Search NASA⌕ Search

SEARCH · Search NASA

Results for “Transient Dynamic 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.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

1,052 records · Page 39
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

REDOTHERM (Redox Countercurrent Thermodynamic Limits Model) [SWR-24-88]

REDOTHERM is an open-source, MATLAB-based thermodynamic modeling framework developed to evaluate the performance of redox-active materials for water (H2O) and carbon dioxide (CO2) splitting. It includes models of all major unit operations and supports comparative analysis of different redox-active material candidates. The model is tailored for systems of moving oxide under a parallel/cocurrent flow (PF) and countercurrent flow (CF) configurations. Unvalidated mixed flow reactor (MFR, also known as CSTR) model is also included as an optional addition.

Lidor, Alon [National Renewable Energy Laboratory ↗
NASA NTRS2026

SERFE Ground Unit EVA Series After Three Year Spacesuit Stowage Period

NASA’s spacesuit government reference design for returning to the Moon is called the Exploration Extravehicular Mobility Unit (xEMU). The xEMU subassembly that provides life support, such as oxygen and thermal control, is the Portable Life Support System (PLSS). Inside the PLSS is a new technology that NASA wanted to test to provide cooling to the crew during EVAs (ExtraVehicular Activity). This technology is called the Spacesuit Water Membrane Evaporator (SWME). In order to test SWME in an improved thermal control loop (TCL) both on Earth and in Space, the Spacesuit Evaporation Rejection Flight Experiment (SERFE) was created. The Ground unit, or testbed at Johnson Space Center (JSC), tested the cooling technology in Earth’s gravity, while the Flight unit, or payload on the International Space Station (ISS), tested the cooling technology in micro-gravity. Since fluids flow differently in micro-gravity, testing in both environments would provide important data for improving cooling performance during EVAs. Both units completed 25 simulated EVAs with the same settings so SWME performance on the ground could be compared to the ISS. The Flight unit was completed first and performed EVAs on the ISS between 2020 and 2022. The Ground unit performed EVAs between 2021 and 2022. When the Flight unit came back from the ISS, it was taken apart for analysis. This analysis looked at how well SWME was able to maintain its heat rejection capability after various dwell times, such as a 90 day Airlock Coolant Loop Recovery (ALCLR) cycle, when the Extravehicular Mobility Unit (EMU) currently on the ISS would normally need maintenance. After a three year simulated spacesuit dwell, the Ground unit performed another EVA series in 2025 to test SWME’s shelf life. The results from this test series will inform mission planning as NASA plans to go back to the Moon and beyond.

SWME↗
NASA NTRS1992

Thermal expansion behavior of LDEF metal matrix composites

The effects of the space environment on the thermal expansion stability of metal matrix composites (graphite/Al and graphite/Mg) will be presented. A sample from each category of metal matrix composites mounted on the leading and trailing edge was chosen for analysis of the temperature-time-thermal strain histories. Typical thermal expansion curves over the same range of temperature were selected at the beginning, mid, and end of the recording duration. The thermal expansion of selected post-flight LDEF samples were measured over the same range of temperature in the laboratory using a Michelson laser interferometer. The thermal strains were monitored concurrently with a laser interferometer and a mounted strain gage.

T. D. Le↗
NASA NTRS2015

Empirical Model Development for Predicting Shock Response on Composite Materials Subjected to Pyroshock Loading: Appendices

The NASA Engineering and Safety Center (NESC) received a request to develop an analysis model based on both frequency response and wave propagation analyses for predicting shock response spectrum (SRS) on composite materials subjected to pyroshock loading. The model would account for near-field environment (approx. 9 inches from the source) dominated by direct wave propagation, mid-field environment (approx. 2 feet from the source) characterized by wave propagation and structural resonances, and far-field environment dominated by lower frequency bending waves in the structure. This document contains appendices to the Volume I report.

Gentz, Steven J↗
DOE OSTI2026

A Data-Driven Method for Modeling Creep-Fatigue Stress- Strain Behavior Using Neural ODEs

In this paper, we introduce a data-driven machine learning approach for modeling one-dimensional stress–strain behavior under cyclic loading, utilizing experimental data from the nickel-based Alloy 617. The study employs uniaxial creep–fatigue test data acquired under various loading histories and compares two distinct neural network-based ODE models. The first model, known as the black-box model, comprehensively describes the strain–stress relationship using a Neural ODE equation. To interpret this black-box model, we apply the Sparse Identification of Nonlinear Dynamical Systems (SINDy) technique, transforming the black-box model into an equation-based model using symbolic regression. The second model, the Neural flow rule model, incorporates Hooke’s Law for the linear elastic component, with the nonlinear part characterized by a Neural ODE. Both models are trained with experimental data to accurately reflect the observed stress–strain behavior. We conduct a detailed comparison with the standard Chaboche model, which includes three back stresses. Our results demonstrate that the neural network-based ODE models precisely capture the experimental creep–fatigue mechanical behavior, exceeding the standard Chaboche model’s accuracy. Furthermore, an interpretable model derived from the black-box neural ODE model through symbolic regression achieves accuracy comparable to the Chaboche model, enhancing its interpretability. The results highlight the potential of neural network-based ODE models to depict complex creep–fatigue behavior, eliminating the necessity for experts to define a specific, material-focused model form.

creep-fatigue↗
NASA NTRSDate not supplied

First-Principles Investigation of Phase Transitions in RuNb, RuTa, and Ru-Nb-Ta Ultrahigh Temperature Shape Memory Alloys

Ultrahigh temperature shape memory alloys (UHT-SMAs) have transition temperatures above 600 C, and they have found applications for sensing and actuating devices in aerospace industry. Among very few such UHT-SMAs currently known are Ru-based alloys such as RuNb and RuTa, whose martensite structures and phase transitions are totally different from those of NiTi-based SMAs and were poorly understood. In this work, we carried out a systematical study of RuNb using first-principles total energy calculations and molecular dynamics (MD) simulations. The transition paths and mechanisms in cubic → tetragonal →monoclinic transitions are revealed. and the transition sequence and martensitic transition temperatures are determined (MTTs) by evaluating the Gibbs free energies using thermodynamic integration. The calculated MTTs are in very good agreement with the experimental data. We found that the monoclinic phase at the second transition has the 𝑃21/m symmetry instead of experimentally identified 𝑃2/m. Our calculations demonstrate that RuTa has very similar phase transitions to those of RuTa. Furthermore, we studied the Ru0.5Nb0.25Ta0.25 ternary. Our results can explain the measured significant drop of MTT (~ 200 C) in the second transition for Ru0.5Nb0.25Ta0.25 compared with those of RuNb and RuTa, while in the first transition its MTT is between MTTs of RuNb and RuTa. The insights gained by this study and the verified ab initio methods for accurate MTT calculations can be applied to fast screen and quantitatively design novel UHT-SMAs having similar properties with desirable MTTs and much reduced cost.

Zhigang Wu↗
DOE OSTI2025

Keldysh tuning of photoluminescence in a lead halide perovskite crystal

In 1964, Keldysh laid the groundwork for strong-field physics in atomic, molecular, and solid-state systems by delineating a ubiquitous transition from multiphoton absorption to quantum electron tunneling under intense AC driving forces. While both processes in semiconductors can generate carriers and result in photon emission through electron-hole recombination, the low quantum yields in most materials have hindered direct observation of the Keldysh crossover. Leveraging the large quantum yields of photoluminescence in lead halide perovskites, we show that we can not only induce bright light emission from extreme sub-bandgap light excitation but also distinguish between photon-induced and electric-field-induced processes. Our results are rationalized by the Landau-Dykhne formalism, providing insights into the non-equilibrium dynamics of strong-field light-matter interactions. These findings open new avenues for light upconversion and sub-bandgap photon detection, highlighting the potential of lead halide perovskites in advanced optoelectronic applications.

FOS: Physical sciences↗
DOE OSTI2025

From PINNs to PIKANs: recent advances in physics-informed machine learning

Physics-Informed Neural Networks (PINNs) have emerged as a key tool in Scientific Machine Learning since their introduction in 2017, enabling the efficient solution of ordinary and partial differential equations using sparse measurements. Over the past few years, significant advancements have been made in the training and optimization of PINNs, covering aspects such as network architectures, adaptive refinement, domain decomposition, and the use of adaptive weights and activation functions. A notable recent development is the Physics-Informed Kolmogorov-Arnold Networks (PIKANS), which leverage a representation model originally proposed by Kolmogorov in 1957, offering a promising alternative to traditional PINNs. In this review, we provide a comprehensive overview of the latest advancements in PINNs, focusing on improvements in network design, feature expansion, optimization techniques, uncertainty quantification, and theoretical insights. We also survey key applications across a range of fields, including biomedicine, fluid and solid mechanics, geophysics, dynamical systems, heat transfer, chemical engineering, and beyond. Lastly, we review computational frameworks and software tools developed by both academia and industry to support PINN research and applications.

Kolmogorov-Arnold networks↗
DOE OSTI2026

Ionic Precursors Transformed Into Vinyl Acetate Synthesis Catalyst via Reaction-Driven Restructuring

Conventional preparation of supported bimetallic catalysts relies on solution-mediated metal salt immobilization and pre-formation of alloy nanoparticles before reaction. Here, we report a fundamentally different synthesis strategy of using a physical mixture of salt precursors to generate an active catalyst during reaction. The catalytic structure is generated in situ from Pd3(OAc)6, Au(OH)3, and KOAc through H2 treatment and reaction-driven restructuring under vinyl acetate monomer (VAM) synthesis conditions. Ascertained from in situ X-ray diffraction and operando infrared spectroscopy analyses, reduction treatment produces segregated Pd and Au domains, and subsequent exposure to a VAM reaction mixture triggers dynamic extraction of Pd from the metal surface. This latter process, mediated by acetate-assisted redox cycles, facilitates Pd migration toward Au domains to form a near-surface localized Pd50Au50 alloy phase. Monometallic Pd domains serve as a reservoir of Pd to the alloy phase, leading to and sustaining a more Pd-enriched active surface and a higher population of accessible Pd sites, compared to a conventionally prepared K-PdAu/SiO2 catalyst. Consequently, this leads to a twofold increase in the VAM formation rate, demonstrating highly active bimetallic catalysts can be generated through the gas-phase treatment of physically mixed ionic precursors.

Cha, Byeong Jun [Rice University]↗
NASA NTRSDate not supplied

Performance Assessment of LunaNet’s Augmented Forward Signal

LunaNet provides a common set of interoperable specifications for communication and position, navigation and time (PNT) services and interfaces soon to be implemented in lunar vicinity. The LunaNet Interoperability Specification (LNIS) provides the design for the GNSS-like Augmented Forward Signal (AFS), which enables orbiting and surface users in lunar space, such as Artemis, to estimate their position, velocity, and time. The specification of AFS defines two orthogonal signal components on a single carrier: the in-phase component (AFS-I), a lower-chip-rate data channel tailored for applications where low SWaP (Size, Weight, and Power) is critical (e.g., IoT devices or search and rescue), and the quadrature component (AFS-Q), a high-chip-rate data-less pilot signal for high-precision, robust lunar navigation and positioning applications. An initial description of AFS was provided in LNIS 2023, with initial analysis results shown in Dafesh 2024 and Dafesh 2025, and the current signal in space description provided in LNIS 2025. As part of NASA's Lunar Communication Relay and Navigation Systems (LCRNS) project, this work expands upon the initial analysis results and proposes a new expanded set of AFS-Q spreading codes that exceed the cross-correlation and autocorrelation sidelobe performance of L1C and other GNSS signals, while providing additional expansion capabilities for future service satellites. A set of 420 codes was selected from a Weil-based code derived from the prime number 10247, which is larger than the 10243 prime number used to derive BeiDou’s B1C Weil sequences. Both the initial set of 210 codes and the expanded set of 420 codes are shown to provide the best cross-correlation of any 10230-chip satellite navigation codes. The performance is demonstrated for hierarchical sets of spreading codes optimized and organized in sets of 30 codes. The work also compares LunaNet’s AFS to terrestrial GNSS signals in terms of acquisition, tracking, and data demodulation performance. Performance is evaluated for receivers that only track the 1.023 MCPS data channel spreading code for low SWaP IoT use cases, as well as for receivers that track both the 1.023 MCPS data channel and the 5.115 MCPS pilot channel spreading code for high-performance use cases. Performance is assessed in the presence of interference and thermal noise. The analysis is performed in terms of expected operating conditions on the lunar surface. Several unique flexibility aspects of the augmented forward signal are described, including the use of the Q channel’s secondary and tertiary codes to enable variable coherent integrations during acquisition. This is compared to GNSS signals such as L5/E5 and MBOC in terms of achievable processing gain for interference mitigation versus acquisition complexity. The work details acquisition and tracking techniques used to optimally acquire and track the primary, secondary, and tertiary codes on the Q channel, as well as acquisition of the I channel spreading code. Acquisition of the 8 ms, Q channel spreading code is also compared to joint acquisition of the I and Q channel primary codes in noise and interference environments

LCRNS↗
DOE OSTI2026

Enhancing Coherence Limits in Superconducting Quantum Systems for Computing and Sensing

This talk will highlight recent efforts at the SQMS Center to develop qudit-based quantum computing architectures using superconducting three-dimensional (3D) cavities, as well as the use of these ultra-coherent cavities for quantum sensing. I will present systematic studies of materials and devices aimed at identifying and mitigating the dominant sources of decoherence—including two-level systems (TLS), quasiparticles, and other noise mechanisms—in both transmons and 3D cavities. These investigations include microwave loss characterization of niobium, tantalum, aluminum, their native oxides, and substrate materials such as silicon and sapphire. By combining measurements on qubits and cavities, we disentangle subsystem-specific loss mechanisms and establish a hierarchy of mitigation strategies, leading to transmon coherence times exceeding one millisecond. I will also discuss studies of quasiparticle dynamics, including quasiparticle bursts observed in qubits operated both above ground and at the Gran Sasso underground laboratory, and the observation that applied magnetic fields can suppress temporal T₁ fluctuations. Building on these advances, we demonstrate a record-coherence two-cell cavity-qudit system with coherence times exceeding 20 milliseconds. Leveraging tunable sideband interactions together with error-resilient protocols, including measurement-based error correction and post-selection, we achieve high-fidelity quantum state control, including the preparation of Fock states up to N=20 with fidelities above 95% and the generation of high-fidelity two-mode entangled states. Finally, I will discuss how these ultra-coherent quantum systems are enabling emerging quantum sensing applications, including searches for dark matter and gravitational waves.

Roy, Tanay [Fermilab] (ORCID:000000019442862X)↗
NASA NTRS2015

Empirical Model Development for Predicting Shock Response on Composite Materials Subjected to Pyroshock Loading: Appendices - Volume 2, Part 1

The NASA Engineering and Safety Center (NESC) received a request to develop an analysis model based on both frequency response and wave propagation analyses for predicting shock response spectrum (SRS) on composite materials subjected to pyroshock loading. The model would account for near-field environment (approximately 9 inches from the source) dominated by direct wave propagation, mid-field environment (approximately 2 feet from the source) characterized by wave propagation and structural resonances, and far-field environment dominated by lower frequency bending waves in the structure. This document contains appendices to the Volume I report.

Steven J Gentz↗
NASA NTRS1977

Riddles About the Origin and Thermal History of the Moon

Analyses of lunar rocks have confirmed that the interior of the Moon has been subjected to magmatic differentiation, as has been determined by calculation of the thermal history of the Moon. However, it appears that differentiation occurred so early that it could not have been related to heating by long-lived radioactive elements, but rather indicates a high initial temperature of the Moon. The reasons for the high initial temperature remain unknown. The presence of a central, partially melted area revealed by seismic observations forces us to reject convective models of the Moon, while the dimensions of this area, together with the measured heat fluxes from the interior, serve as "boundary conditions" for any new calculations of the thermal history for a two-layered differentiated model of the Moon operative over the past 3.5 billion years. If we start with the early differentiation of the Moon and the high content of long-lived radioactive elements in it, explanation of its current thermal properties represents no difficulty. But these prerequisites themselves remain unexplained. Due to the continuing controversy concerning the true nature of the process of magmatic differentiation of the Moon, estimation of the composition of the entire Moon on the basis of analyses of surface rock remains unreliable, and use of these estimates to clarify the origin of the Moon is premature. Existing hypotheses concerning the origin of the Moon encounter dynamic difficulties, as well as difficulties with explanation of the high initial temperature, to say nothing of the difficulty involved in explaining the proposed differences in the composition of the Earth and Moon.

B Yu Levin↗
NASA NTRSDate not supplied

Performance Assessment of LunaNet’s Augmented Forward Signal

LunaNet provides a common set of interoperable specifications for communication and position, navigation and time (PNT) services and interfaces soon to be implemented in lunar vicinity. The LunaNet Interoperability Specification (LNIS) provides the design for the GNSS-like Augmented Forward Signal (AFS), which enables orbiting and surface users in lunar space, such as Artemis, to estimate their position, velocity, and time. The specification of AFS defines two orthogonal signal components on a single carrier: the in-phase component (AFS-I), a lower-chip-rate data channel tailored for applications where low SWaP (Size, Weight, and Power) is critical (e.g., IoT devices or search and rescue), and the quadrature component (AFS-Q), a high-chip-rate data-less pilot signal for high-precision, robust lunar navigation and positioning applications. An initial description of AFS was provided in [1], with initial analysis results shown in [2] and [3] and the current signal in space description provided in [4]. As part of NASA's Lunar Communication Relay and Navigation Systems (LCRNS) project, this work expands upon the initial analysis results and proposes a new expanded set of AFS-Q spreading codes that exceed the cross-correlation and autocorrelation sidelobe performance of L1C and other GNSS signals, while providing additional expansion capabilities for future provider satellites. A set of 420 codes was selected from a Weil-based code derived from the prime number 10247, which is larger than the 10243 prime number used to derive Beidou’s B1C Weil sequences. Both the initial set of 210 codes and the expanded set of 420 codes are shown to provide the best cross-correlation of any 10230-chip satellite navigation codes. The performance is demonstrated for hierarchical sets of spreading codes optimized and organized in sets of 30 codes. The new codes were developed using an optimization approach and correlation methodology described in [5]. The work also compares LunaNet’s AFS to terrestrial GNSS signals in terms of acquisition, tracking, and data demodulation performance. Performance is evaluated for receivers that only track the 1.023 MCPS data channel spreading code for low SWaP IoT use cases, as well as for receivers that track both the 1.023 MCPS data channel and the 5.115 MCPS pilot channel spreading code for high-performance use cases. Performance is assessed in the presence of interference and thermal noise. The analysis is performed in terms of expected operating conditions on the lunar surface. Several unique flexibility aspects of the augmented forward signal are described, including the use of the Q channel’s secondary and tertiary codes to enable variable coherent integrations during acquisition. This is compared to GNSS signals such as L5/E5 and MBOC in terms of achievable processing gain for interference mitigation versus acquisition complexity. The work details acquisition and tracking techniques used to optimally acquire and track the primary, secondary, and tertiary codes on the Q channel, as well as acquisition of the I channel spreading code. Acquisition of the 8 ms Q channel spreading code is also compared to joint acquisition of the I and Q channel primary codes in noise and interference environments.

LCRNS↗
NASA NTRS2026

Performance Optimization for Lunar Extravehicular Activity Readiness (POLAR) Study: Methods Paper

To better understand which aspects of physical fitness may be most related to performance during Lunar surface operations and thereby help to inform the current NASA fitness standards, much can be learned from fields encompassing the “tactical athlete.” Other physically demanding professions such as law enforcement, military, or rescue professionals often require candidates to meet occupationally-relevant fitness standards. The determination of such standards is a multistep process, including both objective and subjective measures, to determine tasks essential to occupational performance and identify the minimal fitness profile needed to meet physical demands of the job. Notably, fitness is only one component which may contribute to the demands of astronaut selection, flight assignment, and occupational performance. Utilizing a framework to systematically determine which domains of fitness most contribute to relevant job tasks can aid in the refinement of current NASA-STD-3001 fitness standards. Therefore, NASA’s Exercise Physiology & Countermeasures Laboratory conducted the Performance Optimization for Lunar Extravehicular Activity Readiness (POLAR) study to identify and examine a comprehensive list of fitness tests (including NASA-STD-3001 assessments: 1-Repetition Maximum [1-RM] bench press and deadlift) and determine preliminary relationships between identified fitness parameters and novel Artemis-relevant tasks to help inform future investigations for the continued development of aerobic and muscular fitness standards for surface EVAs. This was accomplished through 1) a review of the literature relating fitness assessments to simulated or real EVA performance to identify fitness tests that are most correlated with simulated EVA task performance; 2) a task analysis following a modified framework for physical employment standards development to down select mission critical tasks; and 3) development and pilot testing of a novel, portable Artemis-relevant EVA task circuit to relate to a battery of fitness assessments. The current report describes the methodology used to complete the task analysis, EVA task circuit development, and the pilot study.

Nicole C Strock↗
DOE OSTI2026

Sintering effects on NpO 2 grain size and morphology: The role of precursor Np phase

Neptunium dioxide (NpO 2 ) is a key phase in nuclear material processing as a target material for the production of plutonium-238 ( 238 Pu) and has been historically synthesized via the calcination of a Np oxalate precursor. Alternative synthesis methods for NpO 2 are now more prevalent, necessitating their study and comparison with the more common oxalate route. The purpose of this work was to investigate the microstructural properties of NpO 2 synthesized from a nitrate-based Np precursor phase via the assessment of NpO 2 particle size and morphology as a function of calcination temperature and time. Scanning electron microscopy (SEM) was used to probe the primary grain size and morphology of NpO 2 after calcination at temperatures ranging from 700 to 1100 °C and hold times ranging from 1 to 10 h. Post-image analysis using ImageJ software enabled the quantification of mean particle diameter. This analysis indicated that particle diameter increases with both increasing calcination temperature and hold time. Primary particles were shown to be clumped in irregular patterns into the overall rough, blocky aggregates, but this macroscopic morphology was not affected by calcination time or temperature. Although trends in primary grain size of NpO 2 were consistent with available literature from other Np precursor phases, the macroscopic morphology of the NpO 2 aggregates was quite different than reported for other precursors. Through comparison with historical literature on Np oxalate, this work emphasizes the importance of Np precursor on the physical properties of NpO 2 .

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2024

Isotope engineering for spin defects in van der Waals materials

Abstract Spin defects in van der Waals materials offer a promising platform for advancing quantum technologies. Here, we propose and demonstrate a powerful technique based on isotope engineering of host materials to significantly enhance the coherence properties of embedded spin defects. Focusing on the recently-discovered negatively charged boron vacancy center ($${{{{{{{{\rm{V}}}}}}}}}_{{{{{{{{\rm{B}}}}}}}}}^{-}$$ V B − ) in hexagonal boron nitride (hBN), we grow isotopically purified h 10 B 15 N crystals. Compared to$${{{{{{{{\rm{V}}}}}}}}}_{{{{{{{{\rm{B}}}}}}}}}^{-}$$ V B − in hBN with the natural distribution of isotopes, we observe substantially narrower and less crowded$${{{{{{{{\rm{V}}}}}}}}}_{{{{{{{{\rm{B}}}}}}}}}^{-}$$ V B − spin transitions as well as extended coherence timeT 2 and relaxation timeT 1 . For quantum sensing,$${{{{{{{{\rm{V}}}}}}}}}_{{{{{{{{\rm{B}}}}}}}}}^{-}$$ V B − centers in our h 10 B 15 N samples exhibit a factor of 4 (2) enhancement in DC (AC) magnetic field sensitivity. For additional quantum resources, the individual addressability of the$${{{{{{{{\rm{V}}}}}}}}}_{{{{{{{{\rm{B}}}}}}}}}^{-}$$ V B − hyperfine levels enables the dynamical polarization and coherent control of the three nearest-neighbor 15 N nuclear spins. Our results demonstrate the power of isotope engineering for enhancing the properties of quantum spin defects in hBN, and can be readily extended to improving spin qubits in a broad family of van der Waals materials.

Science & Technology - Other Topics↗
DOE OSTI2026

PowerModelsGAT-AI: Physics-Informed Graph Attention for Multi-System Power Flow With Continual Learning

Solving the alternating current power flow equations in real time is essential for secure grid operation, yet classical Newton–Raphson solvers can be slow under stressed conditions. Existing graph neural networks for power flow are typically trained on a single system and often degrade on different systems. We present PowerModelsGAT-AI, a physics-informed graph attention network that predicts bus voltages and generator injections. The model uses bus-type-aware masking to handle different bus types and balances multiple loss terms, including a power-mismatch penalty, using learned weights. We evaluate the model on 14 benchmark systems (4 to 6,470 buses) and train a unified model on 13 of these under contingency conditions with up to two branch outages, achieving an average normalized mean absolute error of 0.89% for voltage magnitudes and R 2 >0.99 for voltage angles. We also show continual learning: when adapting a base model to a new 1,354-bus system, standard fine-tuning causes severe forgetting with error increases exceeding 1000% on base systems, while our experience replay and elastic weight consolidation strategy keeps error increases below 2% and in some cases improves base-system performance. Interpretability analysis shows that learned attention weights correlate with physical branch parameters (susceptance: r=0.38 ; thermal limits: r=0.22 ), and feature importance analysis supports that the model captures established power flow relationships.

24 POWER TRANSMISSION AND DISTRIBUTION↗