Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “weight loss”

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 271 records · Page 15
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Describing hadronization via histories and observables for Monte-Carlo event reweighting

We introduce a novel method for extracting a fragmentation model directly from experimental data without requiring an explicit parametric form, called Histories and Observables for Monte-Carlo Event Reweighting (HOMER), consisting of three steps: the training of a classifier between simulation and data, the inference of single fragmentation weights, and the calculation of the weight for the full hadronization chain. We illustrate the use of HOMER on a simplified hadronization problem, a q\bar{q} q q ‾ string fragmenting into pions, and extract a modified Lund string fragmentation function f(z) f ( z ) . We then demonstrate the use of HOMER on three types of experimental data: (i) binned distributions of high-level observables, (ii) unbinned event-by-event distributions of these observables, and (iii) full particle cloud information. After demonstrating that f(z) f ( z ) can be extracted from data (the inverse of hadronization), we also show that, at least in this limited setup, the fidelity of the extracted f(z) f ( z ) suffers only limited loss when moving from (i) to (ii) to (iii). Public code is available at https://gitlab.com/uchep/mlhad.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2020

Isothermal oxidation of zirconium using steam

Oxidation of zirconium and its alloys under any atmosphere is generally characterised by a protective and non-protective stage. During the protective stage, the kinetics of oxidation are quasi-parabolic, while the oxide scale formed on the surface of the metal is black, compact, and protective in nature. For the non-protective stage, the oxidation rate becomes linear, while the protective oxide film turns into a deteriorated white scale. Breakaway is known as the transition point from protective to non-protective stage. The actual mechanism leading to breakaway of zirconium is still a matter of controversy, however some of the most common factors affecting breakaway are the following: - Impurities on the metal (Carbon, Oxygen, Nitrogen); - Oxidizing atmosphere (Presence of H{sub 2}O (g) or N{sub 2} (g)); - Sample preparation (Chemically etched or polished); - Temperature (LWR, LOCA). This research investigates the effect of temperature on 'breakaway' when zirconium is oxidized isothermally in steam for 60 min at 800 and 900 deg. C using thermogravimetric and optical microscopic imaging techniques. 1. Steam isothermal oxidations were performed on zirconium at 800 and 900 deg. C using a Thermogravimetric Analyzer (TGA). 2. Log-log plots of mass gain vs time were obtained from thermogravimetric results to calculate the reaction orders (n) and reaction rate constants (K) during oxidation. 3. Results of the kinetics of reaction were correlated with optical microscopic images form Light Microscope (LM) and Scanning Electron Microscope (SEM). Breakaway depends very strongly on temperature. At 800 deg. C, the weight gain followed a cubic-parabolic behaviour with evidence of a protective oxide film. On the other hand, oxidation at 900 deg. C showed signs of breakaway since the oxide film was cracked and oxidation rate was almost linear throughout the reaction.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2025

Purification and expression of a novel bacteriocin, JUQZ-1, against Pseudomonas syringae pv. Actinidiae (PSA), secreted by Brevibacillus laterosporus Wq-1, isolated from the rhizosphere soil of healthy kiwifruit

Kiwifruit canker, caused by Pseudomonas syringae pv. actinidiae (PSA), has led to significant losses in the kiwifruit industry each year. Due to the drug resistance feature of PSA, biological control is currently the most promising method. Developing biocontrol bacteria against PSA could help solve the issue of drug resistance generated during the chemical control of PSA to a certain extent. In this research, a Wq-1 strain that demonstrated excellent inhibitory activity against PSA was isolated from the rhizosphere soil of healthy kiwifruit. Based on the morphological characteristics and phylogenetic analysis of the 16S rRNA gene sequence, the isolated strain was identified as Brevibacillus laterosporus Wq-1. Bacteriostatic proteins were isolated from the cell-free culture filtrate of strain Wq-1 and were found to have a molecular weight of approximately 12 kDa, as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Liquid chromatography–tandem mass spectrometry (LC–MS/MS) detection revealed that there were several peptides in the target band that were consistent with protein 01021 in the genome. The gene of the 01021 protein was cloned into the plasmid pPICZa, and the recombinant bacteriocin was successfully expressed using the Pichia pastoris X33 expression system. The recombinant protein 01021 effectively inhibited the growth of PSA. This is the first report of the protein’s antimicrobial activity, distinguishing it from previously identified bacteriocins. Therefore, we named this bacteriocin JUQZ-1. In addition, our results showed that the protein JUQZ-1 not only exhibited a broad bacteriostatic spectrum but also high thermal and pH stability suitable for harsh environmental conditions., JUQZ-1, a protein with antimicrobial properties and strong environmental tolerance, may serve as a promising alternative to antibiotics.

Shuai, Yang↗
DOE OSTI2023

Robust scalable initialization for Bayesian variational inference with multi-modal Laplace approximations

Predictive modeling typically relies on Bayesian model calibration to provide uncertainty quantification. Variational inference utilizing fully independent (“mean-field”) Gaussian distributions are often used as approximate probability density functions. This simplification is attractive since the number of variational parameters grows only linearly with the number of unknown model parameters. However, the resulting diagonal covariance structure and unimodal behavior can be too restrictive to provide useful approximations of intractable Bayesian posteriors that exhibit highly non-Gaussian behavior, including multimodality. High-fidelity surrogate posteriors for these problems can be obtained by considering the family of Gaussian mixtures. Gaussian mixtures are capable of capturing multiple modes and approximating any distribution to an arbitrary degree of accuracy, while maintaining some analytical tractability. Unfortunately, variational inference using Gaussian mixtures with full-covariance structures suffers from a quadratic growth in variational parameters with the number of model parameters. The existence of multiple local minima due to strong nonconvex trends in the loss functions often associated with variational inference present additional complications, These challenges motivate the need for robust initialization procedures to improve the performance and computational scalability of variational inference with mixture models. In this work, we propose a method for constructing an initial Gaussian mixture model approximation that can be used to warm-start the iterative solvers for variational inference. The procedure begins with a global optimization stage in model parameter space. In this step, local gradient-based optimization, globalized through multistart, is used to determine a set of local maxima, which we take to approximate the mixture component centers. Around each mode, a local Gaussian approximation is constructed via the Laplace approximation. Finally, the mixture weights are determined through constrained least squares regression. The robustness and scalability of the proposed methodology is demonstrated through application to an ensemble of synthetic tests using high-dimensional, multimodal probability density functions. Here, the practical aspects of the approach are demonstrated with inversion problems in structural dynamics.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Multi-level, forming and filament free, bulk switching trilayer RRAM for neuromorphic computing at the edge

CMOS-RRAM integration holds great promise for low energy and high throughput neuromorphic computing. However, most RRAM technologies relying on filamentary switching suffer from variations and noise, leading to computational accuracy loss, increased energy consumption, and overhead by expensive program and verify schemes. We developed a filament-free, bulk switching RRAM technology to address these challenges. We systematically engineered a trilayer metal-oxide stack and investigated the switching characteristics of RRAM with varying thicknesses and oxygen vacancy distributions to achieve reliable bulk switching without any filament formation. We demonstrated bulk switching at megaohm regime with high current nonlinearity, up to 100 levels without compliance current. We developed a neuromorphic compute-in-memory platform and showcased edge computing by implementing a spiking neural network for an autonomous navigation/racing task. Our work addresses challenges posed by existing RRAM technologies and paves the way for neuromorphic computing at the edge under strict size, weight, and power constraints.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Enhancing Unknown Waveform Detection by Learning Intra and Inter-domain Dependencies with Advanced Attention Fusion Mechanisms

Detection of unknown waveforms in mission-critical communications is a crucial area of interest for the Department of Energy (DoE). Traditional methods and recent deep learning-based approaches often assume that the training set includes all possible classes, which is impractical for detecting new waveforms. This limitation gives rise to the problem of open-set recognition (OSR), which involves correctly identifying known classes while detecting and rejecting unknown or unseen classes. To address this limitation, we propose a novel dual-domain complex-valued neural architecture that jointly processes time-domain and frequency-domain signal representations using transformer mechanisms. A transformer model is a deep learning architecture that uses self-attention mechanisms to process and learn relationships in sequential data. Our model employs a cosine similarity loss to extract domain-specific features and incorporates a transformer architecture in the latent space to weigh the importance of different features from the time and frequency domains. The transformer layer includes stacked self-attention and cross-attention modules to learn intra-domain and inter-domain dependencies, creating a more holistic signal representation. An attention-based fusion module intelligently combines the time and frequency-domain features using multi-head attention, enabling the network to learn the optimal feature for each domain in each input signal. Quantitative results demonstrate the impact of these architectural choices on overall performance, showing significant improvement after incorporating self and cross-attention modules and using complex attention fusion over simple weighted fusion. Our ongoing work will focus on addressing the limitations of threshold-based OSR methods by developing a novel generative framework that integrates a conditional diffusion probabilistic model (DPM). DPM is a generative framework that learns to synthesize complex data by reversing a gradual noising process using a neural network trained to denoise step-by-step. Our goal is to leverage the inherent strengths of DPMs for identifying unknown signals more robustly. One primary advantage of using a DPM is its ability to provide a more reliable anomaly score based on the model's reconstruction error, rather than relying solely on classifier confidence. Additionally, the iterative denoising process of DPMs makes this approach naturally resilient to low Signal-to-Noise Ratio (SNR) conditions, where traditional methods often fail. By implementing this generative framework, we aim to enhance the model's capability to accurately detect unknown waveforms and maintain performance in challenging environments.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2024

Datasets for Widespread Residential Space Heating Electrification in Texas

In this experiment, we explore long term patterns in electricity demand driven by the dual effects of full electrification of space heating in Texas (by adoption of electric heat pumps), and climate change. We use a predictive model of electricity demand, climate projections, and an open source nodal power system (DC Optimal Power Flow) model of the Electric Reliability Council of Texas (ERCOT) system. Heat pumps are a more energy efficient way of providing space heating and cooling in homes. We attempt to exhaustively investigate the impacts of full residential space heating electrification by adoption of heat pumps for the segment of Texas households that currently rely on fossil fuels (about 40%), while simultaneously incorporating climate change meteorological variables. We explore a range of scenarios of heat pump efficiency and climate uncertainty over a long period of future years (2020-2099). In total, the simulation experiment generates 1,280 simulation years of hourly data. We report and analyze results in form of impacts on residential load, total load, peak load, seasonality of peaking, and reliability measured by occurrence and frequency of loss of load events. While the experiment is for ERCOT, the insights and approach can be applied to other regions. The results from the analysis can inform system planners on a range of potential capacity requirements/ reliability implications and/or risks of full space heating electrification via the adoption of electric heat pumps, given the uncertainty in the scenarios/ climate futures. The dataset includes model output for residential, non residential and total load, and the results from the GO ERCOT model runs for 4 RCP Scenarios (RCP 4.5 Cooler, RCP 4.5 Hotter, RCP 8.5 Cooler, RCP 8.5 Hotter), 4 Heating electrification Scenarios (Base , Standard Efficiency HP, High Efficiency HP, Ultra-High Efficiency HP) over 80 years (2020-2099). The metrological variables at BA scale were weighted weighted using population projections consistent with the SSP3 scenario.

Climate Change↗
DOE OSTI2022

Multi‐objective optimization for cost‐efficient and resilient machining under tool wear

Abstract With the onset and rapid growth of smart manufacturing, there is a constant increase in the demand for automation technologies to enhance productivity while providing uninterrupted, cost‐efficient, and resilient machining. Traditional manufacturing systems, however, suffer from several losses due to machine faults and degradation. Specifically, tool wear directly impacts the precision and quality of the milled parts, which causes an increase in the scrap production. Hence, more attempts are required to meet the desired quota of successful parts, which in turn results in wasted material, longer delays, further tool degradation, and higher energy, machining, and labor costs. As such, this paper develops a multi‐objective optimization framework to generate the optimal control set points (e.g., feed rate and width of cut) that minimize the total cost of machining operations resulting from multiple contradictory cost functions (e.g., material, energy, tardiness, machining, labor, and tool) in the presence of tool wear. Notably, we estimate the total expected cost in dollars, which provides automatic and intuitive weighting in this multi‐objective formulation. The optimization framework is tested on a high‐fidelity face milling model that has been validated on real data from industry. Results show significant dollar savings of up to as compared to the default control scheme.

Wilson, James P.↗
DOE OSTI2022

Age and chemistry of dissolved organic carbon reveal enhanced leaching of ancient labile carbon at the permafrost thaw zone

Abstract. Climate change will alter the balance between frozen and thawed conditions in Arctic systems. Increased temperatures will make the extensive northern permafrost carbon stock vulnerable to decomposition and translocation. Production, cycling, and transport of dissolved organic carbon (DOC) are crucial processes for high-latitude ecosystem carbon loss that result in considerable export off the Arctic landscape. To identify where and under what conditions permafrost DOC is mobilized in an Arctic headwater catchment, we measured radiocarbon (14C) of DOC and assessed DOC composition with ultraviolet–visible spectroscopy (UV–Vis) of surface waters and shallow and deep subsurface porewaters from 17 drainages in the Barrow Environmental Observatory in Alaska. Samples were collected in July and September 2013 to assess changes in age and chemistry of DOC over time. DOC age was highly variable ranging from modern (19 ‰ Δ14C) to approximately 7000 BP (−583 ‰ Δ14C). DOC age increased with depth, over the summer as the active layer deepened, and with increasing drainage size. DOC quality indicators reflected a DOC source rich in high molecular-weight and aromatic compounds, characteristics consistent with vegetation-derived organic matter that had undergone little microbial processing, throughout the summer and a weak relationship with DOC age. In deep porewaters, DOC age was also correlated with several biogeochemical indicators (including dissolved methane concentration, δ13C, and the apparent fractionation factor), suggesting a coupling between carbon and redox biogeochemistry influencing methane production. In the drained thawed lake basins included in this study, DOC concentrations and contributions of vegetation-derived organic matter declined with increasing basin age. The weak relationship between DOC age and chemistry and consistency in DOC chemical indicators over the summer suggest a high lability of old DOC released by thawing permafrost.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Monitoring of Liquid Metal Reactor Heater Zones with Recurrent Neural Network Learning of Temperature Time Series

Advanced high-temperature fluid reactors (ARs), such as sodium fast reactors (SFRs) and molten salt cooled reactors (MSCRs) utilize high-temperature fluids at ambient pressure. To melt the fluid during reactor startup and prevent fluid freezing during cooldown, the thermal–hydraulic systems of such ARs include heater zones consisting of specific heaters with controllers, temperature sensors, and thermal insulation. The failure of heater zones due to insulation material degradation or improper installation, resulting in parasitic heat losses, can lead to fluid freezing. The detection of faults using a heat-transfer model is difficult because of a lack of knowledge of the experimental details. Data-driven machine learning of heater zone temperature time series offers a viable alternative. In this study, we benchmarked the performance of recurrent neural networks (RNNs) in an analysis of heat-up transient temperature time series of heater zones installed on a liquid sodium vessel. The RNN models include long short-term memory (LSTM) and gated recurrent unit (GRU) networks, as well as their bi-directional variants, BiLSTM and BiGRU. Anomalous temperature points were designated using a percentile-based threshold applied to residual fluctuations in the detrended temperature time series. Additionally, the impact of the exponentially weighted moving average (EWMA) method on detection accuracy was examined. The RNN models’ performance was assessed using precision, recall, and F 1 score metrics. Results demonstrated that RNN models effectively detect anomalies in temperature time series with the best models for each heater zone achieving F 1 scores of over 93%. To explain the variations in RNN model performance across different heater zones, we used Kullback–Leibler (KL) divergence to quantify the relative entropy between training and testing data, and the Detrended Fluctuation Analysis (DFA) to assess long-range temporal correlations. For datasets with strong long-range correlations and minimal relative entropy between training and testing data, GRU is the best-performing model. When the data exhibits weaker long-term correlations and a significant relative entropy between training and testing distributions, BiGRU shows the best performance. For the data sets with intermediate values of both KL divergence and DFA, the best performance is obtained with LSTM and BiLSTM, respectively.

gated recurrent unit↗
DOE OSTI2022

gC1qR/C1qBP/HABP-1: Structural Analysis of the Trimeric Core Region, Interactions With a Novel Panel of Monoclonal Antibodies, and Their Influence on Binding to FXII

The protein gC1qR/C1qBP/HABP-1 plays an essential role in mitochondrial biogenesis, but becomes localized at the cellular surface in numerous pathophysiological states. When this occurs on endothelial cells, surface-exposed gC1qR activates the classical pathway of complement. It also promotes assembly of a multi-protein complex comprised of coagulation factor XII (FXII), pre-kallikrein (PK), and high-molecular weight kininogen (HMWK) that activates the contact system and the kinin-generating system. Since surface-exposed gC1qR triggers intravascular inflammatory pathways, there is interest in identifying molecules that block gC1qR function. Here we further that objective by reporting the outcome of a structure/function investigation of gC1qR, its interactions with FXII, and the impact of a panel of monoclonal anti-gC1qR antibodies on FXII binding to gC1qR. Although deletion mutants have been used extensively to assess gC1qR function, none of these proteins have been characterized structurally. To that end, we determined a 2.2 Å resolution crystal structure of a gC1qR mutant lacking both of its acidic loops, but which retained nanomolar-affinity binding to FXII and FXIIa. This structure revealed that the trimeric gC1qR assembly was maintained despite loss of roughly thirty residues. Characterization of a novel panel of anti-gC1qR monoclonal antibodies identified several with biochemical properties distinct from previously described antibodies, as well as one which bound to the first acidic loop of gC1qR. Intriguingly, we found that each of these antibodies could partly inhibit binding of FXII and FXIIa to gC1qR. Based on these results and previously published studies, we offer new perspectives for developing gC1qR inhibitors.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Enzymatic depolymerization of polyester: Foaming as a pretreatment to increase specific surface area

Abstract Poly(ethylene terephthalate) (PET) is widely used for its high strength‐to‐weight ratio, gas barrier properties, and chemical resistance. The growing PET use highlights the demand for a better recycling system. Enzymatic recycling, alongside mechanical and chemical methods, is eco‐friendly and yields properties similar to virgin PET. Substrate properties ( T g , crystallinity, and specific surface area [SSA]) and enzyme stability significantly impact conversion efficiency. Higher SSA and lower crystallinity tend to yield improved depolymerization when employing leaf compost‐cutinase (LCC‐ICCG) enzymes. This study explored melt extrusion and foaming as pretreatment techniques to modify PET structural properties, using a low‐cost chemical foaming agent (CFA). The monomer conversion rate and efficiency during depolymerization were measured and related to the processing, extrudate micro‐ and meso‐structure, and polyester type. Pretreated PET substrates showed reduced T g , crystallinity, density, and enhanced SSA, resulting in a 90% mass loss for foamed RPET and VPET substrates within 2 days. In contrast, PET with ~30% of cyclohexanedimethanol comonomer exhibited a nearly 50% lower depolymerization rate, with zero BHET production. It indicates that the combination of low crystallinity, low T g , and high SSA leads to improved monomer conversion. These findings emphasize the significance of amorphization and foaming in enhancing PET enzymatic depolymerization.

42 ENGINEERING↗
DOE OSTI2023

Removal of Iron from Pyrite-Rich Coal Refuse by Calcination and Magnetic Separation for Hydrometallurgical Extraction of Rare Earth Elements

In the metallurgical extraction of rare earth elements (REEs), the ratio of contaminant ions to REEs in the leachate dictates the cost and operational efficiency of the downstream processes. The current study investigated the potential iron contamination removal from the feed to the hydrometallurgical process by calcination followed by magnetic separation. The 2.20 specific gravity sink fraction of Baker coal seam coarse refuse was pulverized to finer than 180 μm, calcined at various temperatures, and separated into magnetic and non-magnetic fractions using a wet high-intensity magnetic separator at different field strengths. The untreated feed, calcined products, and their subsequent magnetic and non-magnetic fractions were subjected to acid leaching tests with 1.2 M sulfuric acid at 75 °C and 1% w/v solids concentration. The recovery of light and heavy rare earth elements (LREEs and HREEs, respectively) along with the concentration of common contaminant ions (Al, Ca, and Fe) were measured as output variables. The weight percent of magnetic material was maximized at approximately 29% by calcination at a temperature of 400 °C. Magnetic removal of this fraction using a field strength of 1.15 Tesla resulted in the rejection of 81% of the iron. Leaching of the magnetic fraction provided significantly higher Fe recovery relative to untreated feed material and the non-magnetic fraction. The non-magnetic fraction was subsequently calcined at 600 °C to dehydroxylate the clays and released the REE minerals in the same manner as the treatment of the original coarse refuse material. A comparison of the leachate elemental concentrations resulting from the leaching of both the calcined non-magnetic and original coarse refuse showed only a slight reduction in the iron content from the non-magnetic material. This finding combined with the REE loss in the magnetic fraction resulted in the conclusion that the magnetic removal step was unfavorable.

58 GEOSCIENCES↗
DOE OSTI2020

The Catalytic Reduction of Dinitrogen Under Mild Conditions (Final Report)

In order to use molecular hydrogen as an alternative to fossil fuels, large quantities of hydrogen (or protons and electrons) would have to be generated from water using a method based on solar energy and stored as a molecule in which hydrogen is a high weight percentage. The hydrogen must then be generated as needed from the form in which it is stored. An attractive molecule for this purpose is ammonia (~17% hydrogen by weight). Among the many problems that must be addressed for a successful implementation of this approach in the long term is the reduction of nitrogen catalytically to ammonia with protons and electrons under mild conditions, a goal that is addressed in this proposal. The first abiological catalytic reduction of N 2 to NH 3 by protons and electrons under mild conditions (22 °C and 1 atm) employed molybdenum catalysts, [(HIPTNCH 2 CH 2 ) 3 N]Mo (HIPTN 3 N 3- ), where HIPT (hexaisopropylterphenyl) is 3,5- (2,4,6-i-Pr 3 C 6 H 2 ) 2 C 6 H 3 . A total of approximately four turnovers (to give a maximum of eight equivalents of ammonia) along with approximately one equivalent of hydrogen per N 2 reduced, were found. A limitation of this catalytic approach is loss of the ligand from the metal. This project is aimed at developing a complex that contains a ligand that will not be removed from the metal. A macrocyclic TREN-based ligand of this type has been synthesized recently. The project is aimed at making similar ligands and molybdenum complexes that contain them and testing the new catalysts for catalytic reduction of dinitrogen. The results of these investigations will help guide further design of ligands for more efficient reduction of nitrogen to ammonia and ultimately an integrated device in which protons and electrons are generated using solar energy and employed to reduce nitrogen to ammonia using molybdenum catalysts.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Solar Field Layout and Aimpoint Strategy Optimization

The existing methods that determine heliostat aiming strategies for concentrating solar power (CSP) central receiver plants typically use heuristics and/or are computationally expensive, and they lack flexibility for different desired flux profiles and receiver geometries. Because of the interaction between layout and aimpoint strategy, considering the former without accounting for the latter may yield solutions with superfluous heliostats that cannot be used efficiently without compromising receiver flux constraints. To that end, we develop a software decision tool that uses innovative optimization methods to both optimize aimpoint strategies and improve candidate layouts for the solar collection field of a CSP central receiver plant. A CSP plant’s effectiveness relies on the optical efficiency of the solar field, which may be limited by losses due to (i) the cosine effect, (ii) atmospheric attenuation, (iii) interference (i.e., shading and blocking) between heliostats, (iv) spillage as a result of heliostat positioning and geometry, and (iv) some heliostats’ inability to direct irradiance to the receiver without damage due to excessive thermal flux. The goal of this work is to obtain optimized aiming strategies and improved solar field layouts that reduce capital cost and increase field optical efficiency and utilization, while meeting the power requirements of a given CSP receiver design. We formulate the aimpoint optimization problem as a mixed-integer linear programming model, which we then decompose into submodels that we solve in parallel. The decomposition subdivides the solar field into sections, and aimpoint strategies for each section are obtained independently of the others. To improve existing layouts, we develop a utilization-weighted efficiency metric that we use to relocate heliostats to sections of the solar field with similar efficiency and higher utilization. Finally, to connect our software to high-fidelity flux models, we develop a Python application programming interface for SolarPILOT, a mature software package that characterizes solar field performance and generates the heliostat layouts and flux maps that serve as input to our models.

14 SOLAR ENERGY↗
DOE OSTI2022

Molecular weight growth by the phenyl + cyclopentadienyl reaction: Well-skipping, ring-opening, and dissociation

Radical-radical reactions contribute to molecular weight growth of polycyclic aromatic hydrocarbons (PAHs), but their role is not well understood. The phenyl (Ph, C 6 H 5 ) + cyclopentadienyl (C 5 H 5 ) reaction has not been studied but can serve as a surrogate reaction for many reactions of larger aryl-σ radicals with resonance-stabilized π-radicals. The adducts of these reactions easily lose an H atom to produce a new π-radical. Here, through a combination of dilute flash pyrolysis experiments, theory, and simulation, we find that for the title reaction, substantial Ph—C 5 H 4 + H is formed, dominantly by a well-skipping pathway, at conditions around 30 Torr and 1400 K. This radical is thermally stable at moderate temperatures but decomposes at higher temperatures following ring opening of the 5-member ring. Products corresponding to the loss of H, C 2 H 2 , and C 3 H 3 are observed experimentally. Calculation of the C 11 H 9 potential energy surface affirms that these are expected decomposition products and reveals multiple possible pathways.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Optimizing the optimizer for physics-informed neural networks and Kolmogorov-Arnold networks

Physics-Informed Neural Networks (PINNs) have revolutionized the computation of PDE solutions by integrating partial differential equations (PDEs) into the neural network’s training process as soft constraints, becoming an important component of the scientific machine learning (SciML) ecosystem. More recently, physics-informed Kolmogorv-Arnold networks (PIKANs) have also shown to be effective and comparable in accuracy with PINNs. In their current implementation, both PINNs and PIKANs are mainly optimized using first-order methods like Adam, as well as quasi-Newton methods such as BFGS and its low-memory variant, L-BFGS. However, these optimizers often struggle with highly nonlinear and non-convex loss landscapes, leading to challenges such as slow convergence, local minima entrapment, and (non)degenerate saddle points. In this study, we investigate the performance of Self- Scaled BFGS (SSBFGS), Self-Scaled Broyden (SSBroyden) methods and other advanced quasi-Newton schemes, including BFGS and L-BFGS with different line search strategies. These methods dynamically rescale updates based on historical gradient information, thus enhancing training efficiency and accuracy. We systematically compare these optimizers – using both PINNs and PIKANs – on key challenging PDEs, including the Burgers, Allen-Cahn, Kuramoto-Sivashinsky, Ginzburg-Landau, and Stokes equations. Additionally, we evaluate the performance of SSBFGS and SSBroyden for Deep Operator Network (DeepONet) architectures, demonstrating their effectiveness for data-driven operator learning. Our findings provide state-of-the-art results with orders-of-magnitude accuracy improvements without the use of adaptive weights or any other enhancements typically employed in PINNs. More broadly, our work reveal insights into the effectiveness of quasi-Newton optimization strategies in significantly improving the convergence and accurate generalization of PINNs and PIKANs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

The Sensitivity of Variational Bayesian Neural Network Performance to Hyperparameters

In scientific applications, predictive modeling is often of limited use without accurate uncertainty quantification (UQ) to indicate when a model may be extrapolating or when more data needs to be collected. Bayesian Neural Networks (BNNs) produce predictive uncertainty by propagating uncertainty in neural network (NN) weights and offer the promise of obtaining not only an accurate predictive model but also accurate UQ. However, in practice, obtaining accurate UQ with BNNs is difficult due in part to the approximations used for model training (such as those made in variational inference) and in part to the need to choose a suitable set of hyperparameters; these hyperparameters outnumber those needed for traditional NNs and often have opaque effects on the results. We aim to shed light on the effects of hyperparameter choices for variational BNNs by performing a global sensitivity analysis of variational BNN performance under varying hyperparameter settings. Our results indicate that many of the hyperparameters interact with each other to affect both predictive accuracy and UQ. For improved usage of variational BNNs in real-world applications, we suggest that thorough hyperparameter tuning, including tuning of prior hyperparameters and loss function parameters, is essential for accurate UQ in variational BNNs.

97 MATHEMATICS AND COMPUTING↗