Search NASA⌕ Search

SEARCH · Search NASA

Results for “ensemble optimization”

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 667 records · Page 37

Finite-Temperature Structures of Supported Subnanometer Catalysts Inferred via Statistical Learning and Genetic Algorithm-Based Optimization

Single-atom catalysts (SACs) minimize noble metal utilization and can alter the activity and selectivity of supported metal nanoparticles. However, the morphology of active centers, including single atoms and subnanometer clusters of a few atoms, remains elusive due to experimental challenges. The computational cost to describe numerous cluster shapes and sizes makes direct first-principles calculations impractical. We present a computational framework to enable structure determination for single-atom and subnanometer cluster catalysts. As a case study, we obtained the low energy structures of Pd n (n = 1-21) clusters supported on CeO 2 (111), which are critical components of automobile three-way catalysts. Trained on density functional theory data, a three-dimensional cluster expansion is established using statistical learning to describe the Hamiltonian and predict energies of supported Pdn clusters of any structure. Low energy stable and metastable structures are identified using a Metropolis Monte Carlo-based genetic algorithm in the canonical ensemble at 300 K. We observe that supported single atoms sinter to form bilayer clusters and large cluster isomers share similarities in both shape and energy, and elucidate the significance of the support and microstructure on cluster stability. We discovered a simple surrogate structure-energy model, where the energy per atom scales with the square root of the average first coordination number, which can be used to estimate energies and compare the stability of clusters. Our framework, applicable to any metal/support system, fills an important methodological gap to predict the stability of supported metal catalysts in the subnanometer regime.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Measuring, Quantifying, and Predicting the Cost-Accuracy Tradeoff

Exponentially increasing data volumes, coupled with new modes of analysis have created significant new opportunities for data scientists. However, the stochastic nature of many data science techniques results in tradeoffs between costs and accuracy. For example, machine learning algorithms can be trained iteratively and indefinitely with diminishing returns in terms of accuracy. In this paper we explore the cost-accuracy tradeoff through three representative examples: we vary the number of models in an ensemble, the number of epochs used to train a machine learning model, and the amount of data used to train a machine learning model. We highlight the feasibility and benefits of being able to measure, quantify, and predict cost-accuracy tradeoffs by demonstrating the presence and usability of these tradeoffs in two different case studies.

big data analytics↗

Deep Learning Image Segmentation for Atmospheric Rivers

Abstract The identification of atmospheric rivers (ARs) is crucial for weather and climate predictions as they are often associated with severe storm systems and extreme precipitation, which can cause large impacts on society. This study presents a deep learning model, termed ARDetect, for image segmentation of ARs using ERA5 data from 1960 to 2020 with labels obtained from the TempestExtremes tracking algorithm. ARDetect is a convolutional neural network (CNN)-based U-Net model, with its structure having been optimized using automatic hyperparameter tuning. Inputs to ARDetect were selected to be the integrated water vapor transport (IVT) and total column water (TCW) fields, as well as the AR mask from TempestExtremes from the previous time step to the one being considered. ARDetect achieved a mean intersection-over-union (mIoU) rate of 89.04% for ARs, indicating its high accuracy in identifying these weather patterns and a superior performance than most deep learning–based models for AR detection. In addition, ARDetect can be executed faster than the TempestExtremes method (seconds vs minutes) for the same period. This provides a significant benefit for online AR detection, especially for high-resolution global models. An ensemble of 10 models, each trained on the same dataset but having different starting weights, was used to further improve on the performance produced by ARDetect, thus demonstrating the importance of model diversity in improving performance. ARDetect provides an effective and fast deep learning–based model for researchers and weather forecasters to better detect and understand ARs, which have significant impacts on weather-related events such as floods and droughts.

Galea, Daniel↗

Optimization of armored spherical tanks for storage on the lunar surface

A redundancy strategy for reducing micrometeroid armoring mass is investigated, with application to cryogenic reactant storage for a regenerative fuel cell (RFC) on the lunar surface. In that micrometeoroid environment, the cryogenic fuel must be protected from loss due to tank puncture. The tankage must have a sufficiently high probability of survival over the length of the mission so that the probability of system failure due to tank puncture is low compared to the other mission risk factors. Assuming that a single meteoroid penetration can cause a storage tank to lose its contents, two means are available to raise the probability of surviving micrometeoroid attack to the desired level. One can armor the tanks to a thickness sufficient to reduce probability of penetration of any tank to the desired level or add extra capacity in the form of spare tanks that results in survival of a given number out of the ensemble at the desired level. A combination of these strategies (armoring and redundancy) is investigated. The objective is to find the optimum combination which yields the lowest shielding mass per cubic meter of surviving fuel out of the original ensemble. The investigation found that, for the volumes of fuel associated with multikilowatt class cryo storage RFC's, and the armoring methodology and meteoroid models used, storage should be fragmented into small individual tanks. Larger installations (more fuel) pay less of a shielding penalty than small installations. For the same survival probability over the same time period, larger volumes will require less armoring mass per unit volume protected.

Bents, D. J.↗

Optimization of armored spherical tanks for storage on the lunar surface

A redundancy strategy for reducing micrometeroid armoring mass is investigated, with application to cryogenic reactant storage for a regenerative fuel cell (RFC) on the lunar surface. In that micrometeoroid environment, the cryogenic fuel must be protected from loss due to tank puncture. The tankage must have a sufficiently high probability of survival over the length of the mission so that the probability of system failure due to tank puncture is low compared to the other mission risk factors. Assuming that a single meteoroid penetration can cause a storage tank to lose its contents, two means are available to raise the probability of surviving micrometeoroid attack to the desired level. One can armor the tanks to a thickness sufficient to reduce probability of penetration of any tank to the desired level or add extra capacity in the form of space tanks that results in survival of a given number out of the ensemble at the desired level. A combination of these strategies (armor and redundancy) is investigated.

Bents, D. J.↗

Uncertainty quantification of a deep learning fuel property prediction model

Deep learning models are being widely used in the field of combustion. Given the black-box nature of typical neural network based models, uncertainty quantification (UQ) is critical to ensure the reliability of predictions as well as the training datasets, and for a principled quantification of noise and its various sources. Deep learning surrogate models for predicting properties of chemical compounds and mixtures have been recently shown to be promising for enabling data-driven fuel design and optimization, with the ultimate goal of improving efficiency and lowering emissions from combustion engines. In this study, UQ is performed for a multi-task deep learning model that simultaneously predicts the research octane number (RON), Motor Octane Number (MON), and Yield Sooting Index (YSI) of pure components and multicomponent blends. The deep learning model is comprised of three smaller networks: Extractor 1, Extractor 2, and Predictor, and a mixing operator. The molecular fingerprints of individual components are encoded via Extractor 1 and Extractor 2, the mixing operator generates fingerprints for mixtures/blends based on linear mixing operation, and the predictor maps the fingerprint to the target properties. Two different classes of UQ methods, Monte Carlo ensemble methods and Bayesian neural networks (BNNs), are employed for quantifying the epistemic uncertainty. Combinations of Bernoulli and Gaussian distributions with DropConnect and DropOut techniques are explored as ensemble methods. All the DropConnect, DropOut and Bayesian layers are applied to the predictor network. Aleatoric uncertainty is modeled by assuming that each data point has an independent uncertainty associated with it. The results of the UQ study are further analyzed to compare the performance of BNN and ensemble methods. Although this study is confined to UQ of fuel property prediction, the methodologies are applicable to other deep learning frameworks that are being widely used in the combustion community.

33 ADVANCED PROPULSION SYSTEMS↗

Large Scale Study of Ligand–Protein Relative Binding Free Energy Calculations: Actionable Predictions from Statistically Robust Protocols

The accurate and reliable prediction of protein–ligand binding affinities can play a central role in the drug discovery process as well as in personalized medicine. Of considerable importance during lead optimization are the alchemical free energy methods that furnish an estimation of relative binding free energies (RBFE) of similar molecules. Recent advances in these methods have increased their speed, accuracy, and precision. This is evident from the increasing number of retrospective as well as prospective studies employing them. However, such methods still have limited applicability in real-world scenarios due to a number of important yet unresolved issues. Here, we report the findings from a large data set comprising over 500 ligand transformations spanning over 300 ligands binding to a diverse set of 14 different protein targets which furnish statistically robust results on the accuracy, precision, and reproducibility of RBFE calculations. We use ensemble-based methods which are the only way to provide reliable uncertainty quantification given that the underlying molecular dynamics is chaotic. These are implemented using TIES (Thermodynamic Integration with Enhanced Sampling). Results achieve chemical accuracy in all cases. Ensemble simulations also furnish information on the statistical distributions of the free energy calculations which exhibit non-normal behavior. We find that the “enhanced sampling” method known as replica exchange with solute tempering degrades RBFE predictions. We also report definitively on numerous associated alchemical factors including the choice of ligand charge method, flexibility in ligand structure, and the size of the alchemical region including the number of atoms involved in transforming one ligand into another. Our findings provide a key set of recommendations that should be adopted for the reliable application of RBFE methods.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Quantum-Based Molecular Dynamics Simulations Using Tensor Cores

Tensor cores, along with tensor processing units, represent a new form of hardware acceleration specifically designed for deep neural network calculations in artificial intelligence applications. Tensor cores provide extraordinary computational speed and energy efficiency but with the caveat that they were designed for tensor contractions (matrix–matrix multiplications) using only low-precision floating-point operations. Despite this perceived limitation, we demonstrate how tensor cores can be applied with high efficiency to the challenging and numerically sensitive problem of quantum-based Born–Oppenheimer molecular dynamics, which requires highly accurate electronic structure optimizations and conservative force evaluations. The interatomic forces are calculated on-the-fly from an electronic structure that is obtained from a generalized deep neural network, where the computational structure naturally takes advantage of the exceptional processing power of the tensor cores and allows for high performance in excess of 100 Tflops on a single Nvidia A100 GPU. Stable molecular dynamics trajectories are generated using the framework of extended Lagrangian Born–Oppenheimer molecular dynamics, which combines computational efficiency with long-term stability, even when using approximate charge relaxations and force evaluations that are limited in accuracy by the numerically noisy conditions caused by the low-precision tensor core floating-point operations. A canonical ensemble simulation scheme is also presented, where the additional numerical noise in the calculated forces is absorbed into a Langevin-like dynamics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Avalanching strain dynamics during the hydriding phase transformation in individual palladium nanoparticles

Phase transitions in reactive environments are crucially important in energy and information storage, catalysis and sensors. Nanostructuring active particles can yield faster charging/ discharging kinetics, increased lifespan and record catalytic activities. However, establishing the causal link between structure and function is challenging for nanoparticles, as ensemble measurements convolve intrinsic single-particle properties with sample diversity. Here we study the hydriding phase transformation in individual palladium nanocubes in situ using coherent X-ray diffractive imaging. The phase transformation dynamics, which involve the nucleation and propagation of a hydrogen-rich region, are dependent on absolute time (aging) and involve intermittent dynamics (avalanching). A hydrogen-rich surface layer dominates the crystal strain in the hydrogen-poor phase, while strain inversion occurs at the cube corners in the hydrogen-rich phase. A three-dimensional phase-field model is used to interpret the experimental results. In conclusion, our experimental and theoretical approach provides a general framework for designing and optimizing phase transformations for single nanocrystals in reactive environments.

36 MATERIALS SCIENCE↗

SU-E-T-191: First Principle Calculation of Quantum Yield in Photodynamic Therapy

Purpose: We present a first-principle method to calculate the spin transfer efficiency in oxygen induced by any photon fields especially in MeV energy range. The optical pumping is mediated through photosensitizers, e.g., porphyrin and/or ensemble of quantum dots. Methods: Under normal conditions, oxygen molecules are in the relatively non-reactive triplet state. In the presence of certain photosensitizer compounds such as porphyrins, electromagnetic radiation of specific wavelengths can excite oxygen to highly reactive singlet state. With selective uptake of photosensitizers by certain malignant cells, photon irradiation of phosensitized tumors can lead to selective killing of cancer cells. This is the basis of photodynamic therapy (PDT). Despite several attempts, PDT has not been clinically successful except in limited superficial cancers. Many parameters such as photon energy, conjugation with quantum dots etc. can be potentially combined with PDT in order to extend the role of PDT in cancer management. The key quantity for this optimization is the spin transfer efficiency in oxygen by any photon field. The first principle calculation model presented here, is an attempt to fill this need. We employ stochastic density matrix description of the quantum jumps and the rate equation methods in quantum optics based on Markov/Poisson processes and calculate time evolution of the population of the optically pumped singlet oxygen. Results: The results demonstrate the feasibility of our model in showing the dependence of the optical yield in generating spin-singlet oxygen on the experimental conditions. The adjustable variables can be tuned to maximize the population of the singlet oxygen hence the efficacy of the photodynamic therapy. Conclusion: The present model can be employed to fit and analyze the experimental data and possibly to assist researchers in optimizing the experimental conditions in photodynamic therapy.

60 APPLIED LIFE SCIENCES↗

Gearbox bearing crack growth prognostics and uncertainty quantification with physics-informed machine learning

This paper introduces the extreme theory of functional connections (X-TFC), a physics-informed machine learning algorithm, and tailors it to estimate the remaining useful life (RUL) of wind turbine gearbox bearings experiencing fatigue crack growth. Unlike purely data-driven methods, X-TFC embeds a physics model, based on Head's theory in this work, into its training objective. The core of X-TFC is a random-projection single-layer neural network trained via an extreme learning machine, which requires only limited damage progression data and solves for output weights with a least-squares optimization algorithm. A composite loss function balances the network's fit to observed degradation data against the residuals of the governing crack growth differential equation, ensuring the learned damage trajectory remains physically plausible. When applied to a vibration-based health-index (HI) dataset measured during the growth of a crack on the inner ring of a high-speed bearing in a wind turbine gearbox (Bechhoefer and Dubé, 2020), X-TFC achieves near-zero prediction bias. Even when trained on only the first 10 %–20 % of the damage progression data, with sufficient physics weighting its predictions remain monotonic and smooth, delivering high prognosability and trendability. To quantify the epistemic uncertainty, we employ a Monte Carlo ensemble of independently initialized X-TFC models trained on noise-perturbed data, which yields confidence intervals around each RUL estimate and captures both model-parameter and epistemic uncertainty. In addition to a vibration-based HI, we demonstrate that the proposed framework can be directly applied to a supervisory control and data acquisition (SCADA) data-based HI (Eftekhari Milani et al., 2026) measured during similar wind turbine gearbox bearing crack faults, preserving its accuracy and interpretability. This extension shows the versatility of our approach, which is applicable to bearings of multiple gearbox manufacturers, models, and ratings using only SCADA data. By integrating domain knowledge with machine learning, X-TFC offers a rapid, reliable tool for crack prognostics. Its adaptability to other bearing failure modes, such as pitch bearing ring cracks, positions X-TFC as a powerful enabler of data-driven, physics-informed asset management in the wind energy sector and beyond.

17 WIND ENERGY↗

THE IMPACT OF CORRELATED PROJECTIONS ON WEAK LENSING CLUSTER COUNTS

Large-scale structure projections are an obstacle in converting the shear signal of clusters detected in weak-lensing maps into virial masses. However, this step is not necessary for constraining cosmology with the shear-peak abundance, if we are able to predict its amplitude. We generate a large ensemble of N-body simulations spanning four cosmological models, with total volume V{sub tot} approx 1(h {sup -1} Gpc){sup 3} per model. Variations to the matter density parameter and amplitude of fluctuations are considered. We measure the abundance of peaks in the mass density projected in approx100 h {sup -1} Mpc slabs to determine the impact of structures spatially correlated with the simulation clusters, identified by the three-dimensional (3D) friends-of-friends (FoF) algorithm. The halo model shows that the choice of the smoothing filter for the density field is important in reducing the contribution of correlated projections to individual halo masses. Such contributions are less than 2% in the case of the optimal, compensated filter used throughout this analysis. We measure the change in the mass of peaks when projected in slabs of various thicknesses. Peaks in slabs of 26 h {sup -1} Mpc and 102 h {sup -1} Mpc suffer an average mass change of less than 2% compared to their mass in slabs of 51 h {sup -1} Mpc. We then explore the cosmology dependence of the projected-peak mass function, and find that, for a wide range of slab thicknesses (< 500 h {sup -1} Mpc), it scales with cosmology in exactly the same way as the 3D FoF mass function and the Sheth-Tormen (ST) formula. This extends the earlier result of Marian et al. Finally, we show that for all cosmological models considered, the low and intermediate mass bins of the peak abundance can be described using a modified ST functional form to within 10%-20% accuracy.

79 ASTRONOMY AND ASTROPHYSICS↗

Dark Energy Survey Year 3 results: A 2.7% measurement of baryon acoustic oscillation distance scale at redshift 0.835

We present angular diameter measurements obtained by measuring the position of baryon acoustic oscillations (BAO) in an optimized sample of galaxies from the first three years of Dark Energy Survey data (DES Y3). The sample consists of 7 million galaxies distributed over a footprint of 4100 deg 2 with 0.6 < z photo < 1.1 and a typical redshift uncertainty of 0.03 ( 1 + z ) . The sample selection is the same as in the BAO measurement with the first year of DES data, but the analysis presented here uses three times the area, extends to higher redshift, and makes a number of improvements, including a fully analytical BAO template, the use of covariances from both theory and simulations, and an extensive preunblinding protocol. We used two different statistics; angular correlation function and power spectrum, and validate our pipeline with an ensemble of over 1500 realistic simulations. Both statistics yield compatible results. We combine the likelihoods derived from angular correlations and spherical harmonics to constrain the ratio of comoving angular diameter distance D M at the effective redshift of our sample to the sound horizon scale at the drag epoch. We obtain D M ( z eff = 0.835 ) / r d = 18.92 ± 0.51 , which is consistent with, but smaller than, the Planck prediction assuming flat Λ CDM , at the level of 2.3 σ . The analysis was performed blind and is robust to changes in a number of analysis choices. It represents the most precise BAO distance measurement from imaging data to date, and is competitive with the latest transverse ones from spectroscopic samples at z > 0.75 . When combined with DES 3 x 2 pt + SNIa , they lead to improvements in H 0 and Ω m constraints by ∼ 20 % .

79 ASTRONOMY AND ASTROPHYSICS↗

Modular droplet injector for sample conservation providing new structural insight for the conformational heterogeneity in the disease-associated NQO1 enzyme

Droplet injection strategies are a promising tool to reduce the large amount of sample consumed in serial femtosecond crystallography (SFX) measurements at X-ray free electron lasers (XFELs) with continuous injection approaches. Here, we demonstrate a new modular microfluidic droplet injector (MDI) design that was successfully applied to deliver microcrystals of the human NAD(P)H:quinone oxidoreductase 1 (NQO1) and phycocyanin. We investigated droplet generation conditions through electrical stimulation for both protein samples and implemented hardware and software components for optimized crystal injection at the Macromolecular Femtosecond Crystallography (MFX) instrument at the Stanford Linac Coherent Light Source (LCLS). Under optimized droplet injection conditions, we demonstrate that up to 4-fold sample consumption savings can be achieved with the droplet injector. In addition, we collected a full data set with droplet injection for NQO1 protein crystals with a resolution up to 2.7 Å, leading to the first room-temperature structure of NQO1 at an XFEL. NQO1 is a flavoenzyme associated with cancer, Alzheimer's and Parkinson's disease, making it an attractive target for drug discovery. Further, our results reveal for the first time that residues Tyr128 and Phe232, which play key roles in the function of the protein, show an unexpected conformational heterogeneity at room temperature within the crystals. These results suggest that different substates exist in the conformational ensemble of NQO1 with functional and mechanistic implications for the enzyme's negative cooperativity through a conformational selection mechanism. Our study thus demonstrates that microfluidic droplet injection constitutes a robust sample-conserving injection method for SFX studies on protein crystals that are difficult to obtain in amounts necessary for continuous injection, including the large sample quantities required for time-resolved mix-and-inject studies.

59 BASIC BIOLOGICAL SCIENCES↗

Disentangling the impact of quasiparticles and two-level systems on the statistics of superconducting-qubit lifetime

Temporal fluctuations in the superconducting qubit lifetime, T 1 , present additional challenges in the pursuit of fault-tolerant quantum computing. Although the exact mechanisms remain unclear, T 1 fluctuations are generally attributed to strong coupling between the qubit and a few near-resonant two-level systems (TLSs), which can exchange energy with an ensemble of thermally fluctuating two-level fluctuators (TLFs) at low frequencies. Here, we report T 1 measurements of qubits with varying geometrical footprints and surface dielectrics as a function of temperature. By analyzing the noise spectrum of the qubit depolarization rate, Γ 1 = 1 / T 1 , we disentangle the contributions of TLSs, nonequilibrium quasiparticles (QPs), and equilibrium (thermally excited) QPs to the variance in Γ 1 . We find that the Γ 1 variance in qubits with smaller footprints is more susceptible to QP and TLS fluctuations than that in larger-footprint qubits. Furthermore, the QP-induced variances in all qubits align with the theoretical framework of QP diffusion and fluctuation. These findings offer valuable insights for future qubit design and engineering optimization.

Zhu, Shaojiang [Fermilab] (ORCID:0000000293180092)↗

Uncertainty Assessment of Space-Borne Passive Soil Moisture Retrievals

The uncertainty associated with passive soil moisture retrieval is hard to quantify, and known to be underlain by various, diverse, and complex causes. Factors affecting space-borne retrieved soil moisture estimation include: (i) the optimization or inversion method applied to the radiative transfer model (RTM), such as e.g. the Single Channel Algorithm (SCA), or the Land Parameter Retrieval Model (LPRM), (ii) the selection of the observed brightness temperatures (Tbs), e.g. polarization and incidence angle, (iii) the definition of the cost function and the impact of prior information in it, and (iv) the RTM parameterization (e.g. parameterizations officially used by the SMOS L2 and SMAP L2 retrieval products, ECMWF-based SMOS assimilation product, SMAP L4 assimilation product, and perturbations from those configurations). This study aims at disentangling the relative importance of the above-mentioned sources of uncertainty, by carrying out soil moisture retrieval experiments, using SMOS Tb observations in different settings, of which some are mentioned above. The ensemble uncertainties are evaluated at 11 reference CalVal sites, over a time period of more than 5 years. These experimental retrievals were inter-compared, and further confronted with in situ soil moisture measurements and operational SMOS L2 retrievals, using commonly used skill metrics to quantify the temporal uncertainty in the retrievals.

Quets, Jan↗

Specific Features of the Electron Structure of ZnTPP Aggregated Forms: Data of Optical Measurements and Quantum-Chemical Calculations

Zinc tetraphenylporphyrin (ZnTPP) films and ZnTPP-based composites are promising materials of organic photonics. Porphyrins are inclined towards self-assembly and the formation of molecular ensembles or differently structured aggregates. The energies of the formation of aggregates and ordered structures and modifications of their electron structures compared to that of a free ZnTPP molecule have not been adequately explored. In the study, the first comprehensive investigation of the structure, absorption and luminescence spectra, and photoluminescence kinetics of structurally perfect ZnTPP thin films produced under quasi-equilibrium conditions in vacuum is conducted. It is shown that changes in the absorption spectra and the red shift of the luminescence spectra of films by 0.15 eV from the spectra observed for ZnTPP solutions in toluene can be interpreted as a result of the formation of an ordered thin-film structure through the dimerization of porphyrin planar molecules under nearly equilibrium conditions. The optimal geometric structure, the energy of the ground state, and the electronic spectrum of excitations of the dimerized (ZnTPP){sub 2} state are calculated in the context of density functional theory and time-dependent density functional theory. The energy gain on the formation of a dimer of symmetry C{sub i} compared to two separate molecules is 0.23 eV per dimer; the HOMO–LUMO gap for the dimer is decreased by 70 meV. The radiative emission time in the ordered solid phase is an order of magnitude shorter than that in the solution in toluene and corresponds to 277 ps, which is typical of J aggregates of porphyrins.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Uncovering Structure–Conductivity Relationships in Anion Exchange Membranes (AEMs) Using Interpretable Machine Learning

Anion exchange membranes (AEMs) play a vital role in the performance of water electrolyzers and fuel cells, yet their discovery and optimization remain challenging due to the complexity of structure–property relationships. In this study, we introduce a machine learning framework that leverages conditional graph neural networks (cGNNs) and descriptor-based models and a hybrid graph neural network (HGARE) to predict and interpret ionic conductivity. The descriptor-based pipeline employs principal component analysis (PCA), ablation, and SHAP analysis to identify factors governing anion conductivity, revealing electronic, topological, and compositional descriptors as key contributors. Beyond prediction, dimensionality reduction and clustering are performed by employing t-SNE and KMeans as well as SOM, which reveal distinct membranes clusters, some of which were enriched with high anion conductivity. Among graph-based approaches, the graph convolutional (GCN) achieved strong predictive performance, while the Hybrid Graph Autoencoder-Regressor Ensemble (HGARE) achieved the highest accuracy. Additionally, atom-level saliency maps from GCN provide spatial explanations for conductive behavior, revealing the importance of polarizable and flexible regions. This work contributes to the accelerated and data-driven design of high-performance AEMs.

Naghshnejad, Pegah [Department of Chemical Enginee↗