Search NASA⌕ Search

SEARCH · Search NASA

Results for “fast machine learning”

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 145 records · Page 8

MaLTESE: Large-Scale Simulation-Driven Machine Learning for Transient Driving Cycles

Optimal engine operation during a transient driving cycle is the key to achieving greater fuel economy, engine efficiency, and reduced emissions. In order to achieve continuously optimal engine operation, engine calibration methods use a combination of static correlations obtained from dynamometer tests for steady-state operating points and road and/or track performance data. As the parameter space of control variables, design variable constraints, and objective functions increases, the cost and duration for optimal calibration become prohibitively large. In order to reduce the number of dynamometer tests required for calibrating modern engines, a large-scale simulation-driven machine learning approach is presented in this work. A parallel, fast, robust, physics-based reduced-order engine simulator is used to obtain performance and emission characteristics of engines over a wide range of control parameters under various transient driving conditions (drive cycles). We scale the simulation up to 3,906 nodes of the Theta supercomputer at the Argonne Leadership Computing Facility to generate data required to train a machine learning model. The trained model is then used to predict various engine parameters of interest, and the results are compared with those predicted by the engine simulator. Our results show that a deepneural-network-based surrogate model achieves high accuracy: Pearson product-moment correlation values larger than 0.99 and mean absolute percentage error within 1.07% for various engine parameters such as exhaust temperature, exhaust pressure, nitric oxide, and engine torque. Once trained, the deep-neural-network-based surrogate model is fast for inference: it requires about 16 mu s for predicting the engine performance and emissions for a single design configuration compared with about 0.5 s per configuration with the engine simulator. Moreover, we demonstrate that transfer learning and retraining can be leveraged to incrementally retrain the surrogate model to cope with new configurations that fall outside the training data space.

deep learning↗

A review on recent machine learning applications for imaging mass spectrometry studies

Imaging mass spectrometry (IMS) is a powerful analytical technique widely used in biology, chemistry, and materials science fields that continue to expand. IMS provides a qualitative compositional analysis and spatial mapping with high chemical specificity. The spatial mapping information can be 2D or 3D depending on the analysis technique employed. Due to the combination of complex mass spectra coupled with spatial information, large high-dimensional datasets (hyperspectral) are often produced. Therefore, the use of automated computational methods for an exploratory analysis is highly beneficial. The fast-paced development of artificial intelligence (AI) and machine learning (ML) tools has received significant attention in recent years. These tools, in principle, can enable the unification of data collection and analysis into a single pipeline to make sampling and analysis decisions on the go. There are various ML approaches that have been applied to IMS data over the last decade. Here, in this review, we discuss recent examples of the common unsupervised (principal component analysis, non-negative matrix factorization, k-means clustering, uniform manifold approximation and projection), supervised (random forest, logistic regression, XGboost, support vector machine), and other methods applied to various IMS datasets in the past five years. The information from this review will be useful for specialists from both IMS and ML fields since it summarizes current and representative studies of computational ML-based exploratory methods for IMS.

47 OTHER INSTRUMENTATION↗

Safe Reinforcement Learning for Emergency Load Shedding of Power Systems

The paradigm shift in the electric power grid necessitates a revisit of existing control methods to ensure the grid’s security and resilience. In particular, the increased uncertainties and rapidly changing operational conditions in power systems have revealed outstanding issues in terms of either speed, adaptiveness, or scalability of the existing control methods for power systems. On the other hand, the availability of massive real-time data can provide a clearer picture of what is happening in the grid. Recently, deep reinforcement learning (RL) has been regarded and adopted as a promising approach leveraging massive data for fast and adaptive grid control. However, like most existing machine learning (ML)- based control techniques, RL control usually cannot guarantee the safety of the power systems. In this paper, we introduce a novel method for safe RL-based load shedding of power systems that can enhance the safe voltage recovery of the electric power grid after experiencing faults. Numerical simulation on the IEEE 39-bus testcase is performed to demonstrate the effectiveness of the proposed safe RL emergency control, as well as its adaptive capability to faults not seen in the training.

reinforcement learning, emergency control, power g↗

Optimizing Performance on Trinity Utilizing Machine Learning, Proxy Applications and Scheduling Priorities

The sheer number of nodes continues to increase in today’s supercomputers, the first half of Trinity alone contains more than 9400 compute nodes. Since the speed of today’s clusters are limited by the slowest nodes, it more important than ever to identify slow nodes, improve their performance if it can be done, and assure minimal usage of slower nodes during performance critical runs. This is an ongoing maintenance task that occurs on a regular basis and, therefore, it is important to minimize the impact upon its users by assessing and addressing slow performing nodes and mitigating their consequences while minimizing down time. These issues can be solved, in large part, through a systematic application of fast running hardware assessment tests, the application of Machine Learning, and making use of performance data to increase efficiency of large clusters. Proxy applications utilizing both MPI and OpenMP were developed to produce data as a substitute for long runtime applications to evaluate node performance. Machine learning is applied to identify underperforming nodes, and policies are being discussed to both minimize the impact of underperforming nodes and increase the efficiency of the system. In this paper, I will describe the process used to produce quickly performing proxy tests, consider various methods to isolate the outliers, and produce ordered lists for use in scheduling to accomplish this task.

97 MATHEMATICS AND COMPUTING↗

A Gaussian Process Enhancement to Linear Parameter Varying Models

Simulation and analysis for modern engineering systems now routinely requires the merging of multiple disciplines, physical-domains, time-scales, and data sets — all at ever increasing levels. These capabilities are especially needed in the domain of Advanced Air Mobility, where rapidly emerging vehicle designs are significantly more complex, while having to be both cost-effective and safe. To meet these engineering challenges, machine learning methods are an attractive option for merging models and data across multiple areas while providing uncertainty quantification and maintaining computational efficiency. This paper examines the use of Gaussian process machine learning to generalize and enhance the commonly used class of quasi-Linear Parameter Varying models for fast full-envelope simulation while also supporting control system design and analysis with model uncertainty. Gaussian process machine learning is selected because it: can fuse multiple data sets, enables an easy trade-off between data fitting and smoothing, provides model uncertainty quantification, scales well with increasing complexity, and does not generally require starting from a large training data set. To demonstrate the benefits of the approach, a robust stability analysis with Gaussian process uncertainty is shown for a NASA reference design of an electric quad-rotor air-taxi concept vehicle with motor parameter uncertainty.

Gaussian Process↗

Charged Wellbore Casing Controlled Source Electromagnetics (CWC-CSEM) for Reservoir Imaging and Monitoring (Final report)

This project addresses the needs of the U.S. Department of Energy (DOE) to develop advanced monitoring technologies and protocols to track the fate of subsurface carbon dioxide (CO2) plumes for carbon storage. Specifically, the project seeks to develop and test a unique and novel system of technologies consisting of electromagnetic data acquisition, coupled multiphysics imaging, and reservoir model enhancement to understand the migration and long-term distribution of CO2 in the subsurface. The overarching objective is to develop an integrated approach for long term monitoring of carbon storage. The two main components of the project include the methodology development and the test of the method at a field site. The methodology component consists of 1) developing the field procedure and protocol for collecting time-lapse controlled-source electromagnetic (CSEM) data with source electric current injected into the subsurface through wellbore casings; 2) building of background 3D electrical conductivity utilizing multiple sources of data such as supplemental surface transient EM (TEM) surveys, well-logs, and seismic structural information, for enhancing CSEM signal from reservoir depths; and 3) coupled multiphysics simulations and inversion of CSEM data constrained by production data and by structural information from seismic imaging of the reservoir and overlying formations. The testing component used the field site of Bell Creek Oil Field, which served both as a field laboratory for the method development as well as a test site to evaluate the CSEM signal strengths and the methodology developed in this research project. We have accomplished all the proposed tasks and developed the methodology as planned. These include the procedure for time-lapse CSEM data acquisition, data processing techniques, integration with 3D conductivity model building, fast reservoir simulation for history matching using machine learning, and interpreting CSEM data with coupling to the reservoir modeling. Collectively, the outcome of these tasks form a coherent workflow that can be applied to monitor dedicated carbon storage in saline reservoirs. The testing component evaluated the applicability and limitations of the method, and concluded that the method would be ideal for monitoring dedicated carbon storage sites utilizing saline reservoirs.

54 ENVIRONMENTAL SCIENCES↗

OLCF’s Advanced Computing Ecosystem (ACE): FY25 Update for Ongoing Efforts

The advent of widespread use of artificial intelligence (AI) and machine learning (ML) models in science, coupled with fast data production rates of scientific instruments strain the traditional batch-oriented high-performance computing (HPC) environment. As scientific exploration continues to require more data and faster processing and analysis, new emerging technologies and capabilities to enable cross-facility and time-sensitive workflows are required for seamless integration of HPC and experimental facilities. The Advanced Computing Ecosystem (ACE) is a strategic initiative within the Oak Ridge Leadership Computing Facility (OLCF) established in 2024 to support the development of cutting-edge technologies to advance computational research and infrastructure at OLCF and across the Department of Energy (DOE). Several DOE initiatives are spearheading the evolution of the scientific landscape by blurring facility boundaries and connecting the user facilities to advance scientific capabilities and ensure energy dominance. The DOE Integrated Research Infrastructure (IRI) program is one example that is laying a foundation to support complex cross-facility workflows. The IRI program aims to integrate diverse computational resources, data infrastructures, and scientific instruments to facilitate collaboration and accelerate scientific discovery. The Interconnected Science Ecosystem (INTERSECT) initiative at Oak Ridge National Laboratory (ORNL) is another example that aims to revolutionize scientific research through AI-driven, interconnected autonomous laboratories and research facilities. Finally, the American Science Cloud (AmSC), recently announced in the “One Big Beautiful Bill”, aims to leverage prior infrastructure efforts of the IRI and automation and AI efforts of INTERSECT (and others) to build a federated, AI-augmented AmSC platform to unify the DOE’s computing, experimental, and data resources to catalyze scientific innovation.

97 MATHEMATICS AND COMPUTING↗

Equation Discovery Using Fast Function Extraction: a Deterministic Symbolic Regression Approach

Advances in machine learning (ML) coupled with increased computational power have enabled identification of patterns in data extracted from complex systems. ML algorithms are actively being sought in recovering physical models or mathematical equations from data. This is a highly valuable technique where models cannot be built using physical reasoning alone. In this paper, we investigate the application of fast function extraction (FFX), a fast, scalable, deterministic symbolic regression algorithm to recover partial differential equations (PDEs). FFX identifies active bases among a huge set of candidate basis functions and their corresponding coefficients from recorded snapshot data. This approach uses a sparsity-promoting technique from compressive sensing and sparse optimization called pathwise regularized learning to perform feature selection and parameter estimation. Furthermore, it recovers several models of varying complexity (number of basis terms). FFX finally filters out many identified models using non-dominated sorting and forms a Pareto front consisting of optimal models with respect to minimizing complexity and test accuracy. Numerical experiments are carried out to recover several ubiquitous PDEs such as wave and heat equations among linear PDEs and Burgers, Korteweg–de Vries (KdV), and Kawahara equations among higher-order nonlinear PDEs. Additional simulations are conducted on the same PDEs under noisy conditions to test the robustness of the proposed approach.

42 ENGINEERING↗

Understanding Mixed Precision GEMM with MPGemmFI: Insights into Fault Resilience

Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). Thus, one of the critical features of machine-learning-specific accelerators such as NVIDIA Tensor Cores, AMD Matrix Cores, and Google TPUs is the support of mixed-precision enabled GEMM. For DNN models, lower-precision FP data formats and computation offer acceptable correctness but significant performance, area, and memory footprint improvement. While promising, the mixed-precision computation on error resilience remains unexplored. To this end, we develop a fault injection framework that systematically injects fault into the mixed-precision computation results. We investigate how the faults affect the accuracy of machine learning applications. Based on the characteristics of error resilience, we offer lightweight error detection and correction solutions that significantly improve the overall model accuracy by 75% if the models experience hardware faults. The solutions can be efficiently integrated into the accelerator's pipelines.

Fang, Bo↗

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↗

Automatic Loss Factor Modeling and Attribution on Unlabeled PV Energy Data

We present a novel approach for modeling the loss factors of photovoltaic power generation systems (PV systems). This method is a white-box machine learning model built on convex optimization that is fast, interpretable, and auditable. It takes as an input the measured daily energy produced by the system, over a multi-year period, and returns a multiplicative decomposition model of the daily energy signal and full attribution of the total energy loss to each feature. The methods section of this paper has two major components: (1) the description of the signal decomposition (SD) model, expressed in the SD framework, and (2) the attribution of total energy losses via Shapley values. We validate the method on synthetic and open-source data sets and compare to similar methods from the literature.

artificial intelligence↗

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↗

Photometric classification of Hyper Suprime-Cam transients using machine learning

Abstract The advancement of technology has resulted in a rapid increase in supernova (SN) discoveries. The Subaru/Hyper Suprime-Cam (HSC) transient survey, conducted from fall 2016 through spring 2017, yielded 1824 SN candidates. This gave rise to the need for fast type classification for spectroscopic follow-up and prompted us to develop a machine learning algorithm using a deep neural network with highway layers. This algorithm is trained by actual observed cadence and filter combinations such that we can directly input the observed data array without any interpretation. We tested our model with a dataset from the LSST classification challenge (Deep Drilling Field). Our classifier scores an area under the curve (AUC) of 0.996 for binary classification (SN Ia or non-SN Ia) and 95.3% accuracy for three-class classification (SN Ia, SN Ibc, or SN II). Application of our binary classification to HSC transient data yields an AUC score of 0.925. With two weeks of HSC data since the first detection, this classifier achieves 78.1% accuracy for binary classification, and the accuracy increases to 84.2% with the full dataset. This paper discusses the potential use of machine learning for SN type classification purposes.

Takahashi, Ichiro↗

Advancing set-conditional set generation: Diffusion models for fast simulation of reconstructed particles

The computational intensity of detector simulation and event reconstruction poses a significant difficulty for data analysis in collider experiments. This challenge inspires the continued development of machine learning techniques to serve as efficient surrogate models. We propose a fast emulation approach that combines simulation and reconstruction. In other words, a neural network generates a set of reconstructed objects conditioned on input particle sets. To make this possible, we advance set-conditional set generation with diffusion models. Using a realistic, generic, and public detector simulation and reconstruction package (COCOA), we show how diffusion models can accurately model the complex spectrum of reconstructed particles inside jets.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Characterization of mechanical discontinuities based on data-driven classification of compressional-wave travel times

Wave propagation and diffusive transport phenomena are influenced by the mechanical discontinuities in material. Here this study shows that certain bulk properties of the network of low-velocity mechanical discontinuities (e.g. air-filled cracks) in a material can be characterized by processing compressional-wave travel times using traditional data-driven classification techniques. To that end, we perform three tasks in chronological order: (1) use the discrete fracture network (DFN) method to create two-dimensional (2D) numerical models of crack-bearing material embedded with various types of low-velocity mechanical discontinuities, (2) use the fast marching method (FMM) to simulate the propagation of the wave/diffusion front from a single source through the 2D crack-bearing material to multiple receivers placed on the boundary of the material, and (3) train 9 data-driven classifiers to characterize the crack-bearing materials (i.e. bulk properties of the network of mechanical discontinuities in the crack-bearing material) by learning from the simulations of travel times detected by multiple receivers placed around the crack-bearing material. The classifiers identified the orientation, spatial distribution, and dispersion of the low-velocity mechanical discontinuities. Voting classifier performs the best among the 9 classifiers. For the characterization of bulk dispersion and distribution of discontinuities, the sensors located on the adjacent boundaries are more important; whereas for the characterization of bulk orientation of discontinuities, the sensors located on the opposite side are more important.

42 ENGINEERING↗

Jumpstart Opportunities to Unleash Leadership in Energy Storage (JOULES)

Current-generation Li-ion batteries with cobalt- and nickel-containing cathodes and graphite anodes are approaching performance and cost limits. In this program, 24M Technologies, Inc. (24M) is teaming with the Massachusetts Institute of Technology (MIT) and University of Michigan (UM) to develop low cost and fast charging sodium metal batteries with good low-temperature performance and high energy density, building upon previous work performed under ARPA-E programs. Key achievements include optimization of solid electrolyte and anode current collector, optimized cathode active materials, development of high-performance electrolyte formulations, and integration of these components into full cells. The cell design incorporates (1) an ultra-thick cathode (>9 mAh/cm 2 ) comprising advanced cobalt-free, sodium cathode active material, (2) advanced fast-charging electrolyte (up to 12 mS/cm) developed using machine learning and automated high-throughput screening technology by UM, and (3) ceramic modified separator that enable smooth Na transport and deposition, developed at MIT, enabling a high-energy density anode-free configuration and maximizing the energy density of sodium batteries. The team has successfully combined these approaches to sodium chemistry and paved the way to meeting the fast-charging, high-energy density, and low-cost requirements of next-generation drone, electric vertical take-off and -landing, and electric vehicle batteries. Performance for anode-free sodium cells developed under this program is more powerful than the commercial Li-ion batteries. The final deliverable cell design has achieved over 300 Wh/kg and volumetric energy density above 800 Wh/L (Table 1). Additionally, the team has achieved over (1) a lifetime of 340 cycles, (2) 80% capacity retention at -20 °C (compared 25 °C), and (3) the ability to fast charge to 80% SOC in 20 minutes.

25 ENERGY STORAGE↗

Calibrating Bayesian generative machine learning for Bayesiamplification

Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

97 MATHEMATICS AND COMPUTING↗