Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “tailoring”

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 1,081 records · Page 60
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Depicting the roles of CuO secondary phase and heat treatment in driving the magnetic and magnetocaloric features of Pr 2/3 Sr 1/3 MnO 3 manganite

In this work, we report a detailed experimental study regarding the impact of copper oxide (CuO) secondary phase and heat treatments on the structural, magnetic, and magnetocaloric properties of the near-room temperature Pr 2/3 Sr 1/3 MnO 3 (PSMO) magnetic refrigerant. Our investigations are carried out by using structural and microstructural analyses, alongside magnetization measurements. Here, the analysis of X-ray diffraction data of PSMO(95%)-CuO(5%) (PSMO-CuO) samples shows the coexistence of both CuO and PSMO phases. Further, the microstructural analysis of PSMO-CuO reveals that the addition of CuO significantly enhances the grains size. On the other hand, the added secondary phase markedly reduces the Curie temperature (T c ) from 293 K for PSMO to about 273 K for PSMO-CuO composite while increasing the magnetocaloric effect. This decrease in T c is associated with a significant change from 162° to about 156° in the Mn-OII-Mn bond angle respectively. Moreover, performed investigations regarding the role of heat treatments unveil that the observed changes in the structural and magnetic features are mainly driven by the secondary phase that modify grains size and double-exchange interactions in the PSMO compound. Interestingly, our findings demonstrate that the Curie temperature of the PSMO and accordingly its magnetocaloric effect can be tailored by adding small amounts of CuO without need to substitution on cation sites. In the light of obtained results, a multilayered refrigerant composed of PSMO and PSMO-CuO is proposed to cover the magnetic cooling temperature range close to room-temperature. The resulting entropy change remains practically constant between 273 K and 293 K. Such a behavior is highly appreciated from a practical point of view, particularly in cases where the cooling process is carried out by using the AMR and Ericsson cycles.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Graded microstructure and mechanical properties of spark plasma sintered Fe-Cr alloys

Developing graded microstructure and mechanical properties is critical for accelerating the design and optimization of structural materials for a wide range of applications. Here, in this study, Fe-Cr alloys were fabricated by spark plasma sintering (SPS) technique. The microstructure, microhardness, and tensile properties of the as-fabricated and thermally annealed variants were investigated. Graded grain structures were created in the as-fabricated cylinders along both axial and radial directions. Grain size was gradually reduced from the sample periphery to the center. Microhardness measurements reveal a gradual decrease of hardness towards the periphery of the samples. The as-fabricated Fe-Cr alloys show a desired combination of tensile strength and elongation, primarily due to the formation of a high density of oxides, voids, dislocations, and grain boundaries. After in-situ thermal annealing at 600 °C under 60 MPa for 2 h using SPS, the Fe-Cr alloys underwent minimal grain growth, and the graded grain structures were retained. The hardness was more uniformly distributed in the annealed variants, and the tensile strength was reduced with an increase in the total elongation, which is attributed to the dissolution of nano-sized oxide particles and the relief of the residual stress. This study demonstrates that SPS coupled with subsequent heat treatment can tailor the graded microstructure and control the mechanical properties of Fe-Cr alloys, showing potential applications in other alloy systems.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Interplay of pressure and temperature in modulating the magnetic properties of Cr 2 O 3

The spin-lattice interaction is pivotal in tailoring materials properties and is crucial for developing novel spintronic devices. Here, in the current work, we aim to explore the effects of pressure and temperature on the magnetic properties of Cr 2 O 3 to gain insights into its spin-lattice interaction. Through high-pressure neutron diffraction experiments, we observed an enhancement of the magnetic Bragg intensity with increasing pressure and it becomes more pronounced as the temperature increases. Results from first-principles calculations reveal a strengthening of the easy-axis magnetic anisotropy, doubling from 0 to 20 GPa. The exchange parameters, calculated based on this spin orientation, show an enhancement of the dominant magnetic interactions and an increase of magnetic ordering temperature when pressure increases. These findings suggest that the contributions from these two mechanisms are responsible for the observed increase in magnetic Bragg intensity.

Cr2O3↗
DOE OSTI2025

Directed energy deposition of radially functionally graded nickel based superalloys IN625 and IN738: Microstructure and mechanical properties

This study investigates the microstructure and mechanical properties of radially functionally graded (RFG) materials fabricated from nickel-based superalloys IN625 and IN738 using directed energy deposition (DED). RFG cylindrical deposits with radial gradients of IN625-IN738 and IN738-IN625 were successfully fabricated and confirmed to be crack-free through tailored composition grading, scan strategy and process parameters that minimized thermal gradients and solidification velocities. Microstructural analysis revealed distinct variations across the gradient, characterized by differences in dendrite arm spacing, microsegregation, and the formation of distinct γ' precipitates and carbide/boride phases (MC, M 6 C, M 2 C, and MB 2 ) in as-deposited (AD) conditions, which evolved into M 23 C 6 and M 2 B upon subsequent heat treatment (HT). EBSD showed predominant columnar grains with a [100] texture, while microhardness profiles demonstrated consistent variation across the gradients. Mechanical testing highlighted the synergy of the graded structure, combining IN738's strength with IN625's ductility, resulting in enhanced properties such as a yield strength of 1131 MPa, ultimate tensile strength of 1382 MPa, and elongation of 12 % for HT IN625-IN738 grading. Numerical simulations employing CALPHAD thermodynamic calculations and finite element thermal modeling are used to establish phase evolution and microstructural feature correlation with thermal profiles, aiding in the optimization of process-structure-property relationship.

Additive manufacturing↗
DOE OSTI2025

Multi-energy ion irradiation effects and distinctive features of color center generation in yttria-stabilized zirconia driven by structural discrepancy

Regarding a requirement for inert matrix fuel (IMF) to burn minor actinides and plutonium and high-level waste immobilization, Yttria-stabilized zirconia (ZrO 2 -6.5 wt% Y 2 O 3 ) (YSZ) crystals with cubic phase are selected to comprehensively evaluate the irradiation resistance by irradiation-induced structural discrepancies across multiple energy conditions. Electronic excitation and nuclear collisions reveal distinct color centers, influenced by varying component ratios that reflect their formation mechanisms. Key factors driving ultra-fast structural transitions include collision cascades, pressure waves, and energy dissipation from electronic excitation. In this work, we find that in the electronic energy loss (E ele )-dominant region, internal latent tracks with unique surface nanostructures emerge, distinct from the defect clusters seen in the nuclear energy loss (E nuc )-dominant region, while the threshold for latent track formation and melting are identified. In conclusion, structural discrepancies driven by different energy-loss mechanisms promote the generation of singly ionized and nearest-neighbor doubly ionized oxygen vacancies, which respectively dominate the formation of F⁺ and T-centers, resulting in bandgap narrowing (4.80 eV → 4.73 eV), and furthermore triggering visible light emission (2.17 eV → 2.20 eV) in irradiated YSZ crystals, thus, enabling the design of novel irradiation-tailored material functionalities.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Microstructural Engineering of Cu-Rich Nanoprecipitate formation in NiCoFeCrCu0.12 High-Entropy Alloy via Severe Plastic Deformation for Enhanced Irradiation Tolerance

This study demonstrates a defect-engineering approach for controlling Cu-rich precipitates in FeNiCrCoCu0.2 high-entropy alloys (Cu-HEAs), delivering a novel pathway for next-generation nuclear reactor materials with superior irradiation resistance. This work establishes that severe plastic deformation (SPD) processing via Shear Assisted Processing and Extrusion (ShAPE) and Friction Stir Layer Deposition (FSLD) creates dense dislocation networks and subgrain boundaries that fundamentally alter precipitation behavior under identical thermal treatments. Atom probe tomography (APT) indicates that SPD produces a metastable, atomically homogeneous solid solution that, upon moderate heat treatment (500°C/10 hour), develops remarkedly stronger Cu clustering than the as-cast counterpart. High-temperature exposure (800°C/100 h) produces near-pure Cu precipitates (~90 at% Cu) with significantly enhanced defect-sink efficacy in SPD-processed alloys: precipitate sizes of 50-60 nm and number densities of 2.7-3.8 × 10¹7 m?³, compared to 89 nm and 0.44 × 10¹7 m?³ in as-cast materials. Collectively, the findings establish defect-mediated precipitation control as a scalable, high-impact route to tailor sink density and distribution in HEAs, enabling microstructures optimized for irradiation tolerance and mechanical robustness in nuclear reactor environments.

Meher, Subhashish↗
DOE OSTI2026

scPlantAnnotate: an accurate and robust transformer-based model for plant cell type annotation

Accurate cell type annotation remains a major bottleneck in plant single-cell RNA sequencing (scRNA-seq), where existing tools are often adapted from animal studies and perform sub-optimally on plant data. The lack of plant-specific computational frameworks limits the construction of plant cell atlases and downstream biological discovery. We develop and evaluate scPlantAnnotate, a Transformer-based reference annotation framework tailored for plant scRNA-seq data, and benchmark it against state-of-the-art deep learning and conventional methods across multiple plant species. Species-specific scPlantAnnotate models were trained using curated datasets from Arabidopsis thaliana, Zea mays, Oryza sativa, and Glycine max. We compared scPlantAnnotate with leading baselines under both standard random-split evaluation and a more stringent leave-one-dataset-out setting, which tests robustness to completely unseen datasets and tissue types. scPlantAnnotate consistently outperforms existing approaches across all four species under random-split evaluation. In the leave-one-dataset-out setting for A. thaliana, where performance drops markedly for all methods due to strong batch effects and dataset heterogeneity, scPlantAnnotate nonetheless achieves the highest Accuracy, Macro-F1, Balanced Accuracy, and Macro-AUROC on average and ranks first on most held-out datasets. These results demonstrate improved robustness to dataset shifts, a critical yet underexplored challenge in plant scRNA-seq analysis. A freely accessible web server enables users to annotate their own datasets using pretrained models. scPlantAnnotate provides a plant-specific, Transformer-based framework for single-cell annotation that delivers state-of-the-art performance and enhanced robustness to unseen datasets. By addressing limitations of existing tools and enabling scalable reference-based annotation, scPlantAnnotate supports the development of comprehensive plant cell atlases and facilitates broader use of single-cell genomics in plant biology.

Bioinformatics↗
DOE OSTI2021

Distinct protein architectures mediate species-specific beta-glucan binding and metabolism in the human gut microbiota

Complex glycans that evade our digestive system are major nutrients that feed the human gut microbiota (HGM). The prevalence of Bacteroidetes in the HGM of populations worldwide is engendered by the evolution of polysaccharide utilization loci (PULs), which encode concerted protein systems to utilize the myriad complex glycans in our diets. Despite their crucial roles in glycan recognition and transport, cell-surface glycan-binding proteins (SGBPs) remained understudied cogs in the PUL machinery. Here, we report the structural and biochemical characterization of a suite of SGBP-A and SGBP-B structures from three syntenic β(1,3)-glucan utilization loci (1,3GULs) from Bacteroides thetaiotaomicron (Bt), Bacteroides uniformis (Bu), and B. fluxus (Bf), which have varying specificities for distinct β-glucans. Ligand complexes provide definitive insight into β(1,3)-glucan selectivity in the HGM, including structural features enabling dual β(1,3)-glucan/mixed-linkage β(1,3)/β(1,4)-glucan-binding capability in some orthologs. The tertiary structural conservation of SusD-like SGBPs-A is juxtaposed with the diverse architectures and binding modes of the SGBPs-B. Specifically, the structures of the trimodular BtSGBP-B and BuSGBP-B revealed a tandem repeat of carbohydrate-binding module-like domains connected by long linkers. In contrast, BfSGBP-B comprises a bimodular architecture with a distinct β-barrel domain at the C terminus that bears a shallow binding canyon. The molecular insights obtained here contribute to our fundamental understanding of HGM function, which in turn may inform tailored microbial intervention therapies.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Excipient screening by lyophilization provides insights into spray drying formulations for nanoparticle vaccines

Nanoparticles have shown great promise as delivery platforms in the development of tunable and safe vaccines. Nanolipoprotein particles (NLPs), also known as nanodiscs, are discoidal nanoparticles composed of a lipid bilayer stabilized at their periphery by apolipoproteins. Under the right conditions, the NLP self-assembly process is highly customizable in terms of lipids and apolipoprotein constituents, allowing for tunable physical and chemical characteristics. This flexibility allows a wide range of vaccine antigens and adjuvants to be incorporated onto the NLP platform for tailored vaccine design. The stability of NLPs during long term storage is a very important factor in developing a vaccine delivery platform suitable for widespread global use. When stored in a solution for extended periods of time, NLPs dissociate into their corresponding lipids and protein constituents, leading to particle degradation. Proper stabilization of NLPs can often be achieved by lyophilization (i.e. freeze-drying), a method widely used for various applications including pharmaceuticals. This process, however, can be damaging to particles without the presence of lyoprotectants or excipients that help maintain particle stability during lyophilization. Another method used to stabilize vaccines and pharmaceuticals is spray drying, a process that converts liquid formulations into dry powders through controlled heating and airflow. While spray drying is rapid, scalable, and cost-effective, lyophilization is typically a gentler process that better retains biomolecule structure and function. Both processes use excipients for particle stabilization, so lyophilization can be used as a surrogate to test stability of NLPs, to down-select formulations that may withstand the harsher conditions of spray drying. To screen different formulations, NLPs were synthesized and purified to homogeneity by size exclusion chromatography (SEC) and samples were prepared with a wide range of lyoprotectants and/or excipients. To assess the protective effects of excipients on NLPs upon spray drying, both pre- and post-lyophilized samples were analyzed by SEC. To assess the protective effects upon heating (encountered during the spray drying process), NLP samples were incubated at elevated temperatures prior to SEC analysis. The lyoprotectants and excipients evaluated in this study had different efficiencies in protecting NLPs during lyophilization and heating tests. Trehalose, for example, exhibits stabilization on NLPs both upon lyophilization and heating whereas leucine accelerated NLP dissociation. Although some lyoprotectants are effective by themselves, different combinations can decrease the stabilization of NLPs. Shelf-stable vaccines that do not require cold-chain storage are essential for global accessibility and our findings provide fundamental insight into how to advance NLP-based vaccines for these applications.

Serrano, Litzay J [Lawrence Livermore National Lab↗
DOE OSTI2024

EHR-BERT: A BERT-based model for effective anomaly detection in electronic health records

Objective: Physicians and clinicians rely on data contained in electronic health records (EHRs), as recorded by health information technology (HIT), to make informed decisions about their patients. The reliability of HIT systems in this regard is critical to patient safety. Consequently, better tools are needed to monitor the performance of HIT systems for potential hazards that could compromise the collected EHRs, which in turn could affect patient safety. In this paper, we propose a new framework for detecting anomalies in EHRs using sequence of clinical events. This new framework, EHR-Bidirectional Encoder Representations from Transformers (BERT), is motivated by the gaps in the existing deep-learning related methods, including high false negatives, sub-optimal accuracy, higher computational cost, and the risk of information loss. EHR-BERT is an innovative framework rooted in the BERT architecture, meticulously tailored to navigate the hurdles in the contemporary BERT method; thus, enhancing anomaly detection in EHRs for healthcare applications.Methods: The EHR-BERT framework was designed using the Sequential Masked Token Prediction (SMTP) method. This approach treats EHRs as natural language sentences and iteratively masks input tokens during both training and prediction stages. This method facilitates the learning of EHR sequence patterns in both directions for each event and identifies anomalies based on deviations from the normal execution models trained on EHR sequences.Results: Extensive experiments on large EHR datasets across various medical domains demonstrate that EHR-BERT markedly improves upon existing models. It significantly reduces the number of false positives and enhances the detection rate, thus bolstering the reliability of anomaly detection in electronic health records. This improvement is attributed to the model’s ability to minimize information loss and maximize data utilization effectively.Conclusion: EHR-BERT showcases immense potential in decreasing medical errors related to anomalous clinical events, positioning itself as an indispensable asset for enhancing patient safety and the overall standard of healthcare services. The framework effectively overcomes the drawbacks of earlier models, making it a promising solution for healthcare professionals to ensure the reliability and quality of health data.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2022

Dependency of CO 2 methanation on the strong metal-support interaction for supported Ni/CeO 2 catalysts

The strong metal-support interaction (SMSI) for supported Ni/CeO 2 catalysts with different CeO 2 nanomorphologies was systematically explored. The degree of encapsulation of Ni particles originating from the SMSI effect was found to follow the trend of Ni/CeO 2 -(1 1 1) > Ni/CeO 2 -(1 0 0) > Ni/CeO 2 -(110 + 100), which parallels the CO 2 hydrogenation activity. Quasi in situ XPS reveals the presence of Ce 3+ sites in accordance with the formation of an amorphous surface CeOx layer encapsulating the Ni nanoparticles. In situ DRIFTS indicates the reaction pathway and rate-determining step are dependent on the degree of the SMSI effect, leading to distinct selectivities towards CH 4 , especially at a high weight hourly space velocity (WHSV). Finally, these findings present a fundamental strategy about tailoring catalytic performance through support facet dependent susceptibility of SMSI phenomena.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Encapsulation of hygroscopic liquids via polymer precipitation in non-aqueous emulsions

Here, encapsulation of ionic liquids (ILs) and phase change materials (PCMs) can overcome limitations associated with bulk materials, e.g., slow mass transfer rates, high viscosities, or susceptibility to external environment. Single step soft-templated encapsulation methods commonly use interfacial polymerization for shell formation, with a multifunctional monomer in the continuous phase and another in the discontinuous phase, and thus do not give pristine core material. We posit that polymer precipitation onto emulsion droplets in non-aqueous emulsions could produce a robust shell without contamination of the core, ideal for the encapsulation of water-sensitive or water-miscible materials. Solutions of commodity polymers were added to the continuous phase of non-aqueous Pickering emulsions stabilized by alkylated graphene oxide (GO) nanosheets such that the change in solubility of the polymer led to formation of robust shells and the production of capsules that could be isolated. We demonstrate that a polymer precipitation approach can produce capsules with pristine core of the IL 1-ethyl-3-methylimidazolium hexafluorophosphate [Emim][PF 6 ] or the salt hydrate PCM magnesium nitrate hexahydrate (MNH) and shell of nanosheets and polystyrene, poly(methyl methacrylate), or polyethylene. The capsules are approximately 80 wt% [Emim][PF 6 ] or >90 wt% MNH, and the core can undergo multiple cycles of solidification and melting without leakage or destruction. This novel, single-step methodology provides a distinct advantage to access capsules with pristine core composition and is amenable to different core and shell, paving the way for tailoring capsule composition for desired applications.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Growth mechanism of glucose-based hydrochar under the effects of acid and temperature regulation

Improving the tailorability of hydrochar synthesis is an effective way to enhance its performance and utilization efficiency. In this study, the growth rate, morphology, and molecular structure of hydrochar were controlled by regulating the pH and temperature of the hydrothermal carbonization process. Growth process analysis indicates that hydrochar has three growth periods: induction, rapid, and stable growth periods. It is mainly controlled by 5-hydroxymethylfurfural (HMF), which is formed by converting glucose and its transformation products. Further, the regulation of acid can significantly shorten the induction period of hydrochar, even under low-temperature conditions (<180 °C), and increase the growth rate of hydrochar. However, the degree of hydrochar adhesion varies: the lower the temperature, the greater the degree of its adhesion. Molecular structural analysis demonstrates that hydrochar mainly consists of furan structural domains and aromatic clusters, and its surface is rich in oxygen-containing functional groups. The degree at which hydrochar was aromatized was improved by increasing the reaction temperature (160–220 °C); whereas the regulation of acid reduced it and increased the content of oxygen-containing functional groups on the hydrochar surface. Based on these results, it is proposed that hydrochar formation has five stages and three growth periods with or without acid regulation.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Quantitative SANS and multi-model analysis of spacer-dependent micellization of urea-based gemini surfactants

The micellization behavior of urea-based cationic gemini surfactants was investigated using small-angle neutron scattering (SANS) with multi-model form factor analysis. A homologous series of surfactants with urea group included in the hydrophobic tail and polymethylene spacers consisting of two to ten methylene units was analyzed using three form factor models: a core–shell ellipsoid and two variants of homogeneous ellipsoids. The results from all models show a consistent trend of the micelle structures, confirming that the spacer length critically influences micellar geometry, aggregation number, and hydration. The surfactant with four CH 2 groups in the spacer formed the largest micelles with the highest aggregation number, while longer spacers led to progressively smaller, more compact aggregates. The shell hydration—quantified as the volume fraction of heavy water within the hydrophilic region—decreased systematically with increasing spacer length due to enhanced hydrophobicity of the headgroup-spacer region. Intermicellar interactions, modeled as screened Coulomb interaction using the rescaled mean spherical approximation (RMSA), revealed the strongest electrostatic repulsion for the case of four methylene groups in the spacer, corresponding to the highest micellar charge and largest interparticle spacing. The observed spacer-dependent trends were robust across all modeling approaches, demonstrating that the spacer length serves as a key structural determinant of self-assembly in this type of urea-based gemini systems. These findings provide insight into the design of gemini surfactants with tailored aggregation behavior for applications in drug delivery, nanostructure templating, and solubilization technologies.

Core–shell ellipsoid model↗
DOE OSTI2021

Performance of hybridized bamboo-carbon fiber reinforced polypropylene composites processed using wet laid technique

The end-of-life vehicles (ELV) regulations motivate hybrid materials usage in automotive industries to optimize properties at reduced cost and increase eco-friendly designs. This research explores hybrid compositions of natural bamboo fiber and synthetic carbon fiber. The goal of hybridization was to synergistically benefit from each constituent– i.e., sustainability, energy absorption and superior damping from bamboo, and high strength and stiffness from carbon fiber. Carbon fibers (CF), bamboo fibers (BF) and polypropylene (PP) fibers were dispersed in water to produce wet-laid hybrid mats. The mats were compression molded into consolidated panels to obtain the hybrid composite(s) (BF-CF-PP). Four formulations with different fiber-resin weight percent were designed and produced including- BF-PP (30/70), BF-PP (50/50), BF-CF-PP (32/8/60), and BF-CF-PP (8/32/60). The effect of (a) fiber length, (b) surface treatment, (c) fiber content, and (d) consolidation pressure on the mechanical properties were examined. The improved mechanical (flexural strength 76.4 MPa, flexural modulus 4.1 GPa, ILSS 12.4 MPa and impact strength 49.9 KJ/m 2 ) and vibrational damping (1.05%) properties showed that the BF-CF-PP (8/32/60) provided higher properties compared to the other variants. The projected properties at various stoichiometric ratios of carbon and bamboo fiber revealed that the properties of hybrid composites could be tailored to produce desirable, cost-effective, and sustainable automotive components.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Optimal design of chemoepitaxial guideposts for the directed self-assembly of block copolymer systems using an inexact Newton algorithm

Directed self-assembly (DSA) of block copolymers (BCPs) is one of the most promising developments in the cost-effective production of nanoscale devices. The process makes use of the natural tendency for BCP melts to form nanoscale structures upon phase separation. The phase separation can be directed through the use of chemically patterned substrates to promote the formation of morphologies that are essential to the production of semiconductor devices. Moreover, the design of substrate pattern can be formulated as an optimization problem for which we seek optimal substrate designs that effectively produce given target morphologies. In this paper, we adopt a phase field model given by a nonlocal Cahn–Hilliard partial differential equation (PDE) based on the minimization of the Ohta–Kawasaki free energy, and present an efficient PDE-constrained optimization framework for the optimal design problem. The design variables are the locations of circular- or strip-shaped guiding posts that are used to model the substrate chemical pattern. To solve the ensuing optimization problem, we propose a variant of an inexact Newton conjugate gradient algorithm tailored to this problem. Additionally, we demonstrate the effectiveness of our computational strategy on numerical examples that span a range of target morphologies. Owing to our second-order optimizer and fast state solver, the numerical results demonstrate five orders of magnitude reduction in computational cost over previous work. The efficiency of our framework and the fast convergence of our optimization algorithm enable us to rapidly solve the optimal design problem in not only two, but also three spatial dimensions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Multiscale graph neural network autoencoders for interpretable scientific machine learning

The goal of this work is to address two limitations in autoencoder-based models: latent space interpretability and compatibility with unstructured meshes. This is accomplished here with the development of a novel graph neural network (GNN) autoencoding architecture with demonstrations on complex fluid flow applications. To address the first goal of interpretability, the GNN autoencoder achieves reduction in the number nodes in the encoding stage through an adaptive graph reduction procedure. Further, this reduction procedure essentially amounts to flowfieldconditioned node sampling and sensor identification, and produces interpretable latent graph representations tailored to the flowfield reconstruction task in the form of so-called masked fields. These masked fields allow the user to (a) visualize where in physical space a given latent graph is active, and (b) interpret the time-evolution of the latent graph connectivity in accordance with the time-evolution of unsteady flow features (e.g. recirculation zones, shear layers) in the domain. To address the goal of unstructured mesh compatibility, the autoencoding architecture utilizes a series of multi-scale message passing (MMP) layers, each of which models information exchange among node neighborhoods at various lengthscales. The MMP layer, which augments standard single-scale message passing with learnable coarsening operations, allows the decoder to more efficiently reconstruct the flowfield from the identified regions in the masked fields. Analysis of latent graphs produced by the autoencoder for various model settings are conducted using unstructured snapshot data sourced from large-eddy simulations in a backward-facing step (BFS) flow configuration with an OpenFOAM-based flow solver at high Reynolds numbers.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

RandONets: Shallow networks with random projections for learning linear and nonlinear operators

Deep neural networks have been extensively used for the solution of both the forward and the inverse problem for dynamical systems. However, their implementation necessitates optimizing a high-dimensional space of parameters and hyperparameters. This fact, along with the requirement of substantial computational resources, pose a barrier to achieving high numerical accuracy, but also interpretability. Here, to address the above challenges, we present Random Projection-based Operator Networks (RandONets): shallow networks with random projections and tailor-made numerical analysis methods that learn accurately and fast linear and nonlinear operators. Building on previous works, we prove that RandOnets are universal approximators of linear and nonlinear operators. Due to their simplicity, RandONets provide a one-step transformation of the input space, facilitating interpretability. For the evaluation of their performance, we focus on operators of PDEs. We show, that RandONets outperform by several orders of magnitude, both in terms of numerical approximation accuracy and computational cost, the “vanilla” DeepONets. Hence, we believe that our method will trigger further developments in the field of scientific machine learning, for the development of new ‘’light”schemes that will provide high accuracy while reducing dramatically the computational cost. A MATLAB toolbox for RandONets, including demos, is available on GitHub at https://github.com/GianlucaFabiani/RandONets.

Interpretable machine learning↗