Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model transferability”

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 253 records · Page 14

Infrared Emission from Kilonovae: The Case of the Nearby Short Hard Burst GRB 160821B

In this paper, we present constraints on Ks-band emission from one of the nearest short hard gamma-ray bursts, GRB 160821B, at z = 0.16, at three epochs. We detect a red relativistic afterglow from the jetted emission in the first epoch but do not detect any excess kilonova emission in the second two epochs. We compare upper limits obtained with Keck I/MOSFIRE to multi-dimensional radiative transfer models of kilonovae, that employ composition-dependent nuclear heating and LTE opacities of heavy elements. We discuss eight models that combine toroidal dynamical ejecta and two types of wind and one model with dynamical ejecta only. We also discuss simple, empirical scaling laws of predicted emission as a function of ejecta mass and ejecta velocity. Our limits for GRB 160821B constrain the ejecta mass to be lower than 0.03 M ⊙ for velocities greater than 0.1 c. At the distance sensitivity range of advanced LIGO, similar ground-based observations would be sufficiently sensitive to the full range of predicted model emission including models with only dynamical ejecta. Finally, the color evolution of these models shows that I–K color spans 7–16 mag, which suggests that even relatively shallow infrared searches for kilonovae could be as constraining as optical searches.

79 ASTRONOMY AND ASTROPHYSICS↗

An algorithm for physics informed scan path optimization in additive manufacturing

Site specific microstructure control is a critical research area within the field of additive manufacturing due to its potential to revolutionize part performance. One way to achieve site specific microstructure control is through control of the solidification conditions via the construction of intricate scan paths; however, the search space for such a problem is large. Previous attempts only considered the solidification conditions at the top surface while also requiring either lots of manual-fine tuning or large amounts of computational resources. This paper introduces a general method for scan path optimization which considers the solidification conditions in the bulk of the material without an increase in computational expense. This method consists of three core components:1. A heat transfer model for simulating the temperature field at a given time.2. A surrogate model which takes scan pattern information and temperature data and predicts the solidification conditions of the bulk as well as the meltpool depths for a spot melt.3. A decision algorithm to decide which spot melt should be printed next based on the outputs of the surrogate model.Each of these components can be changed without changing the overall method. Within this work, this method is applied in the creation of an algorithm containing a semi-analytic heat transfer model to simulate the temperature field, a fully convolutional neural network (FCNN) as the surrogate model, and a greedy decision algorithm. The resulting algorithm produced complex scan patterns which gave strong results for simulated microstructure control.

36 MATERIALS SCIENCE↗

Modeling Heat Transfer and Pressurization of Polymeric Methylene Diisocyanate (PMDI) Polyurethane Foam in a Sealed Container

Polymer foam encapsulants provide mechanical, electrical, and thermal isolation in engineered systems. It can be advantageous to surround objects of interest, such as electronics, with foams in a hermetically sealed container to protect the electronics from hostile environments, such as a crash that produces a fire. However, in fire environments, gas pressure from thermal decomposition of foams can cause mechanical failure of the sealed system. In this work, a detailed study of thermally decomposing polymeric methylene diisocyanate (PMDI)-polyether-polyol based polyurethane foam in a sealed container is presented. Both experimental and computational work is discussed. Three models of increasing physics fidelity are presented: No Flow, Porous Media, and Porous Media with VLE. Each model us described in detail, compared to experiment, and uncertainty quantification is performed. While the Porous Media with VLE model matches has the best agreement with experiment, it also requires the most computational resources.

36 MATERIALS SCIENCE↗

Dual-pace transient heat conduction in vertically aligned carbon nanotube arrays induced by structure separation

Vertically aligned arrays of carbon nanotubes (CNTs) are attractive for a wide range of macroscopic applications which can exploit the remarkable properties of individual nanotubes. In this work, an abnormal behavior of CNT bundles in heat conduction is discovered under the transient electro-thermal characterization. The measured voltage change over the sample shows a dual-pace thermal response (DTR), which could not be fitted using a single thermal diffusivity heat transfer model. Instead, two thermal diffusivities are determined from the DTR phenomenon. Three rounds of cryogenic experiments are conducted to investigate the physics behind DTR phenomenon. It starts to emerge when the temperature is reduced to a certain level. After two rounds of cryogenic experiments, the DTR phenomenon becomes permanent from 295 K to 10 K. The nano-scale structure separation induced by the increased thermal strain leads to macroscale structure separation, which results in two parallel heat conduction paths responsible for the DTR phenomenon. By building a new parallel heat transfer model taking both the transient and steady-state electrical and thermal response into consideration, the area ratio of separated CNTs is determined. Furthermore, this result uncovers the existence of coiled morphology and nano-structure evolution of CNTs under cryogenic state and its effect on thermal and electrical transport, especially on their transient behaviors. Knowledge of the thermal transport in these CNTs arrays is important considering the fact that transient thermal response strongly affects their mechanical, optical, and electrical behaviors.

42 ENGINEERING↗

Ground-based high spectral resolution observations of the entire terrestrial spectrum under extremely dry conditions: OBSERVATIONS OF THE ENTIRE IR SPECTRUM

A field experiment was conducted in northern Chile at an altitude of 5.3 km to evaluate the accuracy of line­by-­line radiative transfer models in regions of the spectrum that are typically opaque at sea level due to strong water vapor absorption. A suite of spectrally resolved radiance instruments collected simultaneous observations that, for the first time ever, spanned the entire terrestrial thermal spectrum (i.e., from 10 to 3000 cm-1, or 1000 to 3.3 μm). These radiance observations, together with collocated water vapor and temperature profiles, are used to provide an initial evaluation of the accuracy of water vapor absorption in the far-­infrared of two line-­by­line radiative transfer models. These initial results suggest that the more recent of the two models is more accurate in the strongly absorbing water vapor pure rotation band. This result supports the validity of the Turner et al. (2012) study that demonstrated that the use of the more recent water vapor absorption model in climate simulations resulted in significant radiative and dynamical changes in the simulation relative to the older water vapor model.

Turner, D. D.↗

Modeling batch melting: Roles of heat transfer and reaction kinetics

Development of mathematical models of heat and mass transfer in glass-melting furnaces began in the 1970s and progressed rapidly with advances in sophisticated experimental/numerical techniques and increasing computational power. Today, practically all newly built or rebuilt furnaces are optimized with these models to meet stringent quality requirements, reduce the unit costs of manufacturing, or control emissions. One remaining hurdle is to model the batch-to-glass conversion accurately enough to reliably assess the glass production rate. This article summarizes two key aspects of the batch-conversion modeling—the heat transfer and the kinetics of conversion—and reviews the current state-of-the-art approaches to simulating them. We critically examine the advantages of the commonly used heat transfer approach, but also explain that its predictive capabilities are significantly restricted by the dependence of batch thermal properties on the time-temperature history. We argue that kinetic approaches to the batch-conversion modeling would offer a significant improvement when coupled with the heat transfer approach. Finally, we summarize key areas requiring further research on the way toward a realistic model of the batch blanket.

Pokorny, Richard↗

EXTENDED Ly{alpha} EMISSION AROUND STAR-FORMING GALAXIES

Ly{alpha} photons that escape the interstellar medium of star-forming galaxies may be resonantly scattered by neutral hydrogen atoms in the circumgalactic and intergalactic media, thereby increasing the angular extent of the galaxy's Ly{alpha} emission. We present predictions of this extended, low surface brightness Ly{alpha} emission based on radiative transfer modeling in a cosmological reionization simulation. The extended emission can be detected from stacked narrowband images of Ly{alpha} emitters (LAEs) or of Lyman break galaxies (LBGs). Its average surface brightness profile has a central cusp, then flattens to an approximate plateau beginning at an inner characteristic scale below {approx}0.2 Mpc (comoving), then steepens again beyond an outer characteristic scale of {approx}1 Mpc. The inner scale marks the transition from scattered light of the central source to emission from clustered sources, while the outer scale marks the spatial extent of scattered emission from these clustered sources. Both scales tend to increase with halo mass, UV luminosity, and observed Ly{alpha} luminosity. The extended emission predicted by our simulation is already within reach of deep narrowband photometry using large ground-based telescopes. Such observations would test radiative transfer models of emission from LAEs and LBGs, and they would open a new window on the circumgalactic environment of high-redshift star-forming galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗

FIRST SCATTERED-LIGHT IMAGE OF THE DEBRIS DISK AROUND HD 131835 WITH THE GEMINI PLANET IMAGER

We present the first scattered-light image of the debris disk around HD 131835 in the H band using the Gemini Planet Imager. HD 131835 is a ∼15 Myr old A2IV star at a distance of ∼120 pc in the Sco-Cen OB association. We detect the disk only in polarized light and place an upper limit on the peak total intensity. No point sources resembling exoplanets were identified. Compared to its mid-infrared thermal emission, in scattered light the disk shows similar orientation but different morphology. The scattered-light disk extends from ∼75 to ∼210 AU in the disk plane with roughly flat surface density. Our Monte Carlo radiative transfer model can describe the observations with a model disk composed of a mixture of silicates and amorphous carbon. In addition to the obvious brightness asymmetry due to stronger forward scattering, we discover a weak brightness asymmetry along the major axis, with the northeast side being 1.3 times brighter than the southwest side at a 3σ level.

79 ASTRONOMY AND ASTROPHYSICS↗

Transport, vertical structure and radiative properties of dust events in southeast China determined from ground and space sensors

Two dust events were detected over the Yangtze Delta region of China during March 14-17 and April 25-26 in 2009 where such dust events are uncommon. The transport behavior, spatio-temporal evolution, vertical structure, direct radiative effects, as well as induced heating rates, are investigated using a combination of ground-based and satellite-based measurements, a back-trajectory analysis, an aerosol model and a radiative transfer model.

54 ENVIRONMENTAL SCIENCES↗

A LARGE, MASSIVE, ROTATING DISK AROUND AN ISOLATED YOUNG STELLAR OBJECT

We present multi-wavelength observations and a radiative transfer model of a newly discovered massive circumstellar disk of gas and dust which is one of the largest disks known today. Seen almost edge-on, the disk is resolved in high-resolution near-infrared (NIR) images and appears as a dark lane of high opacity intersecting a bipolar reflection nebula. Based on molecular line observations, we estimate the distance to the object to be 3.5 kpc. This leads to a size for the dark lane of {approx}10,500 AU but due to shadowing effects the true disk size could be smaller. In Spitzer/IRAC 3.6 {mu}m images, the elongated shape of the bipolar reflection nebula is still preserved and the bulk of the flux seems to come from disk regions that can be detected due to the slight inclination of the disk. At longer IRAC wavelengths, the flux is mainly coming from the central regions penetrating directly through the dust lane. Interferometric observations of the dust continuum emission at millimeter wavelengths with the Submillimeter Array confirm this finding as the peak of the unresolved millimeter-emission coincides perfectly with the peak of the Spitzer/IRAC 5.8 {mu}m flux and the center of the dark lane seen in the NIR images. Simultaneously acquired CO data reveal a molecular outflow along the northern part of the reflection nebula which seems to be the outflow cavity. An elongated gaseous disk component is also detected and shows signs of rotation. The emission is perpendicular to the molecular outflow and thus parallel to but even more extended than the dark lane in the NIR images. Based on the dust continuum and the CO observations, we estimate a disk mass of up to a few solar masses depending on the underlying assumptions. Whether the disk-like structure is an actual accretion disk or rather a larger-scale flattened envelope or pseudodisk is difficult to discriminate with the current data set. The existence of HCO{sup +}/H{sup 13}CO{sup +} emission proves the presence of dense gas in the disk and the molecules' abundances are similar to those found in other circumstellar disks. We furthermore detected C{sub 2}H toward the objects and discuss this finding in the context of star formation. Finally, we have performed radiative transfer modeling of the K-band scattered light image varying a disk plus outflow two-dimensional density profile and the stellar properties. The model approximately reproduces extent and location of the dark lane, and the basic appearance of the outflow. We discuss our findings in the context of circumstellar disks across all mass regimes and conclude that our discovery is an ideal laboratory to study the early phases in the evolution of massive circumstellar disks surrounding young stellar objects.

79 ASTRONOMY AND ASTROPHYSICS↗

Data for Clumping Index Estimation With 30°-tilted Cameras in Row Crops: Evaluation of Methods and Segment Size Effects

The clumping index (CI) quantifies the spatial distribution of foliage elements and is essential for accurately estimating the plant area index (PAI), canopy radiative transfer, and photosynthesis. Traditionally, the finite-length averaging method (LX), the gap size distribution method (CC), and a combined approach of CC and LX (CLX) have been applied to instruments like TRAC and digital hemispherical photography to estimate CI. However, a comprehensive evaluation of these methods in row crops remains limited, especially regarding the influence of segment size on CI. Meanwhile, digital cameras offer a cost-effective and user-friendly solution for canopy measurements in row crops, yet their application in this context remains underexplored. In this study, we employed a new approach using a 30°-tilted digital camera to estimate CI in corn and soybean fields, applying the LX, CC, and CLX methods. We systematically assessed the performance of these three methods by combining field measurements in real-world fields with simulations using the LESS 3D radiative transfer model. Our results showed that CLX applied to the whole image and 45° segment offered accurate estimation of CI (bias within ±0.1, RMSE < 0.2) and PAI (bias within ±0.4, RMSE < 1) in real-world fields and LESS simulations. The accuracy of the LX method was highly sensitive to segment size, with the best performance observed at the 15° segment (PAI bias within ±0.4). In contrast, the CC method remained stable across different segment sizes, and its performance was generally comparable to that of LX, except at the 15° segment. Across view zenith angles, CI derived from CC generally showed a continuous increase, while those from LX and CLX followed a rising trend at small zenith angles but began to decline at 68°, likely due to an increasing proportion of no-gap segments. Seasonally, LX tended to show decreasing CI during early growth stages but increased as the canopy matured, whereas CC and CLX showed gradually increasing CI before plateauing at peak PAI. The 30°-tilted camera effectively captured CI variations across different angles and growth stages, making it a practical and robust instrument for row crop canopy structure analysis. Applying these CI methods to digital cameras offers a low-cost and accessible CI estimation alternative, improving canopy structure monitoring accuracy in row crops.

Modeling↗

Debris Disk Results from the Gemini Planet Imager Exoplanet Survey's Polarimetric Imaging Campaign

We report the results of a ∼4 yr direct imaging survey of 104 stars to resolve and characterize circumstellar debris disks in scattered light as part of the Gemini Planet Imager (GPI) Exoplanet Survey. We targeted nearby (≲150 pc), young (≲500 Myr) stars with high infrared (IR) excesses (L {sub IR}/L {sub ⋆} > 10{sup −5}), including 38 with previously resolved disks. Observations were made using the GPI high-contrast integral field spectrograph in H-band (1.6 μm) coronagraphic polarimetry mode to measure both polarized and total intensities. We resolved 26 debris disks and 3 protoplanetary/transitional disks. Seven debris disks were resolved in scattered light for the first time, including newly presented HD 117214 and HD 156623, and we quantified basic morphologies of five of them using radiative transfer models. All of our detected debris disks except HD 156623 have dust-poor inner holes, and their scattered-light radii are generally larger than corresponding radii measured from resolved thermal emission and those inferred from spectral energy distributions. To assess sensitivity, we report contrasts and consider causes of nondetections. Detections were strongly correlated with high IR excess and high inclination, although polarimetry outperformed total intensity angular differential imaging for detecting low-inclination disks (≲70°). Based on postsurvey statistics, we improved upon our presurvey target prioritization metric predicting polarimetric disk detectability. We also examined scattered-light disks in the contexts of gas, far-IR, and millimeter detections. Comparing H-band and ALMA fluxes for two disks revealed tentative evidence for differing grain properties. Finally, we found no preference for debris disks to be detected in scattered light if wide-separation substellar companions were present.

79 ASTRONOMY AND ASTROPHYSICS↗

Observation of the spectrally invariant properties of clouds in cloudy-to-clear transition zones during the MAGIC field campaign

We use the spectrally invariant method to study the variability of cloud optical thickness τ and droplet effective radius reff in transition zones (between the cloudy and clear sky columns) observed from Solar Spectral Flux Radiometer (SSFR) and Shortwave Array Spectroradiometer-Zenith (SASZe) during the Marine ARM GPCI Investigation of Clouds (MAGIC) field campaign. The measurements from the SSFR and the SASZe are different, however inter-instrument differences of self-normalized measurements (divided by their own spectra at a fixed time) are small. The spectrally invariant method approximates the spectra in the cloud transition zone as a linear combination of definitely clear and cloudy spectra, where the coefficients, slope and intercept, character-ize the spectrally invariant properties of the transition zone. Simulation results from the SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer) model demonstrate that (1) the slope of the visible band is positively correlated with the cloud optical thickness τ while the intercept of the near-infrared band has high negative cor-relation with the cloud drop effective radius reff even without the exact knowledge of τ; (2) the above relations hold for all Solar Zenith Angle (SZA) and for cloud-contaminated skies. In observations using redundant measure-ments from SSFR and SASZe, we find that during cloudy-to-clear transitions, (a) the slopes of the visible band de-crease, and (b) the intercepts of the near-infrared band remain almost constant near cloud edges. The findings in simulations and observations suggest that, while the optical thickness decreases during the cloudy-to-clear transition, the cloud drop effective radius does not change when cloud edges are approached. These results sup-port the hypothesis that inhomogeneous mixing dominates near cloud edges in the studied cases.

54 ENVIRONMENTAL SCIENCES↗

Early-time radiation flux symmetry optimization and its effect on gas-filled hohlraum ignition targets on the National Ignition Facility

Achieving ignition on the National Ignition Facility (NIF) is tied to our ability to control and minimize deviations from sphericity of the capsule implosion. Low-mode asymmetries of the hot spot result from the combined effect of radiation drive asymmetries throughout the laser pulse and initial roughness on the capsule surface. In this paper, we report on simulations and experiments designed to assess, measure, and correct the drive asymmetries produced by the early-time (≈first 2 ns or “picket”) period of the laser pulse. The drive asymmetry during the picket is commonly thought to introduce distortions in the hot-spot shape at ignition time. However, a more subtle effect not previously considered is that it also leads to an asymmetry in shock velocity and timing, thereby increasing the fuel adiabat and reducing the margin for ignition. It is shown via hydrodynamic simulations that minimizing this effect requires that the early-time asymmetry be kept below 7.5% in the second Legendre mode (P{sub 2}), thus keeping the loss of performance margin below ≈10% for a layered implosion. Asymmetries during the picket of the laser pulse are measured using the instantaneous self-emission of a high-Z re-emission sphere in place of an ignition capsule in a hohlraum with large azimuthal diagnostic windows. Three dimensional simulations using the code HYDRA (to capture the effect of non-azimuthal hohlraum features) coupled to a cross-beam energy transfer model [Michel et al., Phys. Plasmas 17, 056305 (2010)] are used to establish the surrogacy of the re-emit target and to assess the early-time drive symmetry. Calculations using this model exhibit the same sensitivity to variations in the relative input powers between the different cones of NIF beams as measured for the “Rev5” CH target [Haan et al., Phys Plasmas 18, 051001 (2011)] and reported by Dewald et al. [Phys. Rev. Lett. 111, 235001 (2013)]. The same methodology applied to recently improved implosions using different hohlraum geometries and picket powers show good agreement with experimental data.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Recent Updates to NRC Fuel Performance Codes and Plans for Future Improvements

FRAPCON-3.4a and FRAPTRAN 1.4 are the most recent versions of the U.S. Nuclear Regulatory Commission (NRC) steady-state and transient fuel performance codes, respectively. These codes have been assessed against separate effects data and integral assessment data and have been determined to provide a best estimate calculation of fuel performance. Recent updates included in FRAPCON-3.4a include updated material properties models, models for new fuel and cladding types, cladding finite element analysis capability, and capability to perform uncertainty analyses and calculate upper tolerance limits for important outputs. Recent updates included in FRAPTRAN 1.4 include: material properties models that are consistent with FRAPCON-3.4a, cladding failure models that are applicable for loss-of coolant-accident and reactivity initiated accident modeling, and updated heat transfer models. This paper briefly describes these code updates and data assessments, highlighting the particularly important improvements and data assessments. This paper also discusses areas of improvements that will be addressed in upcoming code versions.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

The CU 2-D-MAX-DOAS instrument – Part 2: Raman scattering probability measurements and retrieval of aerosol optical properties

Abstract. The multiannual global mean of aerosol optical depth at 550 nm (AOD550) over land is ∼ 0.19, and that over oceans is ∼ 0.13. About 45 % of the Earth surface shows AOD550 smaller than 0.1. There is a need for measurement techniques that are optimized to measure aerosol optical properties under low AOD conditions. We present an inherently calibrated retrieval (i.e., no need for radiance calibration) to simultaneously measure AOD and the aerosol phase function parameter, g, based on measurements of azimuth distributions of the Raman scattering probability (RSP), the near-absolute rotational Raman scattering (RRS) intensity. We employ radiative transfer model simulations to show that for solar azimuth RSP measurements at solar elevation and solar zenith angle (SZA) smaller than 80°, RSP is insensitive to the vertical distribution of aerosols and maximally sensitive to changes in AOD and g under near-molecular scattering conditions. The University of Colorado two-dimensional Multi-AXis Differential Optical Absorption Spectroscopy (CU 2-D-MAX-DOAS) instrument was deployed as part of the Two Column Aerosol Project (TCAP) at Cape Cod, MA, during the summer of 2012 to measure direct sun spectra and RSP from scattered light spectra at solar relative azimuth angles (SRAAs) between 5 and 170°. During two case study days with (1) high aerosol load (17 July, 0.3 < AOD430 < 0.6) and (2) near-molecular scattering conditions (22 July, AOD430 < 0.13) we compare RSP-based retrievals of AOD430 and g with data from a co-located CIMEL sun photometer, Multi-Filter Rotating Shadowband Radiometer (MFRSR), and an airborne High Spectral Resolution Lidar (HSRL-2). The average difference (relative to DOAS) for AOD430 is +0.012 ± 0.023 (CIMEL), −0.012 ± 0.024 (MFRSR), −0.011 ± 0.014 (HSRL-2), and +0.023 ± 0.013 (CIMELAOD − MFRSRAOD) and yields the following expressions for correlations between different instruments: DOASAOD = −(0.019 ± 0.006) + (1.03 ± 0.02) × CIMELAOD (R2 = 0.98), DOASAOD = −(0.006 ± 0.005) + (1.08 ± 0.02) × MFRSRAOD (R2 = 0.98), and CIMELAOD = (0.013 ± 0.004) + (1.05 ± 0.01) × MFRSRAOD (R2 = 0.99). The average g measured by DOAS on both days was 0.66 ± 0.03, with a difference of 0.014 ± 0.05 compared to CIMEL. Active steps to minimize the error in the RSP help to reduce the uncertainty in retrievals of AOD and g. As AOD decreases and SZA increases, the RSP signal-to-noise ratio increases. At AOD430 ∼ 0.4 and 0.10 the absolute AOD errors are ∼ 0.014 and 0.003 at 70° SZA and 0.02 and 0.004 at 35° SZA. Inherently calibrated, precise AOD and g measurements are useful to better characterize the aerosol direct effect in urban polluted and remote pristine environments.

54 ENVIRONMENTAL SCIENCES↗

State-of-the-art on research and applications of machine learning in the building life cycle

Fueled by big data, powerful and affordable computing resources, and advanced algorithms, machine learning has been explored and applied to buildings research for the past decades and has demonstrated its potential to enhance building performance. This study systematically surveyed how machine learning has been applied at different stages of building life cycle. By conducting a literature search on the Web of Knowledge platform, we found 9579 papers in this field and selected 153 papers for an in-depth review. The number of published papers is increasing year by year, with a focus on building design, operation, and control. However, no study was found using machine learning in building commissioning. There are successful pilot studies on fault detection and diagnosis of HVAC equipment and systems, load prediction, energy baseline estimate, load shape clustering, occupancy prediction, and learning occupant behaviors and energy use patterns. None of the existing studies were adopted broadly by the building industry, due to common challenges including (1) lack of large scale labeled data to train and validate the model, (2) lack of model transferability, which limits a model trained with one data-rich building to be used in another building with limited data, (3) lack of strong justification of costs and benefits of deploying machine learning, and (4) the performance might not be reliable and robust for the stated goals, as the method might work for some buildings but could not be generalized to others. Finally, findings from the study can inform future machine learning research to improve occupant comfort, energy efficiency, demand flexibility, and resilience of buildings, as well as to inspire young researchers in the field to explore multidisciplinary approaches that integrate building science, computing science, data science, and social science.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗