Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “out of memory”

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 379 records · Page 21
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Performance Study of CXL Memory Topology

This paper presents a comprehensive evaluation of the performance impact of various Compute Express Link (CXL) memory topologies, with a particular emphasis on CXL switches, in the context of High- Performance Computing (HPC) and Large Language Model (LLM) inference workloads. Our study unveils significant performance variations across different topologies, demonstrating that certain configurations yield superior performance for specific workloads. These findings underscore the critical importance of tailored topol- ogy selection in optimizing system performance. Additionally, we address the inherent challenges associated with integrating CXL switches, including overhead considerations and routing complex- ities. Our research highlights the necessity for thorough evalua- tion methodologies to fully leverage CXL technology’s potential in contemporary computing environments. These insights provide valuable guidance for system architects and data center operators in designing and optimizing CXL-based infrastructures for diverse workload requirements.

CXL, memory, Artificial Intelligence (AI), HPC↗
DOE OSTI2026

qSIEVE: Efficient qLDPC Memory via Systolic Movement in Atom Arrays

As quantum machines have scaled up in their number of qubits, significant research has turned towards increasing their fidelity with quantum error correction codes. Although promising results have been shown with the surface code, which only requires near-neighbor connections between qubits, the high qubit overhead of such local codes promises to be problematic. Consequently, recent work has explored non-local quantum LDPC (qLDPC) codes, which have good asymptotic encoding rates. Despite theoretical progress, hardware implementations of these codes have been a longstanding challenge. At the experimental level, demonstrations of movement based communication on atom arrays suggest this is a powerful new primitive to achieve non-local connectivity. Leveraging this, we present a protocol for implementing non-local qLDPC codes in hardware. Our protocol, qSIEVE, is a co-design of such codes with movement in atom arrays. qSIEVE defines a restricted family of qLDPC codes that can be implemented efficiently with systolic movement. We then quantify the utility of qSIEVE in the context of a complete fault tolerant architecture. We compare the cost of implementing benchmark programs in a standard, surface code only architecture and a mixed architecture where data is stored in qLDPC memory with qSIEVE and loaded to surface codes for computation.

Quantum error correction↗
DOE OSTI2023

Electrophysiological Signatures of Visual Recognition Memory across All Layers of Mouse V1

In mouse primary visual cortex (V1), familiar stimuli evoke significantly altered responses when compared with novel stimuli. This stimulus-selective response plasticity (SRP) was described originally as an increase in the magnitude of visual evoked potentials (VEPs) elicited in layer 4 (L4) by familiar phase-reversing grating stimuli. SRP is dependent on NMDA receptors (NMDARs) and has been hypothesized to reflect potentiation of thalamocortical (TC) synapses in L4. However, recent evidence indicates that the synaptic modifications that manifest as SRP do not occur on L4 principal cells. To shed light on where and how SRP is induced and expressed in male and female mice, the present study had three related aims: (1) to confirm that NMDAR are required specifically in glutamatergic principal neurons of V1, (2) to investigate the consequences of deleting NMDAR specifically in L6, and (3) to use translaminar electrophysiological recordings to characterize SRP expression in different layers of V1. We find that knock-out (KO) of NMDAR in L6 principal neurons disrupts SRP. Current-source density (CSD) analysis of the VEP depth profile shows augmentation of short latency current sinks in layers 3, 4, and 6 in response to phase reversals of familiar stimuli. Multiunit recordings demonstrate that increased peak firing occurs in response to phase reversals of familiar stimuli across all layers, but that activity between phase reversals is suppressed. Together, these data reveal important aspects of the underlying phenomenology of SRP and generate new hypotheses for the expression of experience-dependent plasticity in V1. SIGNIFICANCE STATEMENT Repeated exposure to stimuli that portend neither reward nor punishment leads to behavioral habituation, enabling organisms to dedicate attention to novel or otherwise significant features of the environment. The neural basis of this process, which is so often dysregulated in neurologic and psychiatric disorders, remains poorly understood. Learning and memory of stimulus familiarity can be studied in mouse visual cortex by measuring electrophysiological responses to simple phase-reversing grating stimuli. The current study advances knowledge of this process by documenting changes in visual evoked potentials (VEPs), neuronal spiking activity, and oscillations in the local field potentials (LFPs) across all layers of mouse visual cortex. In addition, we identify a key contribution of a specific population of neurons in layer 6 (L6) of visual cortex.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Control of a Three-Phase Grid-Connected Voltage-Sourced Converter Using Long Short-Term Memory Networks

With the rise of inverter-based resources (IBRs) within the power system, the control of grid-connected converters (GCCs) has become pertinent due to the fact they interface IBRs to the grid. The conventional method of control for a GCC such as the voltage-sourced converter (VSC) is through a decoupled control loop in the synchronous reference frame. However, this model-based control method is sensitive to parameter changes causing deterioration in controller performance. Data-driven approaches such as machine learning can be utilized to design controllers that are capable of operating GCCs in various system conditions. This work explores a deep learning-based control method for a three-phase grid-connected VSC, specifically utilizing a long short-term memory (LSTM) network for robust control. Simulations of a conventional controlled VSC are conducted using Simulink to collect data for training the LSTM-based controller. The LSTM model is built and trained using the Keras and TensorFlow libraries in Python and tested in Simulink. The performance of the LSTM-based controller is evaluated under different case studies and compared to the conventional method of control. Simulation results demonstrate the effectiveness of this approach by outperforming the conventional controller and maintaining stability under different system parameter changes.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Anomaly Detection in Liquid Sodium Cold Trap Operation with Multisensory Data Fusion Using Long Short-Term Memory Autoencoder

Sodium-cooled fast reactors (SFR), which use high temperature fluid near ambient pressure as coolant, are one of the most promising types of GEN IV reactors. One of the unique challenges of SFR operation is purification of high temperature liquid sodium with a cold trap to prevent corrosion and obstructing small orifices. We have developed a deep learning long short-term memory (LSTM) autoencoder for continuous monitoring of a cold trap and detection of operational anomaly. Transient data were obtained from the Mechanisms Engineering Test Loop (METL) liquid sodium facility at Argonne National Laboratory. The cold trap purification at METL is monitored with 31 variables, which are sensors measuring fluid temperatures, pressures and flow rates, and controller signals. Loss-of-coolant type anomaly in the cold trap operation was generated by temporarily choking one of the blowers, which resulted in temperature and flow rate spikes. The input layer of the autoencoder consisted of all the variables involved in monitoring the cold trap. The LSTM autoencoder was trained on the data corresponding to cold trap startup and normal operation regime, with the loss function calculated as the mean absolute error (MAE). The loss during training was determined to follow log-normal density distribution. During monitoring, we investigated a performance of the LSTM autoencoder for different loss threshold values, set at a progressively increasing number of standard deviations from the mean. The anomaly signal in the data was gradually attenuated, while preserving the noise of the original time series, so that the signal-to-noise ratio (SNR) averaged across all sensors decreased below unity. Results demonstrate detection of anomalies with sensor-averaged SNR < 1.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2021

Coexisting Multiple Martensites in Ni 57-x Mn 21+x Ga 22 Ferromagnetic Shape Memory Alloys: Crystal Structure and Phase Transition

A comprehensive study of the crystal structure and phase transition as a function of temperature and composition in Ni 57–x Mn 21+x Ga 22 (x = 0, 2, 4, 5.5, 7, 8) (at. %) magnetic shape memory alloys was performed by a temperature-dependent synchrotron X-ray diffraction technique and transmission electron microscopy. A phase diagram of this Ni 57–x Mn 21+x Ga 22 alloy system was constructed. The transition between coexisting multiple martensites with monoclinic and tetragonal structures during cooling was observed in the Ni 51.5 Mn 26.5 Ga 22 (x = 5.5) alloy, and it was found that 5M + 7M multiple martensites coexist from 300 K to 160 K and that 5M + 7M + NM multiple martensites coexist between 150 K and 100 K. The magnetic-field-induced transformation from 7M martensite to NM martensite at 140 K where 5M + 7M + NM multiple martensites coexist before applying the magnetic field was observed by in situ neutron diffraction experiments. The present study is instructive for understanding the phase transition between coexisting multiple martensites under external fields and may shed light on the design of novel functional properties based on such phase transitions.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Long Short-Term Memory Networks for Pattern Recognition of Synthetical Complete Earthquake Catalog

Exploring the spatiotemporal distribution of earthquake activity, especially earthquake migration of fault systems, can greatly to understand the basic mechanics of earthquakes and the assessment of earthquake risk. By establishing a three-dimensional strike-slip fault model, to derive the stress response and fault slip along the fault under regional stress conditions. Our study helps to create a long-term, complete earthquake catalog. We modelled Long-Short Term Memory (LSTM) networks for pattern recognition of the synthetical earthquake catalog. The performance of the models was compared using the mean-square error (MSE). Our results showed clearly the application of LSTM showed a meaningful result of 0.08% in the MSE values. Our best model can predict the time and magnitude of the earthquakes with a magnitude greater than Mw = 6.5 with a similar clustering period. These results showed conclusively that applying LSTM in a spatiotemporal series prediction provides a potential application in the study of earthquake mechanics and forecasting of major earthquake events.

58 GEOSCIENCES↗
DOE OSTI2025

Ultra-Fast Non-Volatile Resistive Switching Devices with Over 512 Distinct and Stable Levels for Memory and Neuromorphic Computing

Low-current multilevel programmability with inherent non-volatility and high stability of resistance states is required for both multi-bit memory storage and deep learning accelerators but is difficult to achieve. Here, in a resistive switching system, this work realizes >512 (>9 bits) distinct non-volatile conductance levels with stable retention for each state with current levels down to the nanoampere range, highly promising for potential integration with small processing nodes with ultra-low power consumption requirements. This is achieved by demonstrating a new thin film design concept that encompasses three key features: an ultra-thin epitaxial oxygen ionic switching layer that provides a tunable energy barrier at the bottom electrode, an overcoat amorphous layer that acts as an ion migration barrier for stable state retention, and a partial conductive filament as a localized electronic transport channel to the epitaxial switching layer. A large dynamic resistance range of up to seven orders of magnitude is achieved with reset-free transitions among intermediate states, and programmability is demonstrated with ultra-fast (20 ns) pulses. Artificial neural network (ANN) simulations, based on the experimental performance and its non-idealities, demonstrate close-to-ideal inference accuracies for various Modified National Institute of Standards and Technology (MNIST) data sets.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Machine Learning Reveals Memory of the Parent Phases in Ferroelectric Relaxors Ba(Ti$_{1-x}$,Zr x )O 3

Machine learning has been establishing its potential in multiple areas of condensed matter physics and materials science. Here, in this work, an unsupervised machine learning workflow is developed and used within a framework of first-principles-based atomistic simulations to investigate phases, phase transitions, and their structural origins in ferroelectric relaxors, Ba(Ti 1-x ,Zr x )O 3 . The applicability of the workflow is first demonstrated to identify phases and phase transitions in the parent compound, a prototypical ferroelectric BaTiO 3 . Then the workflow is applied for Ba(Ti 1-x ,Zrx)O 3 with x ≤ 0.25 to reveal i) that some of the compounds bear a subtle memory of BaTiO 3 phases beyond the point of the pinched phase transition, which could contribute to their enhanced electromechanical response; ii) the existence of peculiar phases with delocalized precursors of nanodomains—likely candidates for the controversial polar nanoregions; and iii) nanodomain phases for the largest concentrations of x.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2022

CeO 2 Doping of Hf 0.5 Zr 0.5 O 2 Thin Films for High Endurance Ferroelectric Memories

Ferroelectric switching is demonstrated in CeO 2 -doped Hf 0.5 Zr 0.5 O 2 (HZCO) thin films with application in back-end-of-line compatible embedded memories. At low cerium oxide doping concentrations (2.0–5.6 mol%), the ferroelectric orthorhombic phase is stabilized after annealing at temperatures below 400 °C. HZCO ferroelectrics show reliable switching characteristics beyond 10 11 cycles in TiN/HZCO/TiN capacitors, several orders of magnitude greater than identically processed Hf 0.5 Zr 0.5 O 2 (HZO) capacitors, without sacrificing polarization and retention. Internal photoemission and photoconductivity experiments show that CeO 2 -doping introduces in-gap states in HZCO that are nearly aligned with TiN Fermi level, facilitating electron injection through these states. Furthermore, the enhanced average bulk conduction, which may lead to more uniform thermal dissipation in the HZCO films, delays irreversible degradation via breakdown that leads to device failure after repeated programming cycles.

36 MATERIALS SCIENCE↗
DOE OSTI2023

ReaxFF reactive force field model enables accurate prediction of physiochemical and mechanical properties of crystalline and amorphous shape‐memory polyurethane

Abstract We investigate reactive force field (ReaxFF) prediction of physiochemical and mechanical properties, and stimuli response behavior of a shape‐memory polyurethane (SMPU) rigid segment. We used SMPUs as a platform to link the rigid/hard segment domain crystallinity with properties that can also be experimentally measured. Specifically, we developed models for crystalline and amorphous forms of a common rigid segment of SMPUs, for example, 4,4′‐diphenylmethane diisocyanate (MDI) with n‐butanediol (BDO), denoted as MDI‐BDO. ReaxFF molecular dynamics simulations with equilibrium dynamics are performed by controlling mass, temperature, and pressure or volume to predict the structural and physiochemical properties of crystalline and amorphous rigid segments of SMPU systems, followed by simulating their XRD and FTIR patterns, respectively. Non‐equilibrium ReaxFF simulations are implemented by using uniaxial box deformation technique to study response‐behavior of crystalline SMPU under tensile loading and during stress‐relaxation. The model and simulation scheme presented in this study demonstrates a robust theoretical pathway toward application‐dictated, on‐demand design of high‐performance SMPUs in applications such as soft robotics.

Afroz, Mohammad M.↗
DOE OSTI2024

Connectivity, Pathology, and ApoE4 Interactions Predict Longitudinal Tau Spatial Progression and Memory

ABSTRACT Tau pathology spread into neocortex indicates a transition from healthy aging to Alzheimer's disease (AD). Connectivity between tau epicenters and later accumulating regions of cortex has been proposed as a mechanism of tau spread, but how this relationship changes with greater AD pathology burden or genotype is not understood. We investigated tau accumulation in two key regions, precuneus and inferior temporal cortex, using resting state functional connectivity (rsFC) and longitudinal PET imaging from a multicohort sample of cognitively unimpaired older adults. We examined how baseline tau PET, Aβ PET, and ApoE4 genotype status interact with rsFC between hippocampus and these downstream regions to predict rate of tau accumulation in neocortex. We found that the 3‐way interaction between connectivity, baseline tau, and baseline Aβ or ApoE4 status was associated with neocortical tau accumulation in precuneus and inferior temporal cortex. In addition, baseline tau, Aβ, and ApoE4 status also moderated the association between connectivity and rate of memory decline. Together, these results suggest that the extent and distribution of future tau accumulation may be predicted by the interaction of baseline connectivity, AD pathology, and genetic risk.

Neurosciences & Neurology↗
DOE OSTI2026

Imputation of urban environmental sensor data using gated attention bidirectional long short-term memory (GA-BiLSTM): methods, performance, and implications

Urban environmental monitoring networks frequently encounter significant data gaps due to sensor malfunctions, environmental disturbances, and communication failures. Reliable approaches to address these gaps are essential for ensuring the continuity and quality of environmental data streams. In this study, we developed a gated attention bidirectional long short-term memory (GA-BiLSTM) model to impute missing data in a dense urban monitoring network. Using observations from the CROCUS network in Chicago, we evaluated GA-BiLSTM against widely used approaches (XGBoost and K-nearest neighbors) under scenarios of both short-term intermittent gaps and prolonged outages. GA-BiLSTM consistently outperformed comparative methods, particularly during extended outages of up to ten days, demonstrating its ability to capture spatiotemporal dependencies across sensor nodes. Beyond performance metrics, feature importance and spatial network analyses highlighted the unexpected but critical predictive role of peripheral rural nodes, underlining their strategic value for maintaining robust urban monitoring systems. These results emphasize that advanced imputation methods can substantially improve the reliability of environmental monitoring networks and support more resilient data infrastructures for urban sustainability.

Data imputation↗
DOE OSTI2025

A randomized sketching trust-region secant method for low-memory dynamic optimization

The numerical solution of dynamic optimization problems is often limited by the memory required to store the state trajectory, which is used to evaluate the objective function and its derivatives. Recently, [R. Muthukumar et al., SIAM Journal on Optimization 31(2), pp. 1242–1275 (2021)] introduced a trust-region method for dynamic optimization that employs randomized sketching to compress the state trajectory, resulting in inexact derivative computations. By adaptively learning the sketch rank, the trust-region algorithm achieves rigorous convergence guarantees. Here, we extend this approach to use secant Hessian approximations. Due to the randomness introduced by the sketch, the traditional secant update formulae can produce poor Hessian approximations. In particular, the difference of two gradients, computed from two different sketches, may be inconsistent. To overcome this, we employ a sketched approximation of the Hessian application, in lieu of computing the gradient difference. We numerically demonstrate the improved stability of this approach on an example from PDE-constrained optimization.

dynamic optimization↗
DOE OSTI2023

Caloric Effects Induced by Uniform and Non-uniform Stress in Shape-Memory Materials

A Ginzburg–Landau model is developed that is adequate to describe a square-to-rectangle martensitic transition with associated shape-memory and superelastic properties. Using this model we study caloric effects in the vicinity of the martensitic transition induced by stress and we compare the case of a uniform uniaxial stress and the case of a non-uniform continuous distribution of stresses that produce bending of the material. The former case corresponds to an elastocaloric effect and the latter corresponds to a flexocaloric effect. The aim of the work is to quantitatively compare both cases, which we show must be accomplished in terms of equal amounts of exchanged mechanical work. It is then obtained that the flexocaloric effect is more efficient for low exchanged work but less efficient for large exchanged work.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Asynchronous distributed-memory task-parallel algorithm for compressible flows on unstructured 3D Eulerian grids

Here, we discuss the implementation of a finite element method, used to numerically solve the Euler equations of compressible flows, using an asynchronous runtime system (RTS). The algorithm is implemented for distributed-memory machines, using stationary unstructured 3D meshes, combining data-, and task-parallelism on top of the Charm++ RTS. Charm++’s execution model is asynchronous by default, allowing arbitrary overlap of computation and communication. Task-parallelism allows scheduling parts of an algorithm independently of, or dependent on, each other. Built-in automatic load balancing enables continuous redistribution of computational load by migration of work units based on real-time CPU load measurement. The RTS also features automatic checkpointing, fault tolerance, resilience against hardware failure, and supports power-, and energy-aware computation. We demonstrate scalability up to 25 x 10 9 cells at $\mathscr{O}$10 4 compute cores and the benefits of automatic load balancing for irregular workloads. The full source code with documentation is available at https://quinoacomputing.org.

42 ENGINEERING↗
DOE OSTI2023

Predictive model for real-time energy disaggregation using long short-term memory

To provide affordable energy-saving solutions for the small and medium-sized manufacturers (SMMs), we propose a unified framework for generating predictive models that support real-time disaggregation of power consumption from combined inputs, enabling automatic machine state identification simultaneously for joint analysis of energy usage patterns. Further, the proposed framework transforms raw power consumption into a time series with look-back and bootstrap capabilities for historical pattern detection, while a learning architecture utilizes the stacked long short-term memory (LSTM) layers as encoders for embedding generation with sequential awareness. Experimental results demonstrate 93.65% minimum accuracy in ideal case of real-time energy usage and machine state prediction.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

GrainNN: A neighbor-aware long short-term memory network for predicting microstructure evolution during polycrystalline grain formation

High fidelity simulations of grain formation in alloys are an indispensable tool for process-to-mechanical-properties characterization. Such simulations, however, can be computationally expensive as they require fine spatial and temporal discretizations. Their cost becomes an obstacle to parametric studies and ensemble runs and ultimately makes downstream tasks like optimal control and uncertainty quantification challenging. To enable such downstream tasks, we introduce GrainNN, an efficient and accurate reduced-order model for epitaxial grain growth in additive manufacturing conditions. GrainNN is a sequence-to-sequence long-short-term-memory (LSTM) deep neural network that evolves the dynamics of manually crafted features. Its innovations are (1) an attention mechanism with grain-microstructure-specific transformer architecture; and (2) an overlapping combination of several clones of the network to generalize to grain configurations that are different from those used for training. This design enables GrainNN to predict grain formation for unseen physical parameters, grain number, domain size and geometry. Furthermore, GrainNN not only reconstructs the quantities of interest but also can be pointwise accurate. In our numerical experiments, we use a polycrystalline phase field method to both generate the training data and assess GrainNN. For multiparametric, ensemble simulations with many grains, GrainNN can be orders of magnitude faster than phase field simulations, while delivering 5%–15% pointwise error. Additionally, this speedup includes the cost of the phase field simulations for generating training data.

36 MATERIALS SCIENCE↗