Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Domain reduction method”

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 55 records · Page 3
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

NEAMS Milestone Report: M2MS–20OR030102 FW–CADIS PWR Ex-Core Analysis with Shift Through VERA

This report presents the work completed for the NEAMS milestone M2MS–20OR030102 titled "FW–CADIS PWR Ex-Core Analysis with Shift through VERA." The work completed for this milestone includes the implementation, integration, and optimization of memory and performance improvement methods in Shift for fully coupled ex-core calculations through Virtual Environment for Reactor Applications (VERA). Fully coupled in this context means the transfer of moderator boron concentration, pin-wise fission source, depleted compositions, temperatures, and moderator densities from MPACT (with COBRA-TF (CTF)) to Shift. The ability to run ex-core calculations with VERA has been enabled and used for several years by Consortium for Advanced Simulation of Light Water Reactors (CASL) partners. However, this implementation was limited and potentially computationally burdensome. This work has enabled the ability to run higher-fidelity ex-core calculations on moderate computing clusters by focusing on multithreading, domain decomposition, and Forward-Weighted CADIS (FW-CADIS) variance reduction. Tests performed on small cores, a small modular reactor (SMR), and CASL progression problems show very promising memory reduction and computational performance. Recommendations for settings when running fully coupled high-fidelity ex-core calculations with VERA are documented. Without these optimization methods, many processors on a compute node would be left unused for the entire ex-core calculation. Therefore, these methods enable the user to better use the resources available and reduce computation time.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

Numerical and Analytical Modeling of the Effect of Cracks on the Self-Inductance of a COTS YJ-41003-TC Toroid

COTS inductors and transformers often contain partial cracks whose effect on inductance, a key performance parameter, have not been carefully studied. In this report, the effects of both partial and complete cracks on the self-inductance of a 100 turn square cross section COTS YJ-41003-TC toroid comprised of J Material was comprehensively investigated using both analytically derived closed form expressions and 3D computational techniques employing commercial codes. Both partial (half-penny) and complete (air gap) cracks of 10 and 25 μm were investigated. The crack is defined as the physical distance between two faces of the toroid's magnetic core, such that the surface normal of either face is along the Φ-direction, in alignment with the B-field. For the purposes of validation, two different approaches were incorporated for both the analytical and numerical models. The two analytical methods are comprised of a first principles approach based on the physics of electromagnetics, as well as linear circuit theory. The former directly utilizes the integral form of Maxwell's equations while the latter exploits the interchangeable relationship between electric and magnetic circuits. Validation within the computational scheme is realized through a code-to-code comparison between commercial solvers, COMSOL Multiphysics and CST, with the former employing the Finite Element Method (FEM) and the latter the Finite Difference Time Domain (FDTD) technique. Sound agreement between all four methods (ie., two analytical and two numerical) is observed, with results indicating that only a perturbation in self-inductance occurs for the half-penny cracks, while a substantial reduction takes place for the case of complete cracks. It is important to note that even though a static μ r is applied, representing the linear region of the BH curve (based on manufacturer specifications), the complete crack results still place a lower conservative bound on the inductance. This follows from the fact that even in the case of a half-penny crack, if the magnetic core portion of the crack approaches saturation, the crack begins to behave like an air gap, or complete crack. When an air gap is introduced into a magnetic core, a substantial reduction in inductance can occur due to the significant difference in permeabilities between the two mediums (ie., μ core >> μ air ). The once intact bulk magnetic core of the toroid essentially begins to behave like an air core.

36 MATERIALS SCIENCE↗
DOE OSTI2020

A low-rank solver for the stochastic unsteady Navier–Stokes problem

Here we study a low-rank iterative solver for the unsteady Navier–Stokes equations for incompressible flows with a stochastic viscosity. The equations are discretized using the stochastic Galerkin method, and we consider an all-at-once formulation where the algebraic systems at all the time steps are collected and solved simultaneously. The problem is linearized with Picard’s method. To efficiently solve the linear systems at each step, we use low-rank tensor representations within the Krylov subspace method, which leads to significant reductions in storage requirements and computational costs. Combined with effective mean-based preconditioners and the idea of inexact solve, we show that only a small number of linear iterations are needed at each Picard step. The proposed algorithm is tested with a model of flow in a two-dimensional symmetric step domain with different settings to demonstrate the computational efficiency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Carbonate Management to Enable Energy- and Carbon-Efficient CO 2 Electrolysis (Final Technical Report)

The rapid growth and plummeting cost of solar energy have spurred growing interest in using CO 2 electrolysis to produce chemicals and fuels as an alternative to conventional petrochemical processes. High-temperature (>800 °C) solid oxide electrolyzers that convert CO 2 into CO and O 2 have recently become commercially available. Low-temperature electrolysis cells offer the prospect of more convenient and flexible operation, which is critical for utilizing intermittent solar energy, and provide access to more valuable C 2+ products such as ethylene, ethanol, and propanol. Over the past 10 years, research in this area has yielded substantial progress in both fundamental understanding of the requisite electrocatalytic reactions and design of prototype devices. Leveraging insights from fuel cells and membrane water electrolyzers, researchers have developed electrolysis cells with gas diffusion electrodes (GDE) that have demonstrated high CO 2 electrolysis current densities (>100 mA cm –2 ) as well as promising selectivity and stability. Despite these advances, the energy efficiency (electrical energy-to-product) and carbon efficiency (CO 2 -to-product) of low-temperature CO 2 electrolysis remain far too low for large-scale deployment. A preponderance of evidence indicates that the principal source of efficiency losses is the rapid and thermodynamically favorable reaction of CO 2 with hydroxide (OH – ) to form carbonate (CO 3 2– ). Carbonate formation imposes steady-state electrolysis conditions that result in large voltage and CO 2 losses for all known (photo)electrochemical CO 2 cells. While much current research remains focused on CO 2 reduction catalyst design, this largely overlooked CO 3 2– problem presents a fundamental scientific barrier to creating a viable electrochemical option for converting solar energy into chemicals and fuels. The project pursues an integrated, multi-PI research effort that establishes a fundamental science of CO 3 2– management. PI Kanan and Co-PI Mani’s contribution to the project is to evaluate strategies to mitigate the CO 3 2– problem by changing the properties of the electrolyte and the environment in which CO 2 reduction catalysis takes place. Experimental studies showed that electrolytes composed of a high concentration of both CO 3 2– and HCO 3 – , which serve as moderately alkaline buffers, improved the cell voltage by compared to all-HCO 3 – electrolytes, but these buffered systems still show substantial CO 2 uptake that reduces pH over time. Computational studies developed a homogenized model of a CO 2 reduction catalyst layer that permits a low-cost exploration of the high-dimensional parameter space associated with catalyst layers on gas diffusion electrodes. The model was validated by accurately reproducing experimental data for the related but simpler reaction of CO reduction and then used to probe the effects of catalyst layer architecture on CO 2 reduction. Minimizing the size of catalyst and hydrophobic domains in the catalyst layer is predicted to mitigate CO 3 2– formation and thereby enable prolonged operation at elevated pH. In support of the CO 2 electrolysis studies, a new method for rapidly prototyping electrochemical cells was developed and validated. The method uses a combination of 3D printing and electroless plating to generate conductive cell components for evaluating new cell designs. The carbonate problem encompasses mass transport processes and acid-base reactions that are relevant to many other electrochemical systems. Investigation of strategies to address the carbonate problem led to an additional line of inquiry into the physicochemical phenomena that determine the efficiency of electrochemical acid-base production, which has numerous applications in the broader field of carbon management. New strategies for using the supporting electrolyte to inhibit H + /OH – recombination in electrochemical acid-base production were evaluated, leading to the development of a novel acid-base producing system that eliminates the need for ion exchange membranes and exhibits promising efficiency and current densities for scalable applications.

25 ENERGY STORAGE↗
DOE OSTI2023

Deep learning of electrochemical CO 2 conversion literature reveals research trends and directions

Large-scale and openly available material science databases are mainly composed of computer simulation results rather than experimental data. Some examples include the Materials Project, Open Quantum Materials Database, and Open Catalyst 2022. Unfortunately, building large-scale experimental databases remains challenging due to the difficulties in consolidating locally distributed datasets. In this work, focusing on the catalysis literature of CO 2 reduction reactions (CO 2 RRs), we present a machine learning (ML)-based protocol for selecting highly relevant papers and extracting important experimental data. First, we report a document embedding method (Doc2Vec) for collecting papers of greatest relevance to the specific target domain, which yielded 3154 CO 2 RR-related papers from six publishers. Next, we developed named entity recognition (NER) models to extract twelve entities related to material names (catalyst, electrolyte, etc.) and catalytic performance (Faradaic efficiency, current density, etc.). Further, among several tested models, the MatBERT-based approach achieved the highest accuracy, with an average F1-score of 90.4% and an F1-score of 95.2% in a boundary relaxation evaluation scheme. The accurate and accelerated NER-based data extraction from a large volume of catalysis literature enables temporal trend analyses of the CO 2 RR catalysts, products, and performances, revealing the potentially effective material space in CO 2 RRs. While this work demonstrates the effectiveness of our ML-based text mining methods for specifically CO 2 RR literature, the methods and approach are applicable to and may be used to accelerate the development of other catalytic chemical reactions.

36 MATERIALS SCIENCE↗
DOE OSTI2023

A semi-Lagrangian method for detecting and tracking deep convective clouds in geostationary satellite observations

Automated methods for the detection and tracking of deep convective clouds in geostationary satellite imagery have a vital role in both the forecasting of severe storms and research into their behaviour. Studying the interactions and feedbacks between multiple deep convective clouds (DCC), however, poses a challenge for existing algorithms due to the necessary compromise between false detection and missed detection errors. We utilise an optical flow method to determine the motion of deep convective clouds in GOES-16 ABI imagery in order to construct a semi-Lagrangian framework for the motion of the cloud field, independently of the detection and tracking of cloud objects. The semi-Lagrangian framework allows severe storms to be simultaneously detected and tracked in both spatial and temporal dimensions. For the purpose of this framework we have developed a novel Lagrangian convolution method and a number of novel implementations of morphological image operations that account for the motion of observed objects. These novel methods allow the accurate extension of computer vision techniques to the temporal domain for moving objects such as DCCs. By combining this framework with existing methods for detecting DCCs (including detection of growing cores through cloud top cooling and detection of anvil clouds using brightness temperature), we show that the novel framework enables reductions in errors due to both false and missed detections compared to any of the individual methods, reducing the need to compromise when compared with existing frameworks. The novel framework enables the continuous tracking of anvil clouds associated with detected deep convection after convective activity has stopped, enabling the study of the entire life cycle of DCCs and their associated anvils. Furthermore, we expect this framework to be applicable to a wide range of cases including the detection and tracking of low-level clouds and other atmospheric phenomena. In addition, this framework may be used to combine observations from multiple sources, including satellite observations, weather radar and reanalysis model data.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

A Cross-Domain Optimization Framework of PMU and Communication Placement for Multidomain Resiliency and Cost Reduction

Phasor measurement units (PMUs) play a crucial role in real-time monitoring and control of power grids. They rely on a communication network to transfer measurement data to the phasor data concentrator (PDC) for further processing and analysis. In this paper, a resilient cross-domain PMU and communication link placement method for minimizing the overall installation cost of the wide-area measurement system (WAMS) is proposed. Here, the main idea is to break down the barrier between the power grid domain and the communication domain, and consider the impact of one when design the other. The PMU placement in the power grid domain takes into account the cost of communication links by generating multiple solutions with equally minimum PMU costs for communication link placement evaluation. On the other hand, the communication link placement problem reduces the cost by customizing the routing policies based on the different roles of PMUs in grid observability. The proposed WAMS design is capable of withstanding any single component failure in the power domain (PMU failure or power branch failure) or in the communication domain (communication link failure or PDC failure). Numerical study on the IEEE 57-bus system reveals that the developed cross-domain optimization framework can significantly reduce the overall installation cost of WAMS while attaining multi-domain resiliency.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2022

Parametric model-order reduction for radiation transport using multi-resolution proper orthogonal decomposition

For parametric high-fidelity simulations, it is often desirable to utilize a reduced-order model (ROM) to emulate, at a reduced computational cost, parametric solutions of the governing partial differential equations (PDEs) for unseen parameter values. One commonly employed option is to utilize a data-driven, projection-based ROM supplemented with subspace identification via proper orthogonal decomposition (POD). POD discovers the ROM subspace by computing the singular value decomposition (SVD) of a set of training data from the full-order model (FOM). In streaming-dominated radiation transport simulations with localized sources, solutions often greatly vary over the spatial domain by many orders of magnitude. In such cases, machine-precision arithmetic can be insufficient to obtain an accurate SVD, resulting in a poorly performing ROM. We present a method called multiresolution POD (mrPOD) that mitigates these inaccuracies. The mrPOD method works by decomposing the spatial domain into regions and performing proper orthogonal decomposition on the training dataset separately in each region. In conclusion, mrPOD is tested on single energy group and multigroup atmospheric shielding transport problems and is shown to outperform classic POD.

42 ENGINEERING↗
DOE OSTI2026

Navigating Large Chemical Spaces Using Graph Theory and Integer Programming

Navigating and analyzing large chemical spaces are necessary to accelerate the design and discovery of new molecules and chemical processes. In this work, we introduce a computational framework that integrates graph theory and integer programming to enable the efficient navigation of large chemical spaces. Our framework represents the chemical space as a graph, wherein nodes represent molecules and edges represent the degree of similarity or connectivity based on domain-specific information. Using the graph representation, we identify representative molecules by computing the so-called minimum dominating set (MDS), which in our context is the minimum set of molecules that is connected to all other molecules. We present a suite of solution strategies for the MDS problem including heuristic and rigorous integer programming (IP) approaches. We show that these approaches allow us to capture physicochemical properties and domain-specific logic and constraints, facilitating the identification of molecules with the target properties. We demonstrate the effectiveness of the proposed approach by navigating the chemical space of per- and polyfluoroalkyl substances (PFAS); this comprises approximately 15,000 molecular structures. We compare our framework against traditional dimensionality reduction and clustering methods such as t-SNE and K-means clustering.

Chemical structure↗
DOE OSTI2024

Enhancing the Conductivity and Thermoelectric Performance of Semicrystalline Conducting Polymers through Controlled Tie Chain Incorporation

Abstract Conjugated polymers are promising materials for thermoelectric applications, however, at present few effective and well‐understood strategies exist to further advance their thermoelectric performance. Here a new model system is reported for a better understanding of the key factors governing their thermoelectric properties: aligned, ribbon‐phase poly[2,5‐bis(3‐dodecylthiophen‐2‐yl)thieno[3,2‐b]thiophene] (PBTTT) doped by ion‐exchange doping. Using a range of microstructural and spectroscopic methods, the effect of controlled incorporation of tie‐chains between the crystalline domains is studied through blending of high and low molecular weight chains. The tie chains provide efficient transport pathways between crystalline domains and lead to significantly enhanced electrical conductivity of 4810 S cm −1 , which is not accompanied by a reduction in Seebeck coefficient or a large increase in thermal conductivity. Respectable power factors of 173 µW m −1 K −2 are demonstrated in this model system. The approach is generally applicable to a wide range of semicrystalline conjugated polymers and could provide an effective pathway for further enhancing their thermoelectric properties and overcome traditional trade‐offs in optimization of thermoelectric performance.

Chemistry↗
DOE OSTI2020

Clustering Analysis of Commercial Vehicles Using Automatically Extracted Features from Time Series Data

Standard of practice approaches to time series cluster analysis involve careful feature engineering, often utilizing expert input to tune and select features by hand. In many cases, expert input may not be readily available, or there may not yet exist a community consensus on the ideal features for a given application. This paper compares the results of several cluster analysis methods, using both hand selected features and those extracted automatically, when applied to large geospatial time series telematics data from commercial trucking fleets. The impacts of feature selection, dimensionality reduction, and choice of clustering algorithm on the quality of clustering results are explored. Results from this analysis confirm prior results that domain agnostic features are competitive with the hand engineered features with respect to clustering quality metrics. These results also provide new insight into the most successful strategies for identifying structure in large unstructured vehicle telematics data, and suggest that time series clustering using automatic feature extraction can be an effective approach to extract structure from large scale geospatial time series data in cases when hand selected features are not available.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2022

The Schwarz Alternating Method for the Seamless Coupling of Nonlinear Reduced Order Models and Full Order Models

Projection-based model order reduction allows for the parsimonious representation of full order models (FOMs), typically obtained through the discretization of a set of partial differential equations (PDEs) using conventional techniques (e.g., finite element, finite volume, finite difference methods) where the discretization may contain a very large number of degrees of freedom. As a result of this more compact representation, the resulting projection-based reduced order models (ROMs) can achieve considerable computational speedups, which are especially useful in real-time or multi-query analyses. One known deficiency of projection-based ROMs is that they can suffer from a lack of robustness, stability and accuracy, especially in the predictive regime, which ultimately limits their useful application. Another research gap that has prevented the widespread adoption of ROMs within the modeling and simulation community is the lack of theoretical and algorithmic foundations necessary for the “plug-and-play” integration of these models into existing multi-scale and multi-physics frameworks. This paper describes a new methodology that has the potential to address both of the aforementioned deficiencies by coupling projection-based ROMs with each other as well as with conventional FOMs by means of the Schwarz alternating method [41]. Leveraging recent work that adapted the Schwarz alternating method to enable consistent and concurrent multiscale coupling of finite element FOMs in solid mechanics [35, 36], we present a new extension of the Schwarz framework that enables FOM-ROM and ROM-ROM coupling, following a domain decomposition of the physical geometry on which a PDE is posed. In order to maintain efficiency and achieve computation speed-ups, we employ hyper-reduction via the Energy-Conserving Sampling and Weighting (ECSW) approach [13]. We evaluate the proposed coupling approach in the reproductive as well as in the predictive regime on a canonical test case that involves the dynamic propagation of a traveling wave in a nonlinear hyper-elastic material.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Operando probing dynamic migration of copper carbonyl during electrocatalytic CO2 reduction

Single crystals and shape-controlled nanocrystals are well known to exhibit facet-dependent catalytic properties. However, few studies have investigated how those nanocrystals evolve and (de)activate during reactions, calling for the development of nanoscale time-resolved operando methods. In this context, we have designed Cu nanocubes as a model system to elucidate the underlying driving force of dynamic nanocatalyst reconstruction during the CO2 reduction reaction (CO2RR). Operando electrochemical liquid-cell scanning transmission electron microscopy (EC-STEM) and synchrotron-based X-ray spectroscopy reveal the size- and potential-dependent complete transformation from (100)-oriented Cu@Cu2O nanocubes to polycrystalline metallic Cu nanograins under CO2RR conditions. In addition, machine learning-assisted operando four-dimensional STEM reveals that large Cu nanograins derived from nanocubes form mainly crystalline domains, while their smaller counterparts are more amorphous due to faster evolution kinetics. In situ Raman spectroscopy and density functional theory calculations suggest that CO drives the ejection of single Cu atoms, resulting in few-nanometre Cu clusters and the surface migration of highly mobile copper carbonyl (Cu–CO) species. Combined, these multimodal operando methods and theoretical approaches pave the way for understanding the complex structural evolution of energy-related nanocatalysts under electrochemical conditions.

Yang, Yao↗
DOE OSTI2023

Adjoint Waveform Tomography for Crustal and Upper Mantle Structure the Middle East and Southwest Asia for Improved Waveform Simulations Using Openly Available Broadband Data

We present a new model of radially anisotropic seismic wavespeeds for the crust and upper mantle of a broad region of the Middle East and Southwest Asia (MESWA) derived from adjoint waveform tomography. We inverted waveforms from 192 Global Centroid Moment Tensor earthquakes (MW 5.5-7.0) recorded by over 1000 openly available broadband seismic stations from permanent and temporary networks in the region. Spatial coverage of the available data is highly uneven due to earthquakes clustered along plate boundaries and sparse coverage of open seismic networks in the region. We considered three possible starting models: the SPiRaL global model (Simmons et al., 2021); MEC-1 (Kaviani et al., 2020); and CSEM2.0 (Noe et al., 2023). Because the SPiRaL model provides good fits to the observed waveforms measured by the time-bandwidth product of selected windows in several period bands, provides all the necessary parameters and covers the entire domain we used it for the starting model with the period band 50-100 seconds. Inversion iterations proceeded using time-frequency phase misfits in six stages and 54 total iterations reducing the minimum period to 30 seconds. Our final model, MESWA, provides improved waveform fits compared to the starting model for both the data used in the inversion and an independent validation data set of 66 events. Two metrics of waveform fit (the time-frequency phase misfit used in the optimization and normalized L2 misfit) were both reduced by nearly 60% for both data sets and MESWA provides significantly larger misfit reductions relative to the SPiRaL model than the MEC-1 or CSEM models. We also find that MESWA provides a larger time-bandwidth product of selected windows indicating that more information content of the observed waveforms is explained by MESWA than the other models. Our new model reveals tectonic features imaged by other studies and methods but in a new holistic model of shear and compressional wavespeeds (v S and v P , respectively) with anisotropy covering the crust and uppermost mantle of a larger domain. MESWA has smaller scale-length features and tends to sharpen some features relative to the SPiRaL starting model. Examples include: low crustal v S in the TurkishIranian Plateau, Zagros Mountains, Afghan Central Blocks and Sulaiman Fold Belt; low mantle vSfollowing divergent (Gulf of Aden, Red Sea) and transform (Dead Sea Fault) margins of the Arabian Plate; low and high v S in the mantle beneath the Arabian Shield and Platform, respectively. Low vS is imaged below Cenozoic volcanic centers of the Arabian Peninsula, the so-called Mecca-Madina-Nafud (MMN) Line. Positive anisotropy (v SH > v SV ) is inferred for asthenospheric depths across the region except where up/downwelling may influence fabric alignment (e.g. Afar, Red Sea, Arabian Shield). Elevated vS tracks Makran subduction under southeast Iran. MESWA resembles the SPiRaL model in its long-wavelength structure, but enhances shorter wavelengths features on the order of 200 km and smaller. The resulting model could be used for as a starting model for further improvements, say using waveforms from in-country seismic networks that are not openly available or smaller-scale studies targeting shorter period waveforms. The model also could be used for source characterization and moment tensor inversion to improve earthquake hazard studies and nuclear explosion monitoring.

58 GEOSCIENCES↗
DOE OSTI2025

Brillouin Sensing with PCA, and PCA-Based Neural Networks for Efficient Temperature Monitoring

This work explores peak estimation techniques in Brillouin Optical Time Domain Analysis (BOTDA), emphasizing both accuracy and efficiency. Euclidean distance measurement method is applied to principal components derived from Brillouin Gain Spectrum data. It offers a major speed advantage being 180 170 times faster than traditional curve fitting methods such as Lorentzian curve fitting, while maintaining similar accuracy. Additionally, a PCA- based neural network model shows significant reduction of peak estimation time compared to Lorentzian fitting. Results show Brillouin frequency shift errors lie under 0.75 MHz in both Euclidean distance-based and neural network-based methods, both of which utilize PCA components. For large data sets and long length fibers, PCA- assisted neural network for peak estimation would be an efficient solution.

Distributed optical fiber sensing↗
DOE OSTI2022

SMALE: Enhancing Scalability of Machine Learning Algorithms on Extreme-Scale Computing Platforms

Deployment and execution of machine learning tasks on extreme-scale computing platforms face several significant technical challenges: 1) High computing cost incurred by dense networks – The computing workload of deep networks with densely-connected topology increases rapidly with the network size, imposing a non-scalable computing model of extreme-scale computing platforms; 2) Non-optimized workload distribution – Many advanced deep learning algorithms, e.g., sparsification and irregular net-work topology, produce very unbalanced workload distribution on extreme-scale computing platforms. The computation efficiency is greatly hindered by the incurred data and computation redundancies as well as long tails of the node with extensive workload; 3) Constraints in data movement and I/O bottle-neck – Inter-node data movement in extreme-scale computing platforms are associated with high energy and latency costs, and subject to the constraints of I/O bandwidth; and 4) Generalization of algorithm realization and acceleration on computing platforms – The large varieties of machine learning algorithms and structures of extreme-scale computing platforms make the derivation of a generalized algorithm realization and acceleration method very challenging, which, however, is the requirement by domain scientists and interested users. We call the above challenges Smale’s Problems in Machine Learning and Understanding for High-Performance Computing Scientific Discovery. The objective of our three-year research project is to develop a holistic innovation set at structure, assembly, and acceleration layers of machine learning algorithms to address the above challenges in algorithm deployment and execution. Three tasks are particularly performed, including: At the algorithm structure level, we investigate the techniques that can structurally sparsify on the topology of deep networks for computing workload reduction. We also study clustering and pruning techniques that can optimize the workload distributions over the extreme-scale computing platforms; At the algorithm assembly level, we derive a unified learning framework for unsupervised transfer learning and dynamic growing capabilities. Novel training methods are also exploited to enhance the training efficiency of the proposed framework; At the algorithm acceleration level, we will develop a series of techniques that can accelerate the computation of sparse matrix operations, which are one of the core executions in deep learning and optimize memory access of the concerned platforms. Our proposed techniques attack the fundamental problems in machine learning algorithms running on extreme-scale computing platforms by vertically integrating the solutions at three closely entangled layers, paving the long-term scaling path of machine learning applications under DOE context. Three tasks corresponding to the above respective research orientations are performed during the three-year project period with our collaborators at ORNL. The outcome of the proposed project is anticipated to form a holistic solution set of novel algorithms and network topologies, efficient training techniques, and fast acceleration methods to promote the computing scalability of the machine learning applications of particular interest to DOE.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Effect of fuel composition and EGR on spark-ignited engine combustion with LPG fueling: Experimental and numerical investigation

This paper presents an experimental and numerical investigation of a spark-ignited (SI) cooperative fuel research (CFR) engine fueled with different liquefied petroleum gas (LPG) fuels and exhaust gas recirculation (EGR). Here, the effects of LPG fuel composition on engine combustion characteristics are initially evaluated at two different compression ratios (CR). Results show normal combustion at CR 7 and heavy knocking combustion at CR 10 for all the tested fuels, with a more substantial impact for the LPG fuel with high proportions of n-butane species. The Livengood-Wu (LW) integral method is then used to analyze the knock occurrence risk of individual fuel based on the reactivity of the tested fuels. The introduction of EGR then demonstrates the potential of knock intensity reduction below the borderline knock limit. A zonal-based kinetic interactions study is also performed to understand the knock mitigation effectiveness of EGR over the pressure–temperature domain relevant to SI engine operation. Finally, a multidimensional, computational fluid dynamics (CFD) simulation model is shown to predict the LPG combustion characteristics and presents the evolution of in-cylinder temperature and chemical species to demonstrate the development of end-gas autoignition events without and with EGR.

02 PETROLEUM↗
DOE OSTI2021

Transfer learning of memory kernels for transferable coarse-graining of polymer dynamics

The present work concerns the transferability of coarse-grained (CG) modeling in reproducing the dynamic properties of the reference atomistic systems across a range of parameters. In particular, we focus on implicit-solvent CG modeling of polymer solutions. The CG model is based on the generalized Langevin equation, where the memory kernel plays the critical role in determining the dynamics in all time scales. Thus, we propose methods for transfer learning of memory kernels. The key ingredient of our methods is Gaussian process regression. By integration with the model order reduction via proper orthogonal decomposition and the active learning technique, the transfer learning can be practically efficient and requires minimum training data. Through two example polymer solution systems, we demonstrate the accuracy and efficiency of the proposed transfer learning methods in the construction of transferable memory kernels. The transferability allows for out-of-sample predictions, even in the extrapolated domain of parameters. Built on the transferable memory kernels, the CG models can reproduce the dynamic properties of polymers in all time scales at different thermodynamic conditions (such as temperature and solvent viscosity) and for different systems with varying concentrations and lengths of polymers.

Ma, Zhan↗