Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “surrogate approximator”

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 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
NASA NTRS2009

Defining Nitrogen Kinetics for Air Break in Prebreathe

Actual tissue nitrogen (N2) kinetics are complex; the uptake and elimination is often approximated with a single half-time compartment in statistical descriptions of denitrogenation [prebreathe (PB)] protocols. Air breaks during PB complicate N2 kinetics. A comparison of symmetrical versus asymmetrical N2 kinetics was performed using the time to onset of hypobaric decompression sickness (DCS) as a surrogate for actual venous N2 tension. Published results of 12 tests involving 179 hypobaric exposures in altitude chambers after PB, with and without air breaks, provide the complex protocols from which to model N2 kinetics. DCS survival time for combined control and air breaks were described with an accelerated log logistic model where N2 uptake and elimination before, during, and after the air break was computed with a simple exponential function or a function that changed half-time depending on ambient N2 partial pressure. P1N2-P2 = delta P defined DCS dose for each altitude exposure, where P2 was the test altitude and P1N2 was computed N2 pressure at the beginning of the altitude exposure. The log likelihood (LL) without DCS dose (null model) was -155.6, and improved (best-fit) to -97.2 when dose was defined with a 240 min half-time for both N2 elimination and uptake during the PB. The description of DCS survival time was less precise with asymmetrical N2 kinetics, for example, LL was -98.9 with 240 min half-time elimination and 120 min half-time uptake. The statistical regression described survival time mechanistically linked to symmetrical N2 kinetics during PBs that also included air breaks. The results are data-specific, and additional data may change the conclusion. The regression is useful to compute additional PB time to compensate for an air break in PB within the narrow range of tested conditions.

Conkin, Johnny↗
NASA NTRS2010

Defining Nitrogen Kinetics for Air Break in Prebreath

Actual tissue nitrogen (N2) kinetics are complex; the uptake and elimination is often approximated with a single half-time compartment in statistical descriptions of denitrogenation [prebreathe(PB)] protocols. Air breaks during PB complicate N2 kinetics. A comparison of symmetrical versus asymmetrical N2 kinetics was performed using the time to onset of hypobaric decompression sickness (DCS) as a surrogate for actual venous N2 tension. METHODS: Published results of 12 tests involving 179 hypobaric exposures in altitude chambers after PB, with and without airbreaks, provide the complex protocols from which to model N2 kinetics. DCS survival time for combined control and airbreaks were described with an accelerated log logistic model where N2 uptake and elimination before, during, and after the airbreak was computed with a simple exponential function or a function that changed half-time depending on ambient N2 partial pressure. P1N2-P2 = (Delta)P defined decompression dose for each altitude exposure, where P2 was the test altitude and P1N2 was computed N2 pressure at the beginning of the altitude exposure. RESULTS: The log likelihood (LL) without decompression dose (null model) was -155.6, and improved (best-fit) to -97.2 when dose was defined with a 240 min half-time for both N2 elimination and uptake during the PB. The description of DCS survival time was less precise with asymmetrical N2 kinetics, for example, LL was -98.9 with 240 min half-time elimination and 120 min half-time uptake. CONCLUSION: The statistical regression described survival time mechanistically linked to symmetrical N2 kinetics during PBs that also included airbreaks. The results are data-specific, and additional data may change the conclusion. The regression is useful to compute additional PB time to compensate for an airbreak in PB within the narrow range of tested conditions.

Conkin, Johnny↗
NASA NTRS1990

Surrogate oracles, generalized dependency and simpler models

Software reliability models require the sequence of interfailure times from the debugging process as input. It was previously illustrated that using data from replicated debugging could greatly improve reliability predictions. However, inexpensive replication of the debugging process requires the existence of a cheap, fast error detector. Laboratory experiments can be designed around a gold version which is used as an oracle or around an n-version error detector. Unfortunately, software developers can not be expected to have an oracle or to bear the expense of n-versions. A generic technique is being investigated for approximating replicated data by using the partially debugged software as a difference detector. It is believed that the failure rate of each fault has significant dependence on the presence or absence of other faults. Thus, in order to discuss a failure rate for a known fault, the presence or absence of each of the other known faults needs to be specified. Also, in simpler models which use shorter input sequences without sacrificing accuracy are of interest. In fact, a possible gain in performance is conjectured. To investigate these propositions, NASA computers running LIC (RTI) versions are used to generate data. This data will be used to label the debugging graph associated with each version. These labeled graphs will be used to test the utility of a surrogate oracle, to analyze the dependent nature of fault failure rates and to explore the feasibility of reliability models which use the data of only the most recent failures.

Wilson, Larry↗
DOE OSTI2024

Assessing the reliability of medical resource demand models in the context of COVID-19

Abstract Background Numerous medical resource demand models have been created as tools for governments or hospitals, aiming to predict the need for crucial resources like ventilators, hospital beds, personal protective equipment (PPE), and diagnostic kits during crises such as the COVID-19 pandemic. However, the reliability of these demand models remains uncertain. Methods Demand models typically consist of two main components: hospital use epidemiological models that predict hospitalizations or daily admissions, and a demand calculator that translates the outputs of the epidemiological model into predictions for resource usage. We conducted separate analyses to evaluate each of these components. In the first analysis, we validated various hospital use epidemiological models using a recent validation framework designed for epidemiological models. This allowed us to quantify the accuracy of the models in predicting critical aspects such as the date and magnitude of local COVID-19 peaks, among other factors. In the second analysis, we evaluated a range of demand calculators for ventilators, medical gowns, and COVID-19 test kits. To achieve this, we decoupled these demand calculators from the underlying epidemiological models and provided ground truth data for their inputs. This approach enabled a direct comparison of the demand calculators, comparing them against each other and actual usage data when available. The code is available athttps://doi.org/10.5281/zenodo.13712387. Results Performance varied greatly across the epidemiological models, with greater variability in COVID-19 hospital use predictions than for COVID-19 deaths as analyzed previously. Some models did not have any peaks. Among those that did, the models under-estimated date of peak approximately as often as they over-estimated, but were more likely to under-estimate magnitude of peak, with typical relative errors around 50%. Regarding demand calculator predictions, there was significant variability, including five-fold differences in predictions for gown models. Validation against actual or surrogate usage data illustrated the potential value of demand models while demonstrating their limitations. Conclusions The emerging field of demand modeling holds promise in averting medical resource shortages during future public health emergencies. However, achieving this potential necessitates focused efforts on standardization, transparency, and rigorous model validation before placing reliance on demand models in critical public health decision-making.

Medical Informatics↗
DOE OSTI2021

Continued Investigations of Respirable Release Fractions for Stress Corrosion Crack-Like Geometries

The formation of a stress corrosion crack (SCC) in the canister wall of a dry cask storage system (DCSS) has been identified as a potential issue for the long-term storage of spent nuclear fuel. The presence of an SCC in a storage system could represent a through-wall flow path from the canister interior to the environment. Modern, vertical DCSSs are of particular interest due to the commercial practice of using relatively high backfill pressures (up to approximately 800 kPa) in the canister to enhance internal natural convection. This pressure differential offers a comparatively high driving potential for blowdown of any particulates that might be present in the canister. In this study, the rates of gas flow and aerosol transmission of a spent fuel surrogate through an engineered microchannel with dimensions representative of an SCC were evaluated experimentally using coupled mass flow and aerosol analyzers. The microchannel was formed by mating two gage blocks with a linearly tapering slot orifice nominally 13 μm (0.005 in.) tall on the upstream side and 25 μm (0.0010 in.) tall on the downstream side. The orifice is 12.7 mm (0.500 in.) wide by 8.89 mm (0.350 in.) long (flow length). Surrogate aerosols of cerium oxide, CeO 2 , were seeded and mixed with either helium or air inside a pressurized tank. The aerosol characteristics were measured immediately upstream and downstream of the simulated SCC at elevated and ambient pressures, respectively. These data sets are intended to demonstrate a new capability to characterize SCCs under well-controlled boundary conditions. Modeling efforts were also initiated that evaluate the depletion of aerosols in a commercial dry storage canister. These preliminary modeling and ongoing testing efforts are focused on understanding the evolution in both size and quantity of a hypothetical release of aerosolized spent fuel particles from failed fuel to the canister interior and ultimately through an SCC.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2023

Black and gray box learning of amplitude equations: Application to phase field systems

Here, we present a data-driven approach to learning surrogate models for amplitude equations and illustrate its application to interfacial dynamics of phase field systems. In particular, we demonstrate learning effective partial differential equations describing the evolution of phase field interfaces from full phase field data. We illustrate this on a model phase field system, where analytical approximate equations for the dynamics of the phase field interface (a higher-order eikonal equation and its approximation, the Kardar-Parisi-Zhang equation) are known. For this system, we discuss data-driven approaches for the identification of equations that accurately describe the front interface dynamics. When the analytical approximate models mentioned above become inaccurate, as we move beyond the region of validity of the underlying assumptions, the data-driven equations outperform them. In these regimes, going beyond black box identification, we explore different approaches to learning data-driven corrections to the analytically approximate models, leading to effective gray box partial differential equations.

42 ENGINEERING↗
DOE OSTI2022

Manifold Learning-Based Polynomial Chaos Expansions for High-Dimensional Surrogate Models

In this work we introduce a manifold learning-based method for uncertainty quantification (UQ) in systems describing complex spatiotemporal processes. Our first objective is to identify the embedding of a set of high-dimensional data representing quantities of interest of the computational or analytical model. For this purpose, we employ Grassmannian diffusion maps, a two-step nonlinear dimension reduction technique which allows us to reduce the dimensionality of the data and identify meaningful geometric descriptions in a parsimonious and inexpensive manner. Polynomial chaos expansion is then used to construct a mapping between the stochastic input parameters and the diffusion coordinates of the reduced space. An adaptive clustering technique is proposed to identify an optimal number of clusters of points in the latent space. The similarity of points allows us to construct a number of geometric harmonic emulators which are finally utilized as a set of inexpensive pretrained models to perform an inverse map of realizations of latent features to the ambient space and thus perform accurate out-of-sample predictions. Thus, the proposed method acts as an encoder-decoder system which is able to automatically handle very high-dimensional data while simultaneously operating successfully in the small-data regime. The method is demonstrated on two benchmark problems and on a system of advection-diffusion-reaction equations which model a first-order chemical reaction between two species. In all test cases, the proposed method is able to achieve highly accurate approximations which ultimately lead to the significant acceleration of UQ tasks.

42 ENGINEERING↗
DOE OSTI2025

An efficient surrogate model of secondary electron formation and evolution

This work extends the adjoint-deep learning framework for runaway electron (RE) evolution, developed by McDevitt et al. [Phys. Plasmas 32, 042503 (2025)], to account for large-angle collisions. By incorporating large-angle collisions, the framework allows the avalanche of REs to be captured, an essential component of RE dynamics. This extension is accomplished by using a Rosenbluth–Putvinski approximation to estimate the distribution of secondary electrons generated by large-angle collisions. By evolving both the primary and multiple generations of secondary electrons, the present formulation can capture both the detailed temporal evolution of a RE population beginning from an arbitrary initial momentum space distribution, along with providing approximations to the saturated growth and decay rates of the RE population. Predictions of the adjoint-deep learning framework are verified against a traditional RE solver, with good agreement present across a broad range of parameters.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Spatiotemporal forecasting of the edge localized modes in tokamak plasmas using neural networks

Artificial intelligence techniques have been increasingly adopted by the plasma and fusion science to address problems like plasma reconstruction, surrogate modeling, and tokamak/stellarator optimization. A key focus in sustained fusion research is the prediction and mitigation of edge-localized-modes (ELMs), instabilities that occur in short, periodic bursts and can cause erosion to the tokamak vessel wall. Recent research has demonstrated the power of neural networks in approximating continuous functions. In this work, we build spatiotemporal forecasting models that can predict the onset of ELMs and their evolution at early stages. We leverage recent advances in generative modeling, sequence-to-sequence modeling, and Fourier neural operators to propose architectures and training strategies that can learn to forecast short to long term dynamics of the noisy signals due to ELMs. We benchmark the developed model against a state-of-the-art foundation model using the beam emission spectroscopy (BES) data that captures the plasma fluctuations due to ELMs over a 8 x 8 spatial grid. Our models demonstrate high accuracy, outperforming the baselines, in predicting the evolution of BES signals during ELM events. Furthermore, the developed models exhibit high accuracy in predicting the rapid rise and relaxation of the signals due to ELMs within 30–80 µs.

edge localized modes↗
DOE OSTI2025

Water reactive polyurethane grouting for deep vadose zone contaminant immobilization

Contaminants in the deep vadose zone (DVZ) pose a long-term threat to groundwater, human health, and the environment. Polyurethane grouting is a type of chemical grouting that is potentially advantageous in immobilizing contaminants in DVZs. Polyurethane resin has low viscosity that makes it feasible to penetrate and fill the pore space between fine particles. Here, in this study a series of laboratory pressure grouting and leaching tests were conducted to simulate and assess the effectiveness of polyurethane grouting for immobilizing contaminants in DVZs. Soil with fine particles was prepared with 127 I (as iodide) that served as a non-radioactive surrogate for radioactive 129 I. After grouting and curing, leaching tests were used to measure and compare contaminant diffusivity and leachability index values. Additionally, changes in porosity and saturated hydraulic conductivity of grouted soil were measured. X-Ray Computer Tomography (XCT) results showed that the cured polyurethane was distributed nearly homogeneously and approximately half of the voids were filled with cured polyurethane. Grouting reduced the saturated hydraulic conductivity of soil by 37 %. The effective diffusivity decreased by more than 80 % as compared with the ungrouted soil. The leachability index of the grouted soil was 6.5; meeting the criteria established by the U.S. Nuclear Regulatory Commission (NRC) standard. The results obtained in this study provide a valuable assessment of polyurethane grouting for iodide immobilization in the DVZ and indicate this approach may be a viable method for contaminant remediation in DVZ soils.

Contaminant Diffusion↗
DOE OSTI2024

Flame kinetics at scramjet-engine-relevant conditions: Role of prompt dissociation of weakly-bound radicals

Combustion in high-speed ram-based propulsion engines occurs under distinct thermodynamic conditions of high reactant temperatures (greater than 1000 K) and relatively low pressures (<5 atm). There is a lack of fundamental flame measurements at such conditions that result in adiabatic flame temperatures (T ad ) exceeding 2500 K. In this work, we have measured laminar flame speeds of oxygen-enriched CH 4 /oxidizer mixtures at sub-atmospheric conditions to probe kinetics at high T ad using the isobaric spherically expanding flame approach. Simulations with recent kinetic models revealed increasing differences between data and model predictions with increasing T ad , reaching up to 25 %. Kinetic analyses reveal that at the thermodynamic conditions in these O 2 -enriched flames, i.e., lower pressures and higher T ad , the effects of HCO prompt dissociation are accentuated. In addition to HCO, the prompt dissociations of CH 2 OH and C 2 H 5 are also considered. Here, the prompt dissociations of all three radicals were evaluated and their effects considered in flame speed simulations. Reaction path analysis for the present flames revealed that approximately half of the reaction flux for HCO formation undergoes prompt dissociation to H + CO. Furthermore, these analyses also revealed that the pathways and sensitive reactions are similar between oxygen-enriched fuel/oxidizer mixtures and preheated fuel/air mixtures, if both have similar T ad . Thus, flames of oxygen-enriched mixtures could be a surrogate to probe the flame chemistry of highly preheated mixtures at relatively low pressures that are often encountered in ram-based propulsion engine combustors.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2026

Estimation of post-burst vibration-induced fuel dispersal thresholds following LOCA-simulant testing

The nuclear industry intends to increase burnup past the regulatory limit of 62 GWd/MTU, and to be successful, the dispersal behavior of fragmented high-burnup fuel during loss-of-coolant accidents must be better understood. Part of this behavior is fuel fragmentation, relocation, and dispersal. This work reports on testing concerning the dispersal of fuel from rods following burst and pressure blowdown through the burst. The Fuel Relocation Induced by Accident Recovery (FRIAR) system was used to examine surrogate material and pre-burst cladding specimens to evaluate the effects of vibrational loading, mixture composition, and burst size on post-burst dispersal. Vibrational loading resulted in maximum dispersal around 15 Hz when the loading was performed with peak-to-peak oscillation amplitudes at 0.5 and 1 mm. Here, under this aggressive vibrational loading, dispersal was found to become significant as burst widths increased to approximately 5 mm and larger. However, below 4 mm widths, little material dispersed out of the opening, even at 1 mm amplitudes and for more dispersive mixtures (smaller average fragments). At and above 7 mm widths, dispersal was rapid for all mixtures. Finally, dispersal in the threshold region (4–5 mm wide bursts) strongly depended on the mixture. For mixtures with large fragments, dispersal was minimized to a few fragments, as the large fragments proved to be effective at minimizing dispersal. Conversely, mixtures with smaller fragments showed increased dispersal in this region.

Burst geometry↗
DOE OSTI2024

Investigating sol–gel matrix loading capacity toward producing surrogate nuclear explosive debris with realistic composition

Post detonation nuclear forensic materials which resemble the size, color, elemental composition, and radionuclide content of real nuclear debris would be valuable for developing and validating new nuclear forensic techniques. As nuclear fallout types vary significantly, the ability to tailor each of these parameters accurately is desired to produce materials capable of testing analytical methods under a wide array of forensic scenarios. Sol–gel synthesis techniques can provide tunability of size, shape and composition for producing a wide variety of solid nuclear forensics benchmarking materials. Further, the sol–gel process consists of forming a metal oxide material, often silica, through polymerization of a metal-alkoxy precursor. In this work, we characterize the ability to load sol–gel particles with secondary elemental components such as iron, aluminum, and calcium toward producing benchmarking materials approximating the elemental composition of historic nuclear debris from the Nevada National Security Site. We also demonstrate quantitative radionuclide encapsulation toward producing benchmarking materials with controllable radionuclide content. Finally, we employ these techniques to produce nuclear debris benchmarking materials with controllable elemental matrix composition and radionuclide content and compare these samples with the composition of a historic fallout sample previously reported from the Nevada National Security Site.

36 MATERIALS SCIENCE↗
DOE OSTI2024

A chemical kinetic analysis of knock propensity of methanol-to-gasoline fuel

Production of low carbon gasoline-like fuels such as methanol-to-gasoline (MTG) is a promising approach to achieve rapid greenhouse gas emission reduction of the transportation sector. Despite the fact that gasoline that meets the ASTM D4814 standard for automotive spark-ignition engine fuel can be readily produced from these processes, it is unclear how the composition of MTG may affect engine performance and emissions. Here, in this paper, a surrogate for an MTG is used to numerically study the effects of gasoline composition on knock propensity and on the sensitivity of knock to thermal and fuel stratification, to oxygen dilution and to nitric oxide from exhaust gas recirculation of residual gases. Simulations were performed in ANSYS CHEMKIN-PRO using a comprehensive chemical kinetic mechanism for gasoline surrogates, and results of the MTG surrogate were compared against those of a petroleum-based regular E10 gasoline, termed PACE-20. A premium-grade MTG fuel was also formulated by adding ethanol to the MTG surrogate, and results were compared against those of four premium-grade, gasoline-like fuels representative of future alternative gasoline formulations. Surrogates and mechanism were evaluated by comparison against experimental engine data, and the model showed high accuracy at stoichiometric conditions (mean absolute error of ignition timing equal to 1.46 crank angle degrees) but larger deviations at lean conditions (mean absolute error of ignition timing equal to 5.52 crank angle degrees). Despite the fact that the MTG surrogate has a RON 1.1 units higher than that of PACE-20, it may show higher knock propensity at medium temperature conditions due to a less intense NTC behavior. MTG autoignition was more temperature- and equivalence ratio-sensitive than that of PACE20, suggesting that MTG can benefit more from naturally-occurring thermal stratification or from induced fuel stratification of the end gas to mitigate knock intensity. The sensitivity of autoignition reactivity to oxygen dilution and to NO concentration was higher for MTG than for regular gasoline at medium loads, but the opposite trend was observed at high loads due to the effect of pressure on the low-temperature chemistry of regular gasoline. Approximately 14 % vol ethanol content was required to upgrade the octane rating of MTG from regular grade to premium grade. Adding 13.6 % vol ethanol made the fuel autoignition less sensitive to both oxygen dilution and NO content (ignition time varies approx. 17 % and 50 % less with oxygen dilution and NO addition, respectively, when adding ethanol at high engine loads).

02 PETROLEUM↗
DOE OSTI2025

Bayesian Calibration of Stochastic Agent Based Model via Random Forest

Agent-based models (ABM) provide an excellent framework for modeling outbreaks and interventions in epidemiology by explicitly accounting for diverse individual interactions and environments. However, these models are usually stochastic and highly parametrized, requiring precise calibration for predictive performance. When considering realistic numbers of agents and properly accounting for stochasticity, this high-dimensional calibration can be computationally prohibitive. This paper presents a random forest-based surrogate modeling technique to accelerate the evaluation of ABMs and demonstrates its use to calibrate an epidemiological ABM named CityCOVID via Markov chain Monte Carlo (MCMC). The technique is first outlined in the context of CityCOVID's quantities of interest, namely hospitalizations and deaths, by exploring dimensionality reduction via temporal decomposition with principal component analysis (PCA) and via sensitivity analysis. The calibration problem is then presented, and samples are generated to best match COVID-19 hospitalization and death numbers in Chicago from March to June in 2020. Further, these results are compared with previous approximate Bayesian calibration (IMABC) results, and their predictive performance is analyzed, showing improved performance with a reduction in computation.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2026

Fermilab Booster loss modelling and rebalancing using Bayesian methods

To meet PIP-II upgrade requirements, Fermilab Booster losses need to be reduced by 50% compared to present levels. So far, simulations are not good enough to predict loss patterns. Thus, an extensive Booster tune up will be necessary to achieve required performance. In this paper we present an effort to build a data-driven loss model using Bayesian techniques, and subsequently to rebalance losses for higher trip margins. We first created several sets of spatially and temporally isolated orbit and optics knobs, and trained Gaussian process models for each beam loss monitor as well as beam current. Novel techniques of uncertainty constraints and approximate GP fitting were introduced to handle safety and timing requirements. We then performed single and multi-objective tuning using scalarized objectives comprised of critical beam loss locations. We achieved significant rebalancing of losses, increasing margins by 25%, as well as an overall improvement in transmission efficiency of 0.4%. Automated data collection is being developed so that more accurate surrogate models can be trained over time.

Kuklev, N. [Fermilab]↗
DOE OSTI2023

Neural network ensembles and uncertainty estimation for predictions of inelastic mechanical deformation using a finite element method-neural network approach

The finite element method (FEM) is widely used to simulate a variety of physics phenomena. Approaches that integrate FEM with neural networks (NNs) are typically leveraged as an alternative to conducting expensive FEM simulations in order to reduce the computational cost without significantly sacrificing accuracy. However, these methods can produce biased predictions that deviate from those obtained with FEM, since these hybrid FEM-NN approaches rely on approximations trained using physically relevant quantities. In this work, an uncertainty estimation framework is introduced that leverages ensembles of Bayesian neural networks to produce diverse sets of predictions using a hybrid FEM-NN approach that approximates internal forces on a deforming solid body. The uncertainty estimator developed herein reliably infers upper bounds of bias/variance in the predictions for a wide range of interpolation and extrapolation cases using a three-element FEM-NN model of a bar undergoing plastic deformation. This proposed framework offers a powerful tool for assessing the reliability of physics-based surrogate models by establishing uncertainty estimates for predictions spanning a wide range of possible load cases.

42 ENGINEERING↗
DOE OSTI2026

Component-Level Inverse Design of Transmon Qubits Using Neural Networks

Designing a superconducting qubit to realize specific Hamiltonian parameters typically requires iterating through a time and compute-intensive forward loop in which the designer chooses a layout geometry, simulates it, extracts circuit parameters such as capacitances, and refines the geometry. We study the inverse version of this task using a neural-network workflow that maps target Hamiltonian parameters directly to component-level layout parameters, which we subsequently demonstrate on a planar transmon layout. During training, we pair the inverse model with a frozen forward surrogate model and evaluate the loss in Hamiltonian space rather than in layout-parameter space. In validation against a conventional EM solver, 97% of generated designs produce usable geometries, and the inverse-plus-surrogate pipeline reaches mean percent errors of 0.73% for qubit frequency and 1.58% for anharmonicity, comparable to or below the fabrication and simulation-to-measurement uncertainty expected for academic-process transmon devices of this type. A single pipeline query takes ~60 ms on CPU, versus ~2 min for a conventional EM capacitance extraction on the same hardware, a speedup of approximately 2,000x. Batching minimizes the AI model inference overhead, reducing the runtime to 3.1 microseconds per sample on CPU and 2.6 microseconds per sample on GPU at a batch size of 2048, resulting in speedups of 3.9 x 10^7 and 4.6 x 10^7, respectively, relative to a single conventional CPU EM extraction. Our results indicate that component-level inverse design usefully extends and complements conventional EM simulation, including for small datasets on the order of 1,000 samples.

Seidel, Olivia [Fermilab; Texas U., Arlington]↗