning models for galaxy morphology classification

With increased adoption of supervised deep learning methods for work with cosmological survey data, the assessment of data perturbation effects (that can naturally occur in the data processing and analysis pipelines) and the development of methods that increase model robustness are increasingly important. In the context of morphological classification of galaxies, we study the effects of perturbations in imaging data. In particular, we examine the consequences of using neural networks when training on baseline data and testing on perturbed data. We consider perturbations associated with two primary sources: (a) increased observational noise as represented by higher levels of Poisson noise and (b) data processing noise incurred by steps such as image compression or telescope errors as represented by one-pixel adversarial attacks. We also test the efficacy of domain adaptation techniques in mitigating the perturbation-driven errors. We use classification accuracy, latent space visualizations, and latent space distance to assess model robustness in the face of these perturbations. For deep learning models without domain adaptation, we find that processing pixel-level errors easily flip the classification into an incorrect class and that higher observational noise makes the model trained on low-noise data unable to classify galaxy morphologies. On the other hand, we show that training with domain adaptation improves model robustness and mitigates the effects of these perturbations, improving the classification accuracy up to 23% on data with higher observational noise. Domain adaptation also increases up to a factor of ${\approx}2.3$ the latent space distance between the baseline and the incorrectly classified one-pixel perturbed image, making the model more robust to inadvertent perturbations. Successful development and implementation of methods that increase model robustness in astronomical survey pipelines will help pave the way for many more uses of deep learning for astronomy.

79 ASTRONOMY AND ASTROPHYSICS↗