Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “learning rate”

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 829 records · Page 46
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

A safe reinforcement learning algorithm for supervisory control of power plants

Traditional control theory-based methods require tailored engineering for each system and constant fine-tuning. In power plant control, one often needs to obtain a precise representation of the system dynamics and carefully design the control scheme accordingly. Model-free Reinforcement learning (RL) has emerged as a promising solution for control tasks due to its ability to learn from trial-and-error interactions with the environment. It eliminates the need for explicitly modeling the environment’s dynamics, which is potentially inaccurate. However, the direct imposition of state constraints in power plant control raises challenges for standard RL methods. To address this, we propose a chance-constrained RL algorithm based on Proximal Policy Optimization for supervisory control. Our method employs Lagrangian relaxation to convert the constrained optimization problem into an unconstrained objective, where trainable Lagrange multipliers enforce the state constraints. In conclusion, our approach achieves the smallest distance of violation and violation rate in a load-follow maneuver for an advanced Nuclear Power Plant design.

constrained optimization↗
DOE OSTI2020

Automatic Seismic Phase Picking Using Deep Learning for the EGS Collab project

Microseismic monitoring plays an important role in many energy-related and environmental industries.The microseismic event catalog and seismic structure of the subsurface are two of the primary outputsof the microseismic monitoring system. Though rough locations of microseismic events can be estimated automatically, obtaining high-resolution microseismic event locations requires a significant amount of human laborespecially on seismic phase picking. Unlike traditional automatic pickers that are usually less precise than human analysts, a fewrecently proposed algorithms based on deepneural networks(DNN)were able to match or surpass human performance for earthquake signals. Due to differences in the spatial scale of the study area, sensor sampling rate, and geometry of the monitoring system, it is not clear whether these deepneural networkmodels can be used to speed up microseismic data processing. In this paper, we adapted the DNN based technique for automatic phase picking of microseismic signals. We usedmicroseismic data recorded at the experiment 1 site of the enhancedgeothermal system (EGS) Collab project anddesigneda workflowthat we call transfer-learning aided double-difference tomography (TADT),thatcombines transfer learning and seismic tomography. We re-train an existing DNN with our data to obtain a newmodel using around 3500 seismogramsand associated manual phase picks. Thistransfer learnedmodel is able to reach human performance but muchfaster than human analysts. The transfer-learning-derived phase pickswereused to improve microseismic event locations and imagethesubsurface. The results are similar to or slightly better than those obtained with manual phase picks.

Chai, Chengping↗
NASA NTRS2017

From Prototype to Flight: Qualifying a Ka-Band Parabolic Deployable Antenna (KaPDA) for CubeSats

CubeSats have experienced a number of exciting technological advancements in the past several years. However, until recently, there has been very limited development in the area of high gain CubeSat antennas, which are critical for both high data rate communications and radar science. A Ka-band high gain antenna would provide a 10,000 times increase in data communication rates over an X-band patch antenna and a 100 times increase over state-of-the-art S-band parabolic antennas. Because of this, three years ago the Jet Propulsion Laboratory (JPL) initiated a research and technology development effort to advance CubeSat communication capabilities, with one of the key thrusts being the Ka-band parabolic deployable antenna (KaPDA). This antenna started with the ambitious goal of fitting a 42 dB, 0.5 meter, 35 Ghz antenna in a 1.5U (10 cm x 10 cm x 17 cm) canister. This paper discusses the process of taking the antenna from a first prototype to the flight design, which is flying on the RainCube mission, and earth science CubeSat. The prototype antenna was constructed in early 2015, and then upgraded to an engineering model at the end of 2016 to compensate for lessons learned. The flight version is currently under construction, and scheduled to be finished in 2016. KaPDA is the second deployable parabolic antenna to fly on a CubeSat, and the first of its kind to operate at Ka-band enabling a number of opportunities for high rate deep space antenna communications and radar science.

Sauder, Jonathan↗
DOE OSTI2023

Multimodal Characterization of Crystal Structure and Formation in Rubrene Thin Films Reveals Erasure of Orientational Discontinuities

Abstract Multimodal multiscale characterization provide opportunities to study organic semiconducting thin films with multiple length scales, across multiple platforms, to elucidate crystallization mechanisms of the various microstructures that impact functionality. With polarized scanning transmission X‐ray and 4D‐scanning transmission electron microscopy, hybrid crystalline structures in rubrene thin films in which large crystalline domains surround a common nucleus and transition to a spherulite morphology at larger radii is observed. These high‐resolution techniques reveal how azimuthal orientational discontinuities at smaller radii are erased as spherulite morphology takes hold. In situ crystallization in the films with optical microscopy is also captured, discovering the importance of considering the initial temperature increase of a film during thermal annealing over the crystallization timescale. This kinetic information of the radial crystallization rate and of corresponding film heating kinetics is used to estimate the temperature at which the larger crystalline regions transition into a spherulite. By combining the results obtained from the different characterization modes, it is learned that thermal conditions can sensitively affect the crystallization of rubrene and other organic thin films. The observations suggest opportunities for more complex temperature‐dependent processing to maximize hybrid structures’ functionality in organic thin films and demonstrate that multimodal studies deepen the understanding of structure‐function dynamics.

Tan, Jenna A.↗
DOE OSTI2022

Neural Network-Based Electric Vehicle Range Prediction for Smart Charging Optimization

Range prediction is a standard feature in most modern road vehicles, allowing drivers to make informed decisions about when to refuel. Most vehicles make range predictions through data- or model-driven means, monitoring the average fuel consumption rate or using a tuned vehicle model to predict fuel consumption. The uncertainty of future driving conditions makes the range prediction problem challenging, particularly for less pervasive battery electric vehicles (BEV). Most contemporary machine learning-based methods attempt to forecast the battery SOC discharge profile to predict vehicle range. In this work, we propose a novel approach using two recurrent neural networks (RNNs) to predict the remaining range of BEVs and the minimum charge required to safely complete a trip. Each RNN has two outputs that can be used for statistical analysis to account for uncertainties; the first loss function leads to mean and variance estimation (MVE), while the second results in bounded interval estimation (BIE). These outputs of the proposed RNNs are then used to predict the probability of a vehicle completing a given trip without charging, or if charging is needed, the remaining range and minimum charging required to finish the trip with high probability. Training data was generated using a low-order physics model to estimate vehicle energy consumption from historical drive cycle data collected from medium-duty last-mile delivery vehicles. Here, the proposed method demonstrated high accuracy in the presence of day-to-day route variability, with the root-mean-square error (RMSE) below 6% for both RNN models.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

Measuring Q₀ in LCLS-II Cryomodules Using Helium Liquid Level

The nitrogen-doped cavities used in the Linac Coherent Light Source II (LCLS-II) cryomodules have shown an unprecedented high Q₀ in vertical and cryomodule testing compared with cavities prepared with standard methods. While demonstration of high Q₀ in the test stand has been achieved, maintaining that performance in the linac is critical to the success of LCLS-II and future accelerator projects. The LCLS-II cryomodules required a novel method of measuring Q₀, due to hardware incompatibilities with existing procedures. Initially developed at Jefferson Lab during cryomodule acceptance testing before being used in the tunnel at SLAC, we use helium liquid level data to estimate the heat generated by cavities. We first establish the relationship between the rate of helium evaporation from known heat loads using electric heaters, and then use that relationship to determine heat from an RF load. Here we present the full procedure along with the development process, lessons learned, and reproducibility while demonstrating for the first time that world record Q₀ can be maintained within the real accelerator environment.

43 PARTICLE ACCELERATORS↗
DOE OSTI2021

Robust Molecular Predictive Methods for Novel Polymer Discovery and Applications

Polymeric materials are ubiquitous in modern society and they play an instrumental role in almost all industries, undoubtedly including the energy and environment sectors. Increased demand of energy and awareness to sustainability both necessitates the development of novel polymers with enhanced properties. Unfortunately, their structural and behavioral complexity render such discovery challenging and impeded. To address this problem, scientists are developing various computational modeling techniques and leveraging their power to depict the relationship between structural characteristics of polymers and their properties (such as rheological behaviors), and use such prediction to guide the design and syntheses of novel polymeric materials with enhanced performances. Unfortunately, predicting the relationships between polymer structure and composition with rheological properties via atomistic modeling is still a major challenge because of the extended time and length scales involved. Studying dynamic shear viscosity and linear viscoelasticity using molecular models requires capabilities that have been elusive, including representation of large molecular weight chains with an effective internal scale capable of describing entanglement, shear-rates that are in the s-1 scale with accurate quantitative stresses, and chemically-realistic combinations of both homogeneous and heterogeneous systems. Motivated by these unmet challenges, the overall technical objective of this DOE-STTR Phase II project is to develop robust molecular predictive methods for advanced polymer discovery and applications and especially for designing and demonstrating the “smart” polymer-based waterflooding enhanced oil recovery (EOR) process. In particular, we apply state-of-the-art molecular modeling methods developed by our academic partner, Materials Stimulation Center (MSC) at California Institute of Technology (Caltech), to facilitate and accelerate the experimental discovery processes. During the Phase I of this project, we had focused on development and demonstration of the molecular modeling methods to describe rheological properties of non-Newtonian polymer fluids, and to improve our fundamental understandings of shear-thickening mechanism and kinetics. In Phase II, we further apply the theoretical models to guide our experimental programs to improve our design of smart rheology modifier (SRM) polymers and their optimization for EOR. Specifically, we have three objectives in the Phase II study: (1) to further improve out computational modeling methods, coupling with the advanced machine learning algorithms; (2) to develop cost-effective and efficient SRM-flooding process suitable for EOR applications under typical reservoir conditions; and (3) to further explore the application of our molecular predictive models for innovative material discovery in other industrial applications. The recent development of our multiscale predictive framework allows the successful prediction of rheological properties from the chemical structure for polymers of experimentally relevant molecular weights, and provides an in-silico machine learning engine for screening novel compositions and structures with optimized non-Newtonian response, required for both shear-thinning and shear-thickening applications. Our framework provides: (1) procedures and tools for systematic coarsening from atomistic models and reverse mapping of coarse-grain models to atomistic, (2) unique ab initio methods to characterize the atomistic origin of colloidal and interfacial interactions and phenomena, (3) systematic structure and composition builders based on practical descriptors that drive rheological changes in polymer melts and diluted polymer mixtures, (4) a rheological properties engine capable of predicting viscosity in the zero-shear limit and under realistic dynamic conditions (for shear-rates commensurate with experiments) for large heterogeneous systems, (5) coarse-grain force fields with improved non-bond descriptions based on accurate quantum mechanics, (6) an in-silico screening machine learning engine that feeds from the systematic model builders to cover the descriptors search space, computes the rheological properties from converged trajectories spanning sub-milliseconds and ranks them for each structure/composition using an automated viscosity-vs-shear rate fitness function that can be tuned for shear-thickening, shear-thinning and other rheological responses.

02 PETROLEUM↗
DOE OSTI2023

Deep learning at the edge enables real-time streaming ptychographic imaging

Abstract Coherent imaging techniques provide an unparalleled multi-scale view of materials across scientific and technological fields, from structural materials to quantum devices, from integrated circuits to biological cells. Driven by the construction of brighter sources and high-rate detectors, coherent imaging methods like ptychography are poised to revolutionize nanoscale materials characterization. However, these advancements are accompanied by significant increase in data and compute needs, which precludes real-time imaging, feedback and decision-making capabilities with conventional approaches. Here, we demonstrate a workflow that leverages artificial intelligence at the edge and high-performance computing to enable real-time inversion on X-ray ptychography data streamed directly from a detector at up to 2 kHz. The proposed AI-enabled workflow eliminates the oversampling constraints, allowing low-dose imaging using orders of magnitude less data than required by traditional methods.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Adaptive 3D convolutional neural network-based reconstruction method for 3D coherent diffraction imaging

We present a novel adaptive machine-learning based approach for reconstructing three-dimensional (3D) crystals from coherent diffraction imaging. We represent the crystals using spherical harmonics (SH) and generate the corresponding synthetic diffraction patterns. We utilize 3D convolutional neural networks (CNNs) to learn a mapping between 3D diffraction volumes and the SH, which describe the boundary of the physical volumes from which they were generated. We use the 3D CNN-predicted SH coefficients as the initial guesses, which are then fine-tuned using adaptive model-independent feedback for improved accuracy. We also adaptively tune the locations, intensities, and decay rates of collections of radial basis functions in order to reproduce the non-uniform internal structure of 3D objects and demonstrate the method for a synthetic volume that has an internal void and a density ramp.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Machine learning-aided inverse design for biogas upgrading through biological CO 2 conversion

The biogas upgrading process through bioconversion of CO 2 to CH 4 by hydrogenotrophic methanogens is an attractive strategy for energy decarbonation. Many studies have optimized operational parameters to improve key performance indicators such as CH 4 % and H 2 utilization efficiency. However, inconsistent laboratory conditions make it challenging to compare results. Existing models for analyzing operating conditions can only assess the impact of individual conditions and lack the ability to simultaneously optimize multiple conditions. To address this, two XGBoost models were built with R 2 of 0.779 and 0.903 with data collected from literatures and were embedded into multi-objective partitive swarm optimization algorithm to optimal operating conditions. Predictions were compared with experimental validations under optimized conditions, revealing an 8.50% and 2.95% relative error in CH 4 % and H 2 conversion rate, respectively. This approach streamlines biogas upgrading processes, offering a data-driven solution to enhance efficiency and consistency in the pursuit of sustainable methane production.

Biogas upgrading↗
DOE OSTI2020

Hydrogen production from lignocellulosic hydrolysate in an up-scaled microbial electrolysis cell with stacked bio-electrodes

Hydrogen production from renewable resources via microbial electrolysis cells (MECs) is a promising approach for sustainable energy production. Yet high hydrogen yield from real feedstocks has not been demonstrated in up-scaled MECs. In this study, a 10-L single chamber MEC with a high electrode surface area to volume ratio (66 m 2 /m 3 ) was constructed and electroactive cathodic biofilms were enriched for hydrogen evolution reaction. A high hydrogen yield of 91% was achieved using lignocellulosic hydrolysate with a hydrogen production rate of 0.71 L/L/D at an organic loading rate of 0.4 g/D. The anodic and cathodic microbial communities, with Enterococcus spp. as the known electroactive bacteria, were capable of achieving current densities of 13.7 A/m 2 and 16.5 A/m 2 , respectively. Finally, a machine learning algorithm was used to investigate the correlation between community data and electrochemical performance, and the critical genera on determining current density were identified.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

On the integration of molecular dynamics, data science, and experiments for studying solvent effects on catalysis

Computational workflows that combine molecular dynamics (MD) simulations and emerging data-centric (DC) methods can accelerate the screening and analysis of solvent systems experimentally and computationally. Here, MD simulations provide atomic positions and velocities of reactant, solvent, and catalyst materials that can be manipulated into data representations that in turn can be used by DC techniques to conduct predictive modeling, feature extraction, and experimental design. For liquid-phase catalytic applications, emerging DC techniques such as Convolutional and Graph Neural Networks (CNN/GNN), Topological Data Analysis (TDA), and Active Learning (AL) can leverage MD and experimental data to quickly predict solvent effects on reaction outcomes. For instance, in recent studies, 3D solvent environments obtained with MD have been exploited by CNNs to predict experimental reaction rates for homogeneous acid-catalyzed lignocellulosic processes. In this perspective, we discuss basic principles of DC methods and how these can be combined with MD to enable high-throughput screening of solvent selection for diverse catalysis applications.

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

ECNet is an evolutionary context-integrated deep learning framework for protein engineering

Abstract Machine learning has been increasingly used for protein engineering. However, because the general sequence contexts they capture are not specific to the protein being engineered, the accuracy of existing machine learning algorithms is rather limited. Here, we report ECNet (evolutionary context-integrated neural network), a deep-learning algorithm that exploits evolutionary contexts to predict functional fitness for protein engineering. This algorithm integrates local evolutionary context from homologous sequences that explicitly model residue-residue epistasis for the protein of interest with the global evolutionary context that encodes rich semantic and structural features from the enormous protein sequence universe. As such, it enables accurate mapping from sequence to function and provides generalization from low-order mutants to higher-order mutants. We show that ECNet predicts the sequence-function relationship more accurately as compared to existing machine learning algorithms by using ~50 deep mutational scanning and random mutagenesis datasets. Moreover, we used ECNet to guide the engineering of TEM-1 β-lactamase and identified variants with improved ampicillin resistance with high success rates.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Fatty acid capped, metal oxo clusters as the smallest conceivable nanocrystal prototypes

Metal oxo clusters of the type M 6 O 4 (OH) 4 (OOCR) 12 (M = Zr or Hf) are valuable building blocks for materials science. Here, we synthesize a series of zirconium and hafnium oxo clusters with ligands that are typically used to stabilize oxide nanocrystals (fatty acids with long and/or branched chains). The fatty acid capped oxo clusters have a high solubility but do not crystallize, precluding traditional purification and single-crystal XRD analysis. We thus develop alternative purification strategies and we use X-ray total scattering and Pair Distribution Function (PDF) analysis as our main method to elucidate the structure of the cluster core. We identify the correct structure from a series of possible clusters (Zr3, Zr4, Zr6, Zr12, Zr10, and Zr26). Excellent refinements are only obtained when the ligands are part of the structure model. Further evidence for the cluster composition is provided by nuclear magnetic resonance (NMR), infrared spectroscopy (FTIR), thermogravimetry analysis (TGA), and mass spectrometry (MS). We find that hydrogen bonded carboxylic acid is an intrinsic part of the oxo cluster. Using our analytical tools, we elucidate the conversion from a Zr6 monomer to a Zr12 dimer (and vice versa), induced by carboxylate ligand exchange. Finally, we compare the catalytic performance of Zr12-oleate clusters with oleate capped, 5.5 nm zirconium oxide nanocrystals in the esterification of oleic acid with ethanol. The oxo clusters present a five times higher reaction rate, due to their higher surface area. Since the oxo clusters are the lower limit of downscaling oxide nanocrystals, we present them as appealing catalytic materials, and as atomically precise model systems. In addition, the lessons learned regarding PDF analysis are applicable to other areas of cluster science as well, from semiconductor and metal clusters, to polyoxometalates.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Deep Learning Accelerated History Matching and Forecasting in Geologic CO 2 Sequestration [Slides]

We have successfully developed a predictive workflow for the Geologic CO 2 Sequestration using a deep learning model based on the Fourier Neural Operator. The workflow has high accuracy for predicting pressure & saturation during the injection and post-injection periods and has decent accuracy for predicting water and CO 2 production rates during injection period. It is necessary to train exclusive DNN models for pressure prediction in long term GCS, but for saturation prediction, we can use a single model to predict it.

58 GEOSCIENCES↗
DOE OSTI2022

Application of FARM to an IES scenario within the FORCE ecosystem

The FARM (Feasible Actuator Range Modifier) software module is a component of the RAVEN-based FORCE framework for analysis of Integrated Energy Systems (IES). FARM supports the HERON software module in the evaluation of the optimal dispatch by evaluating feasible set-points for the different IES unit components. Set-points are required to satisfy limits on both production variables (i.e., the variables to be optimized such as the electrical power, the hydrogen production rate, etc.) and process variables tied to the service life of equipment (e.g., steam flowrate, vessel pressure, turbine firing temperature, etc.). This problem is addressed by adopting a two-stage approach. First, the HERON power dispatcher determines set-points that meet the constraints on the former variables (e.g., power levels and power ramp rate limits). These constraints are called explicit constraints. Then, FARM adjusts these set-points to ensure the respect of the limits on the latter variables given the knowledge of the system physics acquired through machine learning algorithms. These constraints are called implicit constraints. From this standpoint, FARM constitutes a bridge between the HERON power dispatcher that adopts a simplified description of the IES unit (low-resolution physics) and the HYBRID high-fidelity models (high-resolution physics). The original version of the FARM software module (FARM-Alpha) was released by Argonne National Laboratory in January 2021. In the latest version of the code released in July 2022 (FARM-Delta), the Reference Governor (RG) algorithm was upgraded to a Multi-Input Multi-Output version from its original Single-Input Single-Output form. The RG algorithm acts to enforce constraints. With this improvement an IES unit is now treated as a single dynamic system from the standpoint of control. The crosstalk among components in an IES unit is now fully considered thereby ensuring a true optimization is obtained for those units that have multiple set-points. In this report, the capabilities of FARM-Delta operating within the FORCE ecosystem are demonstrated for an IES test case. The specific configuration of IES unit for this case was selected by the IES team with consultation from the Advanced Reactor IES Expert Group. A full TEA analysis that invoked HERON, HYBRID, FARM, and RAVEN was performed and serves to demonstrate how the latest modification to FARM algorithms (i.e., state variable selection, state-space matrices derivation, set-point verification) can shape setpoints that might otherwise compromise the health of equipment through accelerated wear and tear. In this specific test case, it was demonstrated that these algorithms ensure a more efficient utilization of steam resources to be shared by two different subsystems, namely Balance of Plant (BOP) and High-Temperature Steam Electrolysis (HTSE). Finally, some code improvements that can further enhance the user-friendliness are suggested.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Efficiency of ML Anomaly Detection Triggers for Emerging Jets

Novel machine learning-based anomaly detection Level 1 (L1) triggers are currently under development at CMS, namely AXOL1TL and CICADA. The former employs a variational autoencoder, while the latter utilizes a convolutional autoencoder. These triggers aim to balance rate reduction with model independence, enabling the selection of potentially significant events that might be overlooked by traditional triggers relying on basic kinematic variable selections. Consequently, they have the potential to enhance signals indicative of physics beyond the Standard Model, such as those associated with emerging jets. Such signals are predicted by models featuring a composite dark sector where long-lived particles decay into Standard Model jets with displaced tracks and numerous vertices. This study evaluates the efficiency of these anomaly detection triggers in selecting events with emerging jets produced via the s-channel production of two dark quarks.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

The ParFlow Sandtank: An interactive educational tool making invisible groundwater visible

Physical aquifer models are a highly effective teaching tool for hydrology education, however they come with inherent limitations that include the high cost to purchase, the static configuration of the model materials, the time required to visualize hydrogeological phenomena, and the effort to reset and clean them over time. To address these and other limitations, we have developed an interactive computer simulation of a physical aquifer model called the ParFlow Sandtank. In this gamified interface, users run the simulation using a familiar web-app like interface with sliders and buttons while learning real hydrologic concepts. Our user interface allows participants to dive into the world of hydrology, understanding assumptions about model parameters such as hydraulic conductivity, making decisions about inputs to groundwater aquifer systems such as pumping rates, visualizing outputs such as stream flow, transport, and saturation, and exploring various factors that impact real environmental systems such as climate change. The ParFlow Sandtank has already been used in a variety of educational settings with more than 9,000 users per year, and we feel this emerging educational tool can be used broadly in educational environments and can be scaled-up to provide greater accessibility for students and educators. Here we present the capabilities and workflow of the ParFlow Sandtank, two use cases, and additional tools and custom templates that have been developed to support and enhance the reach of the ParFlow Sandtank.

54 ENVIRONMENTAL SCIENCES↗