SEARCH · Search NASA

Results for “Modeling and Performance Analysis”

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 577 records · Page 32

All sources NASA NTRS DOE OSTI

NASA NTRS1977

A study of key features of random atmospheric disturbance models for the approach flight phase

An analysis and brief simulator experiment were performed to identify and classify important features of random turbulence for the landing approach flight phase. The analysis of various wind models was carried out within the context of the longitudinal closed-loop pilot/vehicle system. The analysis demonstrated the relative importance of atmospheric disturbance scale lengths, horizontal versus vertical gust components, decreasing altitude, and spectral forms of disturbances versus the pilot/vehicle system. Among certain competing wind models, the analysis predicted no significant difference in pilot performance. This was confirmed by a moving base simulator experiment which evaluated the two most extreme models. A number of conclusions were reached: attitude constrained equations do provide a simple but effective approach to describing the closed-loop pilot/vehicle. At low altitudes the horizontal gust component dominates pilot/vehicle performance.

Heffley, R. K.↗

DOE OSTI2022

Finding Hidden Patterns in High Resolution Wind Flow Model Simulations

Wind flow data is critical in terms of investment decisions and policy making. High resolution data from wind flow model simulations serve as a supplement to the limited resource of original wind flow data collection. Given the large size of data, finding hidden patterns in wind flow model simulations are critical for reducing the dimensionality of the analysis. In this work, we first perform dimension reduction with two autoencoder models: the CNN-based autoencoder (CNN-AE) [1], and hierarchical autoencoder (HIER-AE) [2], and compare their performance with the Principal Component Analysis (PCA). We then investigate the super-resolution of the wind flow data. By training a Generative Adversarial Network (GAN) with 300 epochs, we obtained a trained model with 2× resolution enhancement. We compare the results of GAN with Convolutional Neural Network (CNN), and GAN results show finer structure as expected in the data field images. Also, the kinetic energy spectra comparisons show that GAN outperforms CNN in terms of reproducing the physical properties for high wavenumbers and is critical for analysis where high-wavenumber kinetics play an important role.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Parametric and Sensitivity Analysis of a Steam Generator Model Using Python and Machine-Learning Tools

For this study, we used Python and machine-learning tools to perform a comprehensive parametric and sensitivity analysis on a steam generator (SG) model. (The Python model was based on a previously completed MATLAB framework for the Holtec SMR-160 SG.) We investigated the influence of various input parameters (e.g., heat transfer coefficient [HTC], Nusselt number, and heat exchanger effectiveness) on the system’s output. With machine-learning tools such as the Risk Analysis Virtual Environment (RAVEN), which was developed at Idaho National Laboratory, we were then able to perform an automated analysis of the SG inputs’ effect on the HTC. The analysis results give valuable insights into the performance and optimization of SG systems. We found the inlet mass flow rate (MFR) to have the greatest impact on the HTC, followed closely by the inlet temperature, and then pressure. Shifting of the input parameters causes the location of the maximum HTC along the SG length to change incrementally. The cold leg (CL) MFR was also found to impact the HTC magnitude as well as the location of the maximum HTC. At between 0.4–0.9 of the total SG length, the input parameters experience maximum impact on the HTC, leading us to suggest that sensors be efficiently placed on the SG so as to closely and effectively monitor thermal-hydraulic properties during reactor operation. We also found that the sensitivity data calculated manually agrees with the RAVEN – based data, confirming the same range of maximum sensitivity. However, the RAVEN-based analysis showed that cold leg pressure and hot leg temperature have a greater impact on the heat transfer coefficient than the mass flow rate, implying that a manual sensitivity study taking only two samples is not accurate.

20 FOSSIL-FUELED POWER PLANTS↗

DOE OSTI2024

Comparing net zero pathways across the Atlantic A model inter-comparison exercise between the Energy Modeling Forum 37 and the European Climate and Energy Modeling Forum

Europe and North America account for 32 % of current carbon emissions. Due to distinct legacy systems, energy infrastructure, socioeconomic development, and energy resource endowment, both regions have different policy and technological pathways to reach net zero by the mid-century. Against this background, our paper examines the results from the net zero emission scenarios for Europe and North America that emerged from the collaboration of the European and American Energy Modeling Forums. Here, in our analysis, we perform an inter-comparison of various integrated assessments and bottom-up energy system models. A clear qualitative consensus emerges on five main points. First, Europe and the United States reach net zero targets with electrification, demand-side reductions, and carbon capture and sequestration technologies. Second, the use of carbon capture and sequestration is more predominant in the United States due to a steeper decarbonization schedule. Third, the buildings sector is the easiest to electrify in both regions. Fourth, the industrial sector is the hardest to electrify in the United States and transportation in Europe. Fifth, in both regions, the transition in the energy mix is driven by the substitution of coal and natural gas with solar and wind, but to a different extent.

100 % renewables↗

NASA NTRSDate not supplied

A Simplified Model of VIPER Thermal Management System. Part I: Loop Heat Pipe

NASA’s Volatiles Investigating Polar Exploration Rover (VIPER) uses a combination of loop heat pipes (LHPs) and heaters as the primary hardware to modulate the temperature of the electronic boxes and payloads instruments. LHPs have been designed and instrumented to be capable of passive and active shutdown. The goal of the passive shutdown, using a thermal control valve (TCV) to limit the flow of the working fluid, is to preserve thermal energy and avoid unnecessary heat leaks during lunar night. The goal of the active shutdown on the other hand, is to forcefully increase the vehicle internal temperature to bakeout the payload instruments. Several numerical modeling tools are available to analyze LHP performance under the complex set of operational requirements and environments levied on the VIPER Thermal Management System (TMS). Yet, while those numerical modeling tools are capable of detailed performance analysis, they tend to take large computational resources and long computational time. In this paper, a model of a single LHP developed based on control volume approach is presented. By using the control volume approach, the complex geometry of the LHP, and the components attached to it, are reduced to thermal lump capacitances, thus reducing the computation effort. This paper also describes the effort to correlate the proposed LHP model using thermal vacuum (TVAC) test data.

Loop Heat Pipe↗

NASA NTRSDate not supplied

A Simplified Model of VIPER Thermal Management System. Part I: Loop Heat Pipe

Loop Heat Pipe↗

DOE OSTI2022

Investigating key cell types and molecules dynamics in PyMT mice model of breast cancer through a mathematical model

The most common kind of cancer among women is breast cancer. Understanding the tumor microenvironment and the interactions between individual cells and cytokines assists us in arriving at more effective treatments. Here, we develop a data-driven mathematical model to investigate the dynamics of key cell types and cytokines involved in breast cancer development. We use time-course gene expression profiles of a mouse model to estimate the relative abundance of cells and cytokines. We then employ a least-squares optimization method to evaluate the model’s parameters based on the mice data. The resulting dynamics of the cells and cytokines obtained from the optimal set of parameters exhibit a decent agreement between the data and predictions. We perform a sensitivity analysis to identify the crucial parameters of the model and then perform a local bifurcation on them. The results reveal a strong connection between adipocytes, IL6, and the cancer population, suggesting them as potential targets for therapies.

59 BASIC BIOLOGICAL SCIENCES↗

NASA NTRS1975

Analysis of pilot-aircraft performance and reliability via the application of man-machine models

The present approach to analytic modeling, which utilizes human response theory together with modern control theory, is discussed. Analytic modeling of human performance has progressed to the point where it can be used with some confidence in the design and performance analysis of man-machine systems. An example of its use in helping design a flight director for the hover control of a VTOL vehicle is described. How the method could be used for analysis of human response to system failure is sketched.

Kleinman, D. L.↗

DOE OSTI2017

Performance Assessment of a New Gamma Spectroscopy Toolkit: Peak Doctor and the Spectral Nondestructive Assay Platform (SNAP) - 17397

High resolution gamma spectroscopy (HRGS) is a widely used technique for non-destructive characterization of nuclear materials, and assay of radioactive waste. The ability to accurately evaluate raw spectral data is an important step in producing high-quality results. A spectral analysis application known as Peak Doctor has been developed under a Cooperative Research and Development Agreement (CRADA) between LANL and Pajarito Scientific Corporation (PSC). The software has undergone several years of testing and validation by the Radioactive Waste Characterization Team of the LANL Waste Management Services Group (WM-SVS). PeakDoctor is designed to work with the Spectral Nondestructive Assay Platform (SNAP{sup TM}) to provide a complete HRGS analysis engine. SNAP{sup TM} performs model specific corrections for attenuation and geometry. These applications are the major components of a new toolkit designed primarily for portable far-field non-destructive assay. Peak Doctor performs a very accurate peak fit analysis with the goal of improving upon the capabilities of existing commercially available routines. It is designed to be user-configurable and can be customized to any given type of detector and any given type of spectrum. Peak Doctor was developed to resolve many of the inherent complexities in gamma ray spectra including overlapping peaks, complex multiplet regions, large step peaks, backscatter regions and Compton edges. Unlike most spectral analysis platforms, Peak Doctor does not rely on the energy calibration and detector resolution parameters that are saved by the operator in the spectrum file. Instead, key parameters are pre-defined for individual detectors. The software fits a continuum using a series of splines between a series of junctions or 'knots'. Fitting techniques are used to join the splines together to create a near-continuous multi-element curve. Optimizing the location of the knots is one of Peak Doctor's most powerful features. The knots are initially configured using a pre-defined list and are removed or added automatically during the fitting process. Knots are also created when new features are identified such as Step Peaks and Compton Edges. The user can also manually specify knots in order to improve the continuum fit. The initial fit is improved on by removing any outlier regions (which are converted to peaks) and the continuum fitting process is repeated. At this stage, an experienced user can work interactively with the fit by manually converting peak regions to continuum or vice versa. The final step is to fit a series of peaks above the continuum based on detector parameters. Single and multiplet peaks are then processed by either the tailing (piecewise) or non-tailing (Gaussian) Levenberg Marquardt non-linear peak fitting routine depending on the user defined selection. Only a rough knowledge of the detector characteristics (energy calibration and resolution) is required for the fitting process to proceed with good accuracy. Any peaks that are missed can be manually added by the user and will also be fitted. Discontinuities in the peak / continuum fit can be rapidly identified by reviewing the 'residuals scale' plot. For waste streams that are well characterized, it is often the case that routinely encountered spectra will have similar peaks and continuum features. For these cases a handy 'One Step' method is available that proceeds in an automated fashion without user interaction. Peak Doctor has been configured with a set of default spectrum types including plutonium / transuranic waste (including Weapons Grade Pu, Heat Source Pu, Am-241, Am-243 and Np-237), depleted or enriched uranium, nuclear power plant / intermediate level waste (with fission and activation products such as Cs-137 and Co-60) as well as Naturally Occurring Radioactive Materials (NORM). The user can easily add new spectrum types for specific waste streams or other types of radiological / special nuclear materials. The output from PeakDoctor (including net peak area, centroid energy and a list of minimum detectable activity levels at pre-defined regions) is used by SNAP{sup TM} to complete the data analysis process. Isotope activities are calculated using configurable template-driven geometry and attenuation correction methods. The software accounts for matrix and container wall effects, collimator / angular corrections, detector efficiency (based on calibration curves) and stand-off distance. SNAP{sup TM} uses published algorithms incorporating a series of decision points during its operation to allow the analyst to provide logical input to guide the software during the analysis of each unique spectrum. The Peak Doctor / SNAP{sup TM} toolkit has been extensively tested and validated under a variety of conditions using 'benchmark' reference spectra. Test scenarios reported include HPGe-based measurements of well characterized uranium and plutonium standards and radioactive sources with a variety of different geometries and matrix materials including surrogate drums and boxes. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2018

Optimization study for the experimental configuration of CMB-S4

The CMB Stage 4 (CMB-S4) experiment is a next-generation, ground-based experiment that will measure the cosmic microwave background (CMB) polarization to unprecedented accuracy, probing the signature of inflation, the nature of cosmic neutrinos, relativistic thermal relics in the early universe, and the evolution of the universe. CMB-S4 will consist of O (500,000) photon-noise-limited detectors that cover a wide range of angular scales in order to probe the cosmological signatures from both the early and late universe. It will measure a wide range of microwave frequencies to cleanly separate the CMB signals from galactic and extra-galactic foregrounds. To advance the progress towards designing the instrument for CMB-S4, we have established a framework to optimize the instrumental configuration to maximize its scientific output. The framework combines cost and instrumental models with a cosmology forecasting tool, and evaluates the scientific sensitivity as a function of various instrumental parameters. The cost model also allows us to perform the analysis under a fixed-cost constraint, optimizing for the scientific output of the experiment given finite resources. In this paper, we report our first results from this framework, using simplified instrumental and cost models. Here, we have primarily studied two classes of instrumental configurations: arrays of large-aperture telescopes with diameters ranging from 2–10 m, and hybrid arrays that combine small-aperture telescopes (0.5-m diameter) with large-aperture telescopes. We explore performance as a function of telescope aperture size, distribution of the detectors into different microwave frequencies, survey strategy and survey area, low-frequency noise performance, and balance between small and large aperture telescopes for hybrid configurations. Both types of configurations must cover both large (~ degree) and small (~ arcmin) angular scales, and the performance depends on assumptions for performance vs. angular scale. The configurations with large-aperture telescopes have a shallow optimum around 4–6 m in aperture diameter, assuming that large telescopes can achieve good performance for low-frequency noise. We explore some of the uncertainties of the instrumental model and cost parameters, and we find that the optimum has a weak dependence on these parameters. The hybrid configuration shows an even broader optimum, spanning a range of 4–10 m in aperture for the large telescopes. We also present two strawperson configurations as an outcome of this optimization study, and we discuss some ideas for improving our simple cost and instrumental models used here. There are several areas of this analysis that deserve further improvement. In our forecasting framework, we adopt a simple two-component foreground model with spatially varying power-law spectral indices. We estimate de-lensing performance statistically and ignore non-idealities such as anisotropic mode coverage, boundary effect, and possible foreground residual. Instrumental systematics, which is not accounted for in our analyses, may also influence the conceptual design. Further study of the instrumental and cost models will be one of the main areas of study by the entire CMB-S4 community. Finally, we hope that our framework will be useful for estimating the influence of these improvements in the future, and we will incorporate them in order to further improve the optimization.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2025

A Model Based Approach to Extract Health Information from Textual Data

In current nuclear power plants (NPPs) a large amount of condition-based data is being generated and stored to assess and monitor component health and performance. The format of this data can be either numeric (e.g., pump vibration data) or textual (e.g., condition report which assess component health). While assessing component health from numeric data can be performed with a large variety of methods, the extraction of information from textual data still remains a challenge. Natural language processing (NLP) methods are starting to be deployed in current NPPs mainly to filter out incident reports (IRs) that are not safety related by employing supervised machine learning methods. However, these methods do not really provide the quantitative information that might be contained in IRs. This paper presents an approach to extract information from textual data (e.g., from IRs, maintenance reports) that is based on NLP data analytics methods coupled with model-based system engineer (MBSE) models. NLP methods are employed to perform syntactic and semantic analyses. Syntactic analysis analyzes the grammatical structure of a sentence; such analysis includes: part of speech (POS) tagging (i.e., identification of grammatic elements of each string - e.g., nouns, verbs), named entity recognition (i.e., identification of text entities - e.g., names, dates, events), and relation extraction (e.g., coreference resolution). On the other hand, semantic analysis is designed to analyze the logic structure of a sentence. Through a specific set of rules, our methods can identify whether a sentence contains health information of a component (e.g., degraded performance, anomaly behavior) or the causal relationship between two events (i.e., a cause-effect pair). An innovative element of our approach is that semantic analysis relies on MBSE models to identify links between textual elements. MBSE are diagrams designed to represent system and component dependencies (from both a form and functional point of view). In our approach, MBSE models emulate system engineer knowledge about component/system architecture. This paper presents in detail how the integration of NLP methods and MBSE models is performed. Few analysis examples focusing on centrifugal pumps are presented.

97 - MATHEMATICS AND COMPUTING↗

DOE OSTI2016

Wind Farm Flow Modeling using an Input-Output Reduced-Order Model

Wind turbines in a wind farm operate individually to maximize their own power regardless of the impact of aerodynamic interactions on neighboring turbines. There is the potential to increase power and reduce overall structural loads by properly coordinating turbines. To perform control design and analysis, a model needs to be of low computational cost, but retains the necessary dynamics seen in high-fidelity models. The objective of this work is to obtain a reduced-order model that represents the full-order flow computed using a high-fidelity model. A variety of methods, including proper orthogonal decomposition and dynamic mode decomposition, can be used to extract the dominant flow structures and obtain a reduced-order model. In this paper, we combine proper orthogonal decomposition with a system identification technique to produce an input-output reduced-order model. This technique is used to construct a reduced-order model of the flow within a two-turbine array computed using a large-eddy simulation.

17 WIND ENERGY↗

NASA NTRS1994

Control of maglev vehicles with aerodynamic and guideway disturbances

A modeling, analysis, and control design methodology is presented for maglev vehicle ride quality performance improvement as measured by the Pepler Index. Ride quality enhancement is considered through active control of secondary suspension elements and active aerodynamic surfaces mounted on the train. To analyze and quantify the benefits of active control, the authors have developed a five degree-of-freedom lumped parameter model suitable for describing a large class of maglev vehicles, including both channel and box-beam guideway configurations. Elements of this modeling capability have been recently employed in studies sponsored by the U.S. Department of Transportation (DOT). A perturbation analysis about an operating point, defined by vehicle and average crosswind velocities, yields a suitable linearized state space model for multivariable control system analysis and synthesis. Neglecting passenger compartment noise, the ride quality as quantified by the Pepler Index is readily computed from the system states. A statistical analysis is performed by modeling the crosswind disturbances and guideway variations as filtered white noise, whereby the Pepler Index is established in closed form through the solution to a matrix Lyapunov equation. Data is presented which indicates the anticipated ride quality achieved through various closed-loop control arrangements.

Flueckiger, Karl↗

DOE OSTI2024

Training custom light curve models of SN Ia subpopulations selected according to host galaxy properties

ABSTRACT Type Ia supernova (SN Ia) cosmology analyses include a luminosity step function in their distance standardization process to account for an observed yet unexplained difference in the post-standardization luminosities of SNe Ia originating from different host galaxy populations [e.g. high-mass ($M \gtrsim 10^{10} \, {\rm M}_{\odot }$) versus low-mass galaxies]. We present a novel method for including host-mass correlations in the SALT3 (Spectral Adaptive Light curve Template 3) light curve model used for standardizing SN Ia distances. We split the SALT3 training sample according to host-mass, training independent models for the low- and high-host-mass samples. Our models indicate that there are different average Si ii spectral feature strengths between the two populations, and that the average spectral energy distribution of SNe from low-mass galaxies is bluer than the high-mass counterpart. We then use our trained models to perform an SN cosmology analysis on the 3-yr spectroscopically confirmed Dark Energy Survey SN sample, treating SNe from low- and high-mass host galaxies as separate populations throughout. We find that our mass-split models reduce the Hubble residual scatter in the sample, albeit at a low statistical significance. We do find a reduction in the mass-correlated luminosity step but conclude that this arises from the model-dependent re-definition of the fiducial SN absolute magnitude rather than the models themselves. Our results stress the importance of adopting a standard definition of the SN parameters (x0, x1, c) in order to extract the most value out of the light curve modelling tools that are currently available and to correctly interpret results that are fit with different models.

Taylor, G. (ORCID:0000000157563259)↗

DOE OSTI2017

The Herschel Planetary Nebula Survey (HerPlaNS): A Comprehensive Dusty Photoionization Model of NGC6781

We perform a comprehensive analysis of the planetary nebula (PN) NGC 6781 to investigate the physical conditions of each of its ionized, atomic, and molecular gas and dust components and the object’s evolution, based on panchromatic observational data ranging from UV to radio. Empirical nebular elemental abundances, compared with theoretical predictions via nucleosynthesis models of asymptotic giant branch (AGB) stars, indicate that the progenitor is a solar-metallicity, 2.25--3.0 M{sub ⊙} initial-mass star. We derive the best-fit distance of 0.46 kpc by fitting the stellar luminosity (as a function of the distance and effective temperature of the central star) with the adopted post-AGB evolutionary tracks. Our excitation energy diagram analysis indicates high-excitation temperatures in the photodissociation region (PDR) beyond the ionized part of the nebula, suggesting extra heating by shock interactions between the slow AGB wind and the fast PN wind. Through iterative fitting using the Cloudy code with empirically derived constraints, we find the best-fit dusty photoionization model of the object that would inclusively reproduce all of the adopted panchromatic observational data. The estimated total gas mass (0.41 M{sub ⊙}) corresponds to the mass ejected during the last AGB thermal pulse event predicted for a 2.5 M{sub ⊙} initial-mass star. A significant fraction of the total mass (about 70%) is found to exist in the PDR, demonstrating the critical importance of the PDR in PNe that are generally recognized as the hallmark of ionized/H{sup +} regions.

79 ASTRONOMY AND ASTROPHYSICS↗

NASA NTRS2017

Computational and Experimental Fluid-Structure Interaction Analysis of a High-Lift Wing with a Slat-Cove Filler for Noise Reduction

Airframe noise is a significant component of overall noise produced by transport aircraft during landing and approach (low speed maneuvers). A significant source for this noise is the cove of the leading-edge slat. The slat-cove filler (SCF) has been shown to be effective at mitigating slat noise. The objective of this work is to understand the fluid-structure interaction (FSI) behavior of a superelastic shape memory alloy (SMA) SCF in flow using both computational and physical models of a high-lift wing. Initial understanding of flow around the SCF and wing is obtained using computational fluid dynamics (CFD) analysis at various angles of attack. A framework compatible with an SMA constitutive model (implemented as a user material subroutine) is used to perform FSI analysis for multiple flow and configuration cases. A scaled physical model of the high-lift wing is constructed and tested in the Texas A&M 3 ft-by-4-foot wind tunnel. Initial validation of both CFD and FSI analysis is conducted by comparing lift, drag and pressure distributions with experimental results.

Scholten, William D.↗

NASA NTRS1993

Convection equation modeling: A non-iterative direct matrix solution algorithm for use with SINDA

The determination of the boundary conditions for a component-level analysis, applying discrete finite element and finite difference modeling techniques often requires an analysis of complex coupled phenomenon that cannot be described algebraically. For example, an analysis of the temperature field of a coldplate surface with an integral fluid loop requires a solution to the parabolic heat equation and also requires the boundary conditions that describe the local fluid temperature. However, the local fluid temperature is described by a convection equation that can only be solved with the knowledge of the locally-coupled coldplate temperatures. Generally speaking, it is not computationally efficient, and sometimes, not even possible to perform a direct, coupled phenomenon analysis of the component-level and boundary condition models within a single analysis code. An alternative is to perform a disjoint analysis, but transmit the necessary information between models during the simulation to provide an indirect coupling. For this approach to be effective, the component-level model retains full detail while the boundary condition model is simplified to provide a fast, first-order prediction of the phenomenon in question. Specifically for the present study, the coldplate structure is analyzed with a discrete, numerical model (SINDA) while the fluid loop convection equation is analyzed with a discrete, analytical model (direct matrix solution). This indirect coupling allows a satisfactory prediction of the boundary condition, while not subjugating the overall computational efficiency of the component-level analysis. In the present study a discussion of the complete analysis of the derivation and direct matrix solution algorithm of the convection equation is presented. Discretization is analyzed and discussed to extend of solution accuracy, stability and computation speed. Case studies considering a pulsed and harmonic inlet disturbance to the fluid loop are analyzed to assist in the discussion of numerical dissipation and accuracy. In addition, the issues of code melding or integration with standard class solvers such as SINDA are discussed to advise the user of the potential problems to be encountered.

Schrage, Dean S.↗

DOE OSTI2025

Deep Learning for Subsurface Flow: A Comparative Study of U‐Net, Fourier Neural Operators, and Transformers in Underground Hydrogen Storage

Subsurface flow research is essential for the sustainable management of natural resources and the environment. Deep learning (DL) has significantly advanced this field by developing efficient and accurate surrogate models to replace computationally expensive physics‐based simulations. These surrogate models are commonly used to predict the spatiotemporal evolution of state variables, such as gas saturation and reservoir pressure, in heterogeneous geological formations. Despite the various DL models applied to this task, there is a lack of studies systematically comparing their performance. This absence of comparative analysis leads to somewhat arbitrary DL model selection in subsurface flow research, resulting in suboptimal performance and potentially inaccurate predictions. To bridge this gap, we conduct a systematic comparison study of three popular DL architectures—U‐Net, Fourier Neural Operators (FNO), and Segmentation Transformer (SETR)—in surrogate modeling of underground hydrogen storage (UHS). We focus on UHS due to its promise of enhancing clean energy resilience and its cyclic operational conditions that represent common scenarios in various subsurface applications. We evaluate the models based on accuracy, training cost, and inference speed. The comparison shows that U‐Net achieves the highest accuracy, followed by SETR and FNO. Despite its lower accuracy, FNO has the highest inference speed. SETR offers competitive accuracy with the least training memory usage, demonstrating the potential of transformers in learning subsurface flow. Our results provide guidance for selecting DL models for surrogate modeling in a wide range of subsurface flow problems.

42 ENGINEERING↗