Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “inference accelerators”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

FlowPM: Distributed TensorFlow implementation of the FastPM cosmological N-body solver

Here, we present FlowPM, a Particle-Mesh (PM) cosmological N-body code implemented in Mesh-TensorFlow for GPU-accelerated, distributed, and differentiable simulations. We implement and validate the accuracy of a novel multi-grid scheme based on multiresolution pyramids to compute large-scale forces efficiently on distributed platforms. We explore the scaling of the simulation on large-scale supercomputers and compare it with corresponding Python based PM code, finding on an average 10x speed-up in terms of wallclock time. We also demonstrate how this novel tool can be used for efficiently solving large scale cosmological inference problems, in particular reconstruction of cosmological fields in a forward model Bayesian framework with hybrid PM and neural network forward model. We provide skeleton code for these examples and the entire code is publicly available at https://github.com/modichirag/flowpm[Formula presented].

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Predicting Missing Regions in Charged Particle Tracks Using a Sparse 3D Convolutional Neural Network

The 2x2 Demonstrator is a prototype detector for the Deep Underground Neutrino Experiment (DUNE)'s Near Detector. Both the 2x2 Demonstrator and the Near Detector itself will have inactive regions wherein there is no sensitivity to charge deposition and light signals that arise from charged particle interactions with liquid argon. In the 2x2, these inactive regions are positioned in-between the active detector modules, which introduces the challenge of inferring what charge signals ought to look like in these regions. This study explores the use of a Sparse 3D Convolutional Neural Network (ConvNet) to infer missing regions in charged particle tracks. Hits corresponding to energy depositions are voxelized into a three-dimensional (3D) grid for each track. Inactive regions within the tracks are replaced with a dense, rectangular 3D grid of voxels, ensuring consistent step sizes in X, Y, and Z directions. Voxels in these dense regions are initialized with an energy value of -1, indicating nonphysical energy or charge. The model is trained to predict which voxels should activate as part of the track and which should not, with the goal of eventually inferring the missing charge or energy values in these voxels. Results indicate that the model accurately predicts track voxels within ±1 unit in X, Y, or Z directions and effectively identifies non-track voxels, despite some overprediction. The approach shows promise in prediction of missing track regions with some accuracy.

43 PARTICLE ACCELERATORS↗
DOE OSTI2022

Inferring safety critical events from vehicle kinematics in naturalistic driving environment: Application of deep learning Algorithms

Advances in sensing technology has enabled the collection of countless terabytes of second-by-second kinematics data. Such data provides opportunities for real-time monitoring of driving behavior and identification of safety critical events (SCEs) including crashes and near crashes. The concept of volatility is relevant in this context, which identifies instability and erratic variations in driving behavior prior to involvement in SCEs. This study utilized vehicle kinematics from a large-scale naturalistic driving data to develop a deep learning approach based on 1D convolutional neural networks (CNN) for inferring SCEs. The data are unique in the sense that such accurate pre-crash data at high fidelity are not available in traditional crash repositories. This study contributes to the literature by providing a first attempt at predicting responses to SCEs by developing deep learning-based CNN architectures using novel driving volatility based kinematic thresholds for a sample of 9553 events. The key contribution lies in developing a volatility-based CNN input layout that is acceptable to CNN schemes and represents the motion kinematics such as speed, acceleration and volatility measures. Several 1D-CNN architectures were developed using layers, numbers of convolutions, layer patterns, and kernels. Shallow and deep architectures were tested, revealing higher accuracy of shallow architectures in detecting SCEs. The optimal number of epochs were identified using an early stopping method while the CNN performance was improved by increasing the number of epochs. The ensemble CNN had the highest predictive accuracy of 95.6% for detection of crashes and near crashes, which was 2.5% higher than the optimal CNN using 20% hold out test data. The ensemble CNN also outperformed classical machine learning models and model performance reported in past studies on detection of SCEs. Finally, these results have implications for identification of safety hotspots and providing real-time alerts and warnings in connected and highly automated vehicle environment including society of automotive engineers levels 3–5.

42 ENGINEERING↗
DOE OSTI2026

MIND-MAC: Multi-Level In-memory Quasi Non-Destructive MAC Operation in Compact 2T-nC FeRAM for Efficient DNN Accelerator

We present MIND-MAC, a compact 2T-nC FeRAM architecture that performs multi-level, quasi-non-destructive in-memory multiply–accumulate (MAC) for deep neural networks. By exploiting voltage-controlled partial domain switching in MFM capacitors and read-transistor amplification, the cell stores multi-bit weights and gates bit-serial inputs to produce an accumulated current on shared lines. We combine TCAD-extracted parasitics with experimentally calibrated ferroelectric models in SPICE to validate device-/circuit-level behavior, and validate multi-level sensing and QNRO with measurements on a fabricated 2T-3C test vehicle. An analytical system model maps MIND-MAC to a 6-GB main-memory in-memory compute (IMC) architecture and benchmarks VGG13 inference in 61.08 ms at 964.99 mJ. Results indicate high density, reduced rewrite overhead, and energy efficiency, positioning 2T-nC FeRAM as a promising IMC candidate for next-generation AI hardware.

36 MATERIALS SCIENCE↗
DOE OSTI2023

AutoReP: Automatic ReLU Replacement for Fast Private Network Inference

The proliferation of the Machine-Learning-As-A-Service (MLaaS) market has brought to light a number of clients’ data privacy and security concerns. One promising solution is private inference (PI) techniques using cryptographic primitives. These techniques often come with high computation and communication overhead associated with the non-linear operator such as ReLU. Several approaches have been developed in reducing the number of ReLU operations, however, they either require a heuristic threshold selection or introduce significant accuracy drop. This work presents AutoReP, a gradient-based framework for non-linear operators reduction that aims to mitigate these concerns from a systematic perspective. AutoReP automates the process of discrete selection of ReLU and polynomial functions on neurons to accelerate PI applications. We also introduce distribution-aware polynomial approximation (DaPa) to accurately approximate ReLUs under given distribution, preserving model expressivity. Our experimental results demonstrate significant accuracy improvements of 6.12% (94.31%, 12.9K ReLU budget, CIFAR-10), 8.39% (74.92%, 12.9K ReLU budget, CIFAR-100), and 9.45% (63.69%, 55K ReLU budget, Tiny-ImageNet) over current state-of-the-art methods, e.g., SNL. Morever, AutoReP is applied to EfficientNet-B2 on ImageNet dataset, and achieved 75.55% accuracy with 176.1 × ReLU budget reduction.

Peng, Hongwu↗
DOE OSTI2023

hIPPYlib-MUQ: A Bayesian Inference Software Framework for Integration of Data with Complex Predictive Models under Uncertainty

Bayesian inference provides a systematic framework for integration of data with mathematical models to quantify the uncertainty in the solution of the inverse problem. However, the solution of Bayesian inverse problems governed by complex forward models described by partial differential equations (PDEs) remains prohibitive with black-box Markov chain Monte Carlo (MCMC) methods. We present hIPPYlib-MUQ, an extensible and scalable software framework that contains implementations of state-of-the art algorithms aimed to overcome the challenges of high-dimensional, PDE-constrained Bayesian inverse problems. These algorithms accelerate MCMC sampling by exploiting the geometry and intrinsic low-dimensionality of parameter space via derivative information and low rank approximation. The software integrates two complementary open-source software packages, hIPPYlib and MUQ. hIPPYlib solves PDE-constrained inverse problems using automatically-generated adjoint-based derivatives, but it lacks full Bayesian capabilities. MUQ provides a spectrum of powerful Bayesian inversion models and algorithms, but expects forward models to come equipped with gradients and Hessians to permit large-scale solution. By combining these two complementary libraries, we created a robust, scalable, and efficient software framework that realizes the benefits of each and allows us to tackle complex large-scale Bayesian inverse problems across a broad spectrum of scientific and engineering disciplines. To illustrate the capabilities of hIPPYlib-MUQ, we present a comparison of a number of MCMC methods available in the integrated software on several high-dimensional Bayesian inverse problems. These include problems characterized by both linear and nonlinear PDEs, various noise models, and different parameter dimensions. The results demonstrate that large (~ 50×) speedups over conventional black box and gradient-based MCMC algorithms can be obtained by exploiting Hessian information (from the log-posterior), underscoring the power of the integrated hIPPYlib-MUQ framework.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Optical Imaging of Laser-Driven Fast Electron Weibel-like Filamentation in Overcritical Density Plasma

We report on the measurement of filamented transport of laser-generated fast electron beams in near-critical density plasma. A relativistic intensity long-wave-infrared laser irradiated a hydrodynamically shaped helium gas flow at an electron density n e ≃ 10 25 m − 3 , generating a large flux of fast electrons that propagated beyond the critical surface. The beam-to-background electron density ratio was sufficiently high to drive growth of Weibel-like filamentation, which was measured by optical probing to extend up to 800 μ m with radii ∼ 10 μ m . Particle-in-cell simulations reproduce the main features of the filamentation generation, suggesting that collisionless processes are dominant in these interactions. Expansion of the filaments after formation infers a fast electron heated plasma temperature ∼ 400 eV in the overcritical density plasma. Published by the American Physical Society 2025

43 PARTICLE ACCELERATORS↗
DOE OSTI2023

Improving deep learning model performance under parametric constraints for materials informatics applications

Abstract Modern machine learning (ML) and deep learning (DL) techniques using high-dimensional data representations have helped accelerate the materials discovery process by efficiently detecting hidden patterns in existing datasets and linking input representations to output properties for a better understanding of the scientific phenomenon. While a deep neural network comprised of fully connected layers has been widely used for materials property prediction, simply creating a deeper model with a large number of layers often faces with vanishing gradient problem, causing a degradation in the performance, thereby limiting usage. In this paper, we study and propose architectural principles to address the question of improving the performance of model training and inference under fixed parametric constraints. Here, we present a general deep-learning framework based on branched residual learning (BRNet) with fully connected layers that can work with any numerical vector-based representation as input to build accurate models to predict materials properties. We perform model training for materials properties using numerical vectors representing different composition-based attributes of the respective materials and compare the performance of the proposed models against traditional ML and existing DL architectures. We find that the proposed models are significantly more accurate than the ML/DL models for all data sizes by using different composition-based attributes as input. Further, branched learning requires fewer parameters and results in faster model training due to better convergence during the training phase than existing neural networks, thereby efficiently building accurate models for predicting materials properties.

36 MATERIALS SCIENCE↗
DOE OSTI2022

A laser ultrasonics-based approach for rapid screening of high entropy alloys

The primary objective of this seed project is to develop a laser ultrasonics-based characterization methodology for rapid metallic materials design and discovery via in situ determination of phases, microstructures and elastic properties with respect to temperature. Ultrasonic waves are strongly affected by material microstructure, and therefore, serve as a facile means to probe elastic properties, phase content and their size distributions and volume fractions. In this study, a laser ultrasonic technique will be used to systematically study the evolution of properties in a set of interrelated simple binary alloys and high entropy alloys (HEA). Phase transformations and microstructural changes inferred from the ultrasonic signals will be correlated with electron microscopy data and predictions using calculations of phase diagrams (CALPHAD). The non-contact and non-destructive ultrasonic testing approach developed in this study could overcome several limitations associated with current material characterization methods for materials discovery. It is expected that the products of the proposed work will be utilized to evaluate novel graded composition HEAs currently being developed at the Idaho National Laboratory (INL) using advanced manufacturing methods based on direct energy deposition and spark plasma sintering processes, and contribute to accelerate the discovery of HEAs.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Secure Collaborative Environment for Seamless Sharing of Scientific Knowledge

In a secure collaborative environment, tera-bytes of data generated from powerful scientific instruments are used to train secure machine learning (ML) models on exascale computing systems, which are then securely shared with internal or external collaborators as cloud-based services. Devising such a secure platform is necessary for seamless scientific knowledge sharing without compromising individual, or institute-level, intellectual property and privacy details. By enabling new computing opportunities with sensitive data, we envision a secure collaborative environment that will play a significant role in accelerating scientific discovery. Several recent technological advancements have made it possible to realize these capabilities. In this paper, we present our efforts at ORNL toward developing a secure computation platform. We present a use case where scientific data generated from complex instruments, like those at the Spallation Neutron Source (SNS), are used to train a differential privacy enabled deep learning (DL) network on Summit, which is then hosted as a secure multi-party computation (MPC) service on ORNL’s Compute and Data Environment for Science (CADES) cloud computing platform for third-party inference. In this feasibility study, we discuss the challenges involved, elaborate on leveraged technologies, analyze relevant performance results and present the future vision of our work to establish secure collaboration capabilities within and outside of ORNL.

Yoginath, Srikanth↗
DOE OSTI2020

Temperature Trends in the Northwestern Tibetan Plateau Constrained by Ice Core Water Isotopes Over the Past 7,000 Years

Abstract The reasons for the Holocene temperature conundrum, known as the inconsistency between the reconstructed cooling and the inferred warming simulations during the Holocene, remain unclear. Temperature reconstructions from the Tibetan Plateau (TP) provide important insights for understanding the Holocene temperature conundrum due to enhanced sensitivity to climate at high altitudes. Given the significant positive correlation between air temperature and δ 18 O in precipitation over the northern TP, the stable isotopic records of ice cores recovered from this area are widely used for paleotemperature reconstruction. Here we present a new high‐resolution δ 18 O record from the Chongce ice cores to bedrock, dated back to 7 ka BP by the accelerator mass spectrometry (AMS) 14 C dating technique. Our reconstructed temperature record shows a long‐term warming trend until ~2 ka BP, followed by an abrupt change to a relatively cool period until the start of the industrial‐era warming. This record challenges the widely recognized Holocene reconstruction from the neighboring Guliya ice core. It is also different from many previous temperature reconstructions, most of which have summer biases and show a long‐term cooling trend over the past two millennia. In addition, our record shows that temperatures during the recent decades are almost the highest during the past 7 ka BP, highlighting the unusual warming forced by anthropogenic greenhouse gases.

Pang, Hongxi↗
DOE OSTI2025

A Bayesian framework to investigate radiation reaction in strong fields

Recent experiments aiming to measure phenomena predicted by strong-field quantum electrodynamics (SFQED) have done so by colliding relativistic electron beams and high-power lasers. In such experiments, measurements of collision parameters are not always feasible. However, precise knowledge of these parameters is required to accurately test SFQED. Here, we present a novel Bayesian inference procedure that infers collision parameters that could not be measured on-shot. This procedure is applicable to all-optical non-linear Compton scattering experiments investigating radiation reaction. The framework allows multiple diagnostics to be combined self-consistently and facilitates the inclusion of known information pertaining to the collision parameters. Using this Bayesian analysis, the relative validity of the classical, quantum-continuous and quantum-stochastic models of radiation reaction was compared for several test cases, which demonstrates the accuracy and model selection capability of the framework and highlight its robustness if the experimental values of fixed parameters differ from their values in the models.

47 OTHER INSTRUMENTATION↗
DOE OSTI2023

Hydrologic Impacts of a Strike-Slip Fault Zone: Insights from Joint 3D Body-Wave Tomography of Rock Valley

The Rock Valley fault zone (RVFZ), an intraplate strike-slip fault zone in the southern Nevada National Security Site (NNSS), hosted a series of very shallow (<3 km) earthquakes in 1993. The RVFZ may also have hydrological significance within the NNSS, potentially playing a role in regional groundwater flow, but there is a lack of local hydrological data. In the Spring of 2021, we collected active-source accelerated weight drop seismic data over part of the RVFZ to better characterize the shallow subsurface. Here, we manually picked ~17,000 P-wave travel times and over 14,000 S-wave travel times, which were inverted for P-wave velocity (V P ), S-wave velocity (V S ), and V P /V S ratio in a 3D joint tomographic inversion scheme. Seismic velocities are imaged as deep as ~700 m in areas and generally align with geologic and structural expectations. V P and V S are relatively reduced near mapped and inferred faults, with the most prominent lower V P and V S zone around the densest collection of faults. We image V P /V S ratios ranging from ~1.5 to ~2.4, the extremes of which occur at a depth of ~100 m and are juxtaposed across a fault. One possible interpretation of the imaged seismic velocities is enhanced fault damage near the densest collection of faults with relatively higher porosity and/or crack density at ~100 m depth, with patches of semiperched groundwater present in the sedimentary rock in higher V P /V S areas and drier rock in lower V P /V S areas. A relatively higher V P /V S area beneath the densest faults persists at depth, which suggests percolation of groundwater via the fault damage zone to the regionally connected lower carbonate aquifer. Potentially, the presence and movement of groundwater may have played a role in the 1993 earthquake aftershocks.

58 GEOSCIENCES↗
DOE OSTI2025

Comment on "An implementation of neural simulation-based inference for parameter estimation in ATLAS''

The paper titled "An implementation of neural simulation-based inference for parameter estimation in ATLAS" by the ATLAS collaboration (arXiv:2412.01600v1 [hep-ex]) describes the implementation of neural simulation-based inference for a measurement analysis performed by ATLAS. The uncertainties in the analysis arising from the finiteness of the simulated datasets are estimated using a novel double-bootstrapping technique described in that work. In the present comment, it is claimed and demonstrated, using a toy example, that the double-bootstrapping technique does not actually capture the aforementioned uncertainties.

43 PARTICLE ACCELERATORS↗
DOE OSTI2022

A laser ultrasonics-based approach for rapid screening of high entropy alloys

This project demonstrates a laser ultrasonics-based characterization methodology for rapid metallic materials design and discovery via in situ determination of phases, microstructures and elastic properties with respect to temperature. Ultrasonic waves are strongly affected by material microstructure, and therefore serve as a facile means to probe elastic properties, phase content and their size distributions and volume fractions. In this study, a laser ultrasonic technique is used to systematically study the evolution of properties in a set of interrelated simple binary alloys and high entropy alloys (HEA). Phase transformations and microstructural changes inferred from the ultrasonic signals will be correlated with electron microscopy data and predictions using calculations of phase diagrams (CALPHAD). The non-contact and non-destructive ultrasonic testing approach developed in this study could overcome several limitations associated with current material characterization methods for materials discovery. It is expected that laser ultrasonics-based methodology developed in this work will be utilized to evaluate novel graded composition HEAs currently being developed at the Idaho National Laboratory (INL) using advanced manufacturing methods based on direct energy deposition and spark plasma sintering processes, and contribute to accelerate the discovery of HEAs.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Considerations for Distributed Edge Data Centers and Use of Building Loads to Support Large Interconnections

The rapid expansion of artificial intelligence (AI) and machine learning is driving unprecedented electricity demand from data centers. It is predicted that by 2030, 90% of AI workloads will be inference-based, requiring interconnection of multiple low-latency edge data centers (<20 MW) sited closer to end users - often on already constrained distribution feeders. Although individually small, these loads can aggregate to large loads per feeder, straining infrastructure, creating multi-year interconnection delays, and driving up customer costs. This paper proposes a data center-focused grid-integration framework that combines feeder hosting capacity analysis with building energy efficiency, building load flexibility, and waste heat reuse to expand effective feeder and substation headroom. Such approaches can reduce interconnection delays, lower costs for ratepayers, and accelerate AI-ready infrastructure deployment.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

A Data-Driven Nonparametric Approach for Probabilistic Load-Margin Assessment Considering Wind Power Penetration

A modern power system is characterized by an increasing penetration of wind power, which results in large uncertainties in its states. These uncertainties must be quantified properly; otherwise, the system security may be threatened. Facing this challenge, here we propose a cost-effective, data-driven approach to assessing a power system's load margin probabilistically. Using actual wind data, a kernel density estimator is applied to infer the nonparametric wind speed distributions, which are further merged into the framework of a vine copula. The latter enables us to simulate complex multivariate and highly dependent model inputs with a variety of bivariate copulae that precisely represent the tail dependence in the correlated samples. Furthermore, to reduce the prohibitive computational time of traditional Monte-Carlo simulations that process a large amount of samples, we propose to use a nonparametric, Gaussian-process-emulator-based reduced-order model to replace the original complicated continuation power-flow model through a Bayesian-learning framework. To accelerate the convergence rate of this Bayesian algorithm, a truncated polynomial chaos surrogate, which serves as a highly efficient, parametric Bayesian prior, is developed. This emulator allows us to execute the time-consuming continuation power-flow solver at the sampled values with a negligible computational cost. Results of simulations that are performed on several test systems reveal the impressive performance of the proposed method in the probabilistic load-margin assessment.

17 WIND ENERGY↗
DOE OSTI2022

In-situ Characterizations of Microstructural Degradation of Perovskite Solar Cells

Rapid progress has been achieved in perovskite solar cells (PSCs), and their efficiencies have improved from 3.8 % to 24.2 % in less than a decade. With low-cost processing, PSCs have shown exciting photovoltaic (PV) properties, such as effective optical absorption, a long carrier lifetime, and unique defect tolerance. While recent studies demonstrated improved stability up to 100 days, PSC technology is still challenged to meet the stringent industry requirements for commercialization. Despite considerable efforts, the underlying physical mechanisms for the inferior stability of PSCs are not well understood. One reason for this divergence is that many established measurement techniques (e.g., quantum efficiency, photoluminescence) probe the properties on length scales far greater than that of electronic and/or structural inhomogeneity (i.e., < 1 μm near grain boundaries) and therefore characterize convoluted and/or averaged properties. Ion/electron beam-based techniques have been extensively used to access the microstructures of PSCs, enabling atomic/nanoscale structural, chemical, optical, and electrical characterizations. For example, focused ion beam (FIB) milling produces an atomically smooth surface that minimizes the artifacts attributed to the surface roughness. FIB techniques can also create a well-defined cross-section of PSCs without mechanical damage in a physical cleaving sample preparation. While powerful, there are some concerns about possible beam damage of inorganic-organic perovskites via chemical-bond breakage and local heating. This project aims to comprehensively understand how the microstructural/interfacial properties of PSCs (e.g., Methylammonium Lead Iodide [MAPbI 3 ]) are modified under the irradiating ion beams. Specifically, we investigate the sub-surface properties of PSCs before and after Ar-ion beam injections. Kelvin probe force microscopy (KPFM) measures the contact potential differences (CPDs). Photoluminescence (PL) microscopy in conjunction with Finite-Difference Time-Domain (FDTD) simulations infers the formation of a “dead layer” (< 15 nm) on the subsurface of MAPbI 3 during Ar+ milling processes while preserving the initial bulk properties. The x-ray photoemission spectroscopy (XPS) confirms this modified surface is a lead-rich and iodine-deficient surface. We initiate customizing in-situ measurement setup while measuring the local optical and electrical properties of PSC under thermal (cooling, heating) and light stressors. Our results provide in-depth knowledge of the ion-beam impact on metal-halide perovskites and how this modified sub-surface impacts their properties under accelerated stressors of light and heat. Intensive Monte Carlo simulations of an electron beam interacting with PSCs provide the beam energy distribution in PSCs, proposing possible measurement conditions of using e-beam with minimizing beam damage. Our in-situ measurement platform can accommodate the diverse architecture of PSC devices for studying deterioration mechanisms under mixed environmental stressors.

14 SOLAR ENERGY↗