Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Support vector machine”

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 487 records · Page 27
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

Decentralized control of units in smart grids for the support of renewable energy supply

Due to the significant environmental impact of power production from fossil fuels and nuclear fission, future energy systems will increasingly rely on distributed and renewable energy sources (RES). The electrical feed-in from photovoltaic (PV) systems and wind energy converters (WEC) varies greatly both over short and long time periods (from minutes to seasons), and (not only) by this effect the supply of electrical power from RES and the demand for electrical power are not per se matching. In addition, with a growing share of generation capacity especially in distribution grids, the top-down paradigm of electricity distribution is gradually replaced by a bottom-up power supply. This altogether leads to new problems regarding the safe and reliable operation of power grids. In order to address these challenges, the notion of Smart Grids has been introduced. The inherent flexibilities, i.e. the set of feasible power schedules, of distributed power units have to be controlled in order to support demand–supply matching as well as stable grid operation. Controllable power units are e.g. combined heat and power plants, power storage systems such as batteries, and flexible power consumers such as heat pumps. By controlling the flexibilities of these units we are particularly able to optimize the local utilization of RES feed-in in a given power grid by integrating both supply and demand management measures with special respect to the electrical infrastructure. In this context, decentralized systems, autonomous agents and the concept of self-organizing systems will become key elements of the ICT based control of power units. In this contribution, we first show how a decentralized load management system for battery charging/discharging of electrical vehicles (EVs) can increase the locally used share of supply from PV systems in a low voltage grid. For a reliable demand side management of large sets of appliances, dynamic clustering of these appliances into uniformly controlled appliance sets is necessary. We introduce a method for self-organized clustering for this purpose and show how control of such clusters can affect load peaks in distribution grids. Subsequently, we give a short overview on how we are going to expand the idea of self-organized clusters of units into creating a virtual control center for dynamic virtual power plants (DVPP) offering products at a power market. For an efficient organization of DVPPs, the flexibilities of units have to be represented in a compact and easy to use manner. We give an introduction how the problem of representing a set of possibly 10{sup 100} feasible schedules can be solved by a machine-learning approach. In summary, this article provides an overall impression how we use agent based control techniques and methods of self-organization to support the further integration of distributed and renewable energy sources into power grids and energy markets. - Highlights: • Distributed load management for electrical vehicles supports local supply from PV. • Appliances can self-organize into so called virtual appliances for load control. • Dynamic VPPs can be controlled by extensively decentralized control centers. • Flexibilities of units can efficiently be represented by support-vector descriptions.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2026

CONTROL AND DATA ACQUISITION IN A CYBER-PHYSICAL MIDSTREAM TESTBED

This thesis presents the development of a laboratory-scale cyber–physical midstream pipeline testbed designed to address this gap and support research in industrial control systems security. The platform integrates pumps, valves, sensors, programmable logic controllers (PLCs), and a human–machine interface (HMI) to emulate the monitoring and control architecture of real pipeline operations. The physical process is implemented as a closed-loop liquid circulation system designed to replicate flow behavior characteristic of midstream pipeline infrastructure. The testbed enables real-time data acquisition of key process variables, including flow rate and pressure facilitating the generation of datasets representative of normal pipeline operation. A threat model encompassing common ICS attack vectors was developed, including sensor spoofing, command injection, false data injection, denial-of-service attacks, and relay manipulation. Multiple attack scenarios were implemented and evaluated to demonstrate how cyber intrusions targeting sensors, actuators, networks, and software propagate into measurable physical consequences in pipeline flow and pressure. The developed platform serves as a practical, cost-effective environment for experimentation, education, and future cybersecurity research in midstream pipeline systems.

42 ENGINEERING↗
DOE OSTI2026

Online LIBS–ML Framework for Dynamic Characterization of Heterogeneous Waste-Derived Gasification Feedstocks

LIBS−ML framework for real time feedstock characterization during continuous conveyor transport Heterogeneous waste derived feedstocks (e.g., waste coal, biomass and blends) introduce rapid variability in heating value and ash chemistry that affect gasifier operation, yet conventional laboratory characterization techniques are too slow to support proactive control. To address this gap, this study reports on an online, in situ, dynamic characterization framework that couple’s laser-induced breakdown spectroscopy (LIBS) with leakage safe machine learning (ML) regression to deliver real time, decision quality predictions of gasifier relevant properties. A controlled sample matrix spanning two different waste coals, two different biomasses, and engineered blends under two particle size conditions were constructed and benchmarked using standardized laboratory analyses for proximate/ultimate properties and ash composition. LIBS spectra were acquired dynamically as material flowed on a conveyor belt, using high energy 1064 nm laser ablation and shot averaging to improve repeatability and precision. Supervised regression models (multi layer perceptron (MLP) /artificial neural network (ANN), random forest (RF), and support vector regression (SVR)) and an optimized weighted ensemble were trained on emission line feature sets using nested cross validation with Bayesian hyperparameter tuning and validated against an independent hold out set. The proposed LIBS−ML workflow achieves near laboratory predictive fidelity across parametric targets (including higher heating value (HHV), ash content, fixed carbon, sulfur, major ash forming oxides, and initial deformation temperature (IDT)), with the weighted ensemble providing a robust default predictor under dynamic measurement conditions. These results demonstrate a practical pathway for real time feedstock characterization that can enable feedforward adjustments and more resilient gasifier operation for variable quality waste derived fuels.

Biomass↗
DOE OSTI2023

Envisioning urban environments resilient to vector-borne diseases: a protocol to study dengue in Vietnam

Transmitted primarily by Aedes aegypti (Ae. aegypti) and Aedes albopictus (Ae. albopictus), arboviral diseases pose a major global public health threat. Dengue, chikungunya, and zika are increasingly prevalent in Southeast Asia. Among other arboviruses, dengue and zika are becoming more common in Central and South America. Given human encroachment into previously uninhabited, often deforested areas, to provide new housing in regions of population expansion, conceptualizing built urban environments in a novel way is urgently needed to safeguard against the growing climate change-driven threat of vector-borne diseases. By understanding the spread from a One Health perspective, enhanced control and prevention can be achieved. This is particularly important considering that climate change is likely to significantly impact the persistence of ponded water where mosquitoes breed due to increasing temperature and shifting rainfall patterns with regard to magnitude, duration, frequency, and season. Models can incorporate aquatic mosquito stages and adult spatial dynamics when habitats are heterogeneously available, thereby including dispersal and susceptible-exposed-infected-recovered (SEIR) epidemiology. Coupled with human population distribution (density, locations), atmospheric conditions (air temperature, precipitation), and hydrological conditions (soil moisture distribution, ponding persistence in topographic depressions), modeling has improved predictive ability for infection rates. However, it has not informed interventional approaches from an urban environment perspective which considers the role of ponds/lakes that support green spaces, the density of population that enables rapid spread of disease, and varying micro-habitats for various mosquito stages under climate change. Here, for an example of dengue in Vietnam, a preventive and predictive approach to design resilient urban environments is proposed, which uses data from rapidly expanding metropolitan communities to learn continually. This protocol deploys computational approaches including simulation and machine learning/artificial intelligence, underpinned by surveillance and medical data for validation and adaptive learning. Its application may best inform urban planning in low-middle income countries in tropical zones where arboviral pathogens are prevalent.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

Investigation of Portable Event-Based Monte Carlo Transport Using the NVIDIA Thrust Library

Power consumption considerations are driving future high performance computing platforms toward many-core computing architectures. Los Alamos National Laboratory's Trinity machine, available in 2016, will use both Intel Xeon Haswell processors and Intel Xeon Phi Knights Landing many integrated core (MIC) architecture coprocessors. Lawrence Livermore National Laboratory's Sierra machine, available in 2018, will use an IBM PowerPC architecture along with Nvidia graphics processing unit (GPU) architecture accelerators. These different advanced architectures make the computing landscape in upcoming years complex. Traditional approaches to Monte Carlo transport do not work efficiently on these new computing platforms. MIC architectures require vectorization to operate efficiently, and vectorization is difficult to achieve in Monte Carlo transport. GPU architectures require additional code to explicitly use the hardware, requiring significant code changes or hardware specific branches in the source code. A significant challenge for Monte Carlo transport projects is to simultaneously support within a single source code base efficient simulations for both the current generation of architectures and the different advanced computing architectures. In order to address these challenges, two important changes are typically required: a new algorithmic approach for solving Monte Carlo transport, and explicit use of hardware specific software. In this paper, we describe initial research investigations of an event-based Monte Carlo transport algorithm implemented using the Nvidia Thrust library on a GPU for a Monte Carlo test code. The event-based algorithm targets many-core architectures by increasing SIMD (single instruction multiple data) parallelism, while Thrust potentially provides portable performance by allowing one source code base to compile code targeted for both CPUs and GPUs. We described preliminary investigations of portable event-based Monte Carlo algorithms implemented using the Nvidia Thrust library in a research Monte Carlo test code. We found that an explicit CUDA implementation of an event-based Monte Carlo algorithm performed significantly more efficiently than a Thrust implementation on GPU platforms, most likely as a result of additional flexibility in access to different memory spaces on the GPU. Additionally, we showed that on GPU platforms and at large enough problem sizes the event-based implementations perform more efficiently than the serial history-based implementation running on the host CPU. While investigating this problem, we also discovered that the performance of the event-based algorithm is affected by what tallies are being used. A zonal scalar flux tally requires atomic operations that significantly impacted the performance of the code, in some cases producing slowdowns instead of speedups. We decided to remove the tally in order to focus on the effectiveness of the event-based algorithm. Future work will be required to research more effective ways of handling such tallies. Additionally, we would like to consider new ways for optimizing both the Thrust and CUDA versions, in order to see how much performance we have yet to achieve. The potential trade-off between portability and performance was demonstrated in this investigation. Thrust provides both CPU and GPU versions of the code in one code base, but it does so at a cost. We discovered approximately a factor of 2-5 performance difference between Thrust and CUDA on each GPU platform we tested. Thrust provides a tool to access the GPUs with less effort and specialization, but it does so by giving up fine grained control where extra performance can be found for this application. Future work will be required to determine whether this performance differential between Thrust and CUDA can be reduced for event-based Monte Carlo transport. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

NSTXU Diagnostic Disruption Dynamic Loading Represented by Response Spectra

This article presents the results of transient dynamic simulations of loads due to disruption eddy currents on the NSTXU vacuum vessel. Dynamic loading at diagnostic mounting locations is expressed as response spectra derived from the time history results of the dynamic structural simulations of a variety of disruption scenarios. The disruption simulations draw on a history of the project assessments of worst case disruptions for specific components. Major efforts to assess disruption loading have included the vacuum vessel which is the major structural support for the machine, as well as the passive plates (PPs), high harmonic fast wave (HHFW) antenna, and centerstack casing. Each one of these efforts included transient electromagnetic simulations producing time-dependent eddy current Lorentz loads (and in some cases halo loads) which then were applied to time-dependent structural dynamic analyses intended to obtain the proper dynamic amplification factors. In some instances, the EM model and structural model were identical allowing direct transfer of EM forces to the structural model. In other cases, the EM and structural model were not identical and the vector potential (VP) transfer method was used. The results files from these analyses were available (or re-run) to post process in ANSYS Classic time history postprocessor. In conclusion, the ANSYS command is used to create response spectra from time history data at desired points on the vessel.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Assessment of Machine Learning for Ultrasonic Nondestructive Evaluation of Alkali–Silica Reaction in Concrete

Alkali–silica reaction (ASR) is a type of material degradation in concrete structures that leads to concrete cracking and rebar corrosion, thereby reducing the material’s structural integrity and the overall structure’s lifetime and raising safety concerns. Ultrasonic nondestructive evaluation (NDE) has been proven to be a valuable technique for assessing concrete properties and monitoring ASR progression in concrete. However, the deployment and analysis of ultrasonic NDE and its data requires specialized expertise, often relying on the engineer’s subjective interpretation. With the surge in computational power, artificial intelligence (AI) and machine learning (ML) algorithms have become popular in automating NDE data analysis. Various industrial sectors are increasingly adopting ML algorithms for NDE data analysis with a growing emphasis on AI–assisted automation. Regulatory agencies are also preparing for this technological shift, anticipating corresponding revisions in standards. Thus, there is an urgent need to identify the capabilities and limitations of current ML technologies for the evaluation of concrete material properties and damage status. Furthermore, the effects of various factors on ML model performance must be thoroughly investigated. The study summarized herein evaluated the effectiveness of two ML models (i.e., support vector regression (SVR) and deep neural network (DNN)) in predicting concrete material damage induced by ASR based on the long-term ultrasonic monitoring data. Four distinct concrete specimens were cast with artificially induced ASR, and over a period exceeding 500 days, ultrasonic signals and expansion data were continuously collected. For the SVR model, wave velocity and 12 other wave features were extracted from the ultrasonic signals, with 6 out of 13 features selected as input for the model. Different combinations of training and testing datasets were designed to explore factors influencing prediction performance, including the range of data within training and testing sets, in addition to various signal preprocessing methodologies. These findings suggest the importance of using a training dataset with a broader data range compared with testing datasets for improved model performance alongside consistent signal preprocessing across datasets.

36 MATERIALS SCIENCE↗
DOE OSTI2013

Efficient checkpointing schemes for depletion perturbation solutions on memory-limited architectures

We describe a methodology for decreasing the memory footprint and machine I/O load associated with the need to access a forward solution during an adjoint solve. Specifically, we are interested in the depletion perturbation equations, where terms in the adjoint Bateman and transport equations depend on the forward flux solution. Checkpointing is the procedure of storing snapshots of the forward solution to disk and using these snapshots to recompute the parts of the forward solution that are necessary for the adjoint solve. For large problems, however, the storage cost of just a few copies of an angular flux vector can exceed the available RAM on the host machine. We propose a methodology that does not checkpoint the angular flux vector; instead, we write and store converged source moments, which are typically of a much lower dimension than the angular flux solution. This reduces the memory footprint and I/O load of the problem, but requires that we perform single sweeps to reconstruct flux vectors on demand. We argue that this trade-off is exactly the kind of algorithm that will scale on advanced, memory-limited architectures. We analyze the cost, in terms of FLOPS and memory footprint, of five checkpointing schemes. We also provide computational results that support the analysis and show that the memory-for-work trade off does improve time to solution. (authors)

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Supervised learning with word embeddings derived from PubMed captures latent knowledge about protein kinases and cancer

Abstract Inhibiting protein kinases (PKs) that cause cancers has been an important topic in cancer therapy for years. So far, almost 8% of >530 PKs have been targeted by FDA-approved medications, and around 150 protein kinase inhibitors (PKIs) have been tested in clinical trials. We present an approach based on natural language processing and machine learning to investigate the relations between PKs and cancers, predicting PKs whose inhibition would be efficacious to treat a certain cancer. Our approach represents PKs and cancers as semantically meaningful 100-dimensional vectors based on word and concept neighborhoods in PubMed abstracts. We use information about phase I-IV trials in ClinicalTrials.gov to construct a training set for random forest classification. Our results with historical data show that associations between PKs and specific cancers can be predicted years in advance with good accuracy. Our tool can be used to predict the relevance of inhibiting PKs for specific cancers and to support the design of well-focused clinical trials to discover novel PKIs for cancer therapy.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2019

Multi-Formalism Modeling for Disaster Resilience, Forecasting, and Response

Disaster response preparedness, risk mitigation, and effective post-event response are critical for national security, public health, and economic integrity. For areas impacted by extreme events, timely and accurate impact assessments of lifeline infrastructures are critical. Traditional disaster management and situational awareness platforms are not capable of handling the massive influx of available information from traditional and non-traditional sources. Next-generation analytic assessment approaches, where use of dynamically sourced, heterogeneous data with near real-time availability are integrated into a continually-updated analytics framework. The objective of this research was to develop a foundational multi-formalism modeling platform to integrate novel data sources for use in improving situational awareness after a disaster event. Specifically, to 1) provide an architecture for machine-learning based data fusion various multi-sensor datasets; 2) classify, geolocate, and temporally bin novel data sources for use in spatially-enabled machine-learning based data fusion; and 3) derive confidence-based observation data with machine-learning based prediction to fill spatial and temporal observation gaps. This project developed a reusable cloud-based pipeline architecture for ingesting, storing, and analyzing relevant data, executing a range of deep-learning models, and visualize the information geographically. The pipeline development was informed by previous PNNL investments (StreamKit) and built upon leading-edge technologies consisting of: 1) a raw datastore; 2) a data enrichment pipeline with custom and third-party analytics; 3) enriched data storage; and 4) a custom UI to display the enriched data. Machine learning/deep learning models were selected and tested to classify social media images for relevancy and presence of flood waters. Our custom trained relevance models were run on a validation dataset and the overall accuracy of the relevance classifiers ranged from 65% to 90%. We found that convolutional neural network (CNN) models were both faster and more accurate than more traditional machine learning models. We applied the relevance classifier to remove non-relevant images and then undertook a limited training of CNN models to detect flood on ~1,800 labeled images. The testing accuracy of the flood detection models were ~88%. To improve the geographic location of both social media and oblique aerial images we tested a multi-step workflow built on a pre-trained deep-learning model. To identify similarity features from test and reference scenes, we applied the Xception and Faiss models to a set of test images with inaccurate spatial locations, as well and precisely-located reference scenes from the general area. Matching reference scenes to the test scenes with the closest similarity index had limited success in improving the precision of test scene location and additional methods have been identified for testing. Data fusion between authoritative geographic data (satellite imaging) and non-authoritative crowdsourced data is an underlying theme throughout this research and methods are tested bring these data to a common geographic form through feature matching and transformations. The automated extraction of supporting information for disaster response from oblique aerial images or ground-level social media images is accomplished through semantic image segmentation and transformation. The non-authoritative data are used to validate the authoritative data collections and processing and optionally, t+0 flood models. In addition, the common form geographic data are combined with other static landscape features to develop spatially-explicit training vectors used in deep generative statistical inference models to develop complex non-linear relationships in static and dynamic data and ultimately derive a spatially continuous and probabilistic assessment of flood damage with the event domain.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2019

Creating Feature Representations of Antibody-Antigen Complexes for Fast Binding Prediction with Machine Learning

Antibody-antigen interactions underlie the function of the immune system. Numerous computational tools exist that aim to predict important quantities associated with antibodyantigen interactions, such as binding energy. Efficient and accurate prediction of such quantities would greatly accelerate vaccine (antigen) and therapeutic antibody design. Current methods and tools, however, rely on varying degrees of computationally expensive steps that limit their use on problems that require many predictions, and moreover, no single method produces consistently accurate estimates. In this project, we determined the feasibility of creating feature representations that concisely describe interacting pairs of antibody and antigen as rapidly computable numerical vectors, given the antibody-antigen co-complex. Such a featurization will allow the use of a statistical machine learning model that jointly models the results of the disparate existing computational tools available. In turn, this will greatly enhance their combined predictive performance. We developed a feature representation consisting of 86 descriptors of the antigen-antibody interface which is then reduced into a lower dimensional representation using a neural network dimensionality reduction step. We have demonstrated that this feature representation serves its intended purpose of rapidly describing antigen-antibody pairs for the task of binding strength prediction. While this feasibility study has proven that such a feature representation is feasible, additional research is likely improve its ability to support various tasks, including binding strength prediction.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Neural chaos: A spectral stochastic neural operator

Building surrogate models for operators with uncertainty quantification capabilities is essential for many engineering applications where randomness–such as variability in material properties, boundary conditions, and initial conditions–is unavoidable. Polynomial Chaos Expansion (PCE) is widely recognized as a go-to method for constructing stochastic surrogates in both intrusive and non-intrusive ways, and it has recently been used in the context of operator learning. However, its application becomes challenging for complex or high-dimensional processes, as achieving accuracy requires higher-order polynomials, which can increase computational demand and/or the risk of overfitting. Furthermore, PCE requires specialized treatments to manage random variables that are not independent, and these treatments may be problem-dependent or may fail with increasing complexity. Here, in this work, we adopt the same formalism as the spectral expansion used in PCE; however, we replace the classical polynomial basis functions with neural network (NN) basis functions to leverage their expressivity. To achieve this, we propose an algorithm that identifies NN-parameterized basis functions in a purely data-driven manner, without any prior assumptions about the joint distribution of the random variables involved, whether independent or dependent, or about their marginal distributions. The proposed algorithm identifies each NN-parameterized basis function sequentially, ensuring they are orthogonal with respect to the data distribution. The basis functions are constructed directly on the joint stochastic variables without requiring a tensor product structure or assuming independence of the random variables. This approach may offer greater flexibility for complex stochastic models, while simplifying implementation compared to the tensor product structures typically used in PCE to handle random vectors. This is particularly advantageous given the current state of open-source packages, where building and training neural networks can be done with just a few lines of code and extensive community support. We demonstrate the effectiveness of the proposed scheme through several numerical examples of varying complexity and provide comparisons with classical PCE.

Polynomial chaos expansion↗
DOE OSTI2018

Temporal ensemble learning of univariate methods for short term load forecasting

Short term load forecasting (STLF) is a fundamental component of demand response programs in smart grids. Recent literature has focused on complex neural networks and deep learning models for solving STLF. While these models work well for load forecasting with complex non-linear relationships, they have been shown to be less effective than simpler univariate models for STLF problems with under a 6 hours horizon. Given the lack of multivariate data (such as temperature) in many practical datasets, we need better univariate prediction models for STLF. By partitioning the dataset by temporal features, we develop a novel ensemble learning method that exploits daily seasonality in electricity consumption data to improve accuracy of popular univariate models. We train an ensemble of models from the dataset partitions. We develop a variety of methods, including Ridge Regression, to increase the robustness of the ensemble prediction. To show the effectiveness of our approach, we perform detailed evaluation using an aggregated user electricity consumption dataset collected by the Los Angeles Department of Water and Power (LADWP). We select four popular prediction algorithms in literature for our experiments, including Kernel Regression (KR), Support Vector Regression (SVR) and neural network approaches. We compare the performance of these algorithms applying our ensemble approach to training only one single model. Our approach leads to an 11.2% decrease in mean absolute percentage error (MAPE) and 21.3% decrease in root mean squared error (RMSE) over the single model approach for KR, and a 30% and 32.4% decrease in MAPE and RMSE respectively for SVR. These ensemble models also outperform the neural network approaches.

14 SOLAR ENERGY↗
DOE OSTI2026

Cybersecurity Center for Offshore Wind Energy (Final Project Report)

This project establishes a Cybersecurity Center for Offshore Wind Energy with the objective of designing and operating a cyber-physical testbed for wind energy farms (WEFs) that enables comprehensive cybersecurity research. The testbed incorporates a Supervisory Control and Data Acquisition (SCADA) system connected to turbine models via industrial-grade programmable logic controllers (PLCs) and remote terminal units (RTUs). It supports side-channel data acquisition, implementation and analysis of various cyberattack scenarios, and development of attack detection, mitigation, and best-practice guidance tailored to wind energy systems. During the project, the team expanded the number and fidelity of mathematical turbine models (MTMs), integrated these models with SCADA infrastructure, and deployed a scaled physical turbine and associated sensors. High-resolution operational and side-channel data streams were collected and used to refine machine-learning (ML)-based attack detection systems and to extend the WindCRAFT framework to multi-turbine threat scenarios. The project demonstrated a realistic, scalable environment for evaluating cyber threats, validated attack detection approaches using enriched datasets, and identified new multi-turbine and inter-turbine communication attack vectors. The resulting testbed, models, and security mechanisms provide a foundation for ongoing R&D and deployment of cyber-resilient offshore wind energy systems.

17 WIND ENERGY↗
DOE OSTI2022

Development of Short-Term Forecasting Models Using Plant Asset Data and Feature Selection

Nuclear power plants collect and store large volumes of heterogeneous data from various components and systems. With recent advances in machine learning (ML) techniques, these data can be leveraged to develop diagnostic and short-term forecasting models to better predict future equipment condition. Maintenance operations can then be planned in advance whenever degraded performance is predicted, thus resulting in fewer unplanned outages and the optimization of maintenance activities. This enables lower maintenance costs and improves the overall economics of nuclear power. This paper focuses on developing a short-term forecasting process that leverages a feature selection process to distill large volumes of heterogeneous data and predict specific equipment parameters. A variety of feature selection methods, including Shapley Additive Explanations (SHAP) and variance inflation factor (VIF), were used to select the optimal features as inputs for three ML methods: long short-term memory (LSTM) networks, support vector regression (SVR), and random forest (RF). Each combination of model and input features was used to predict a pump bearing temperature both 1 and 24 hours in advance, based on actual plant system data. The optimal inputs for the LSTM and SVR were selected using the SHAP values, while the optimal input for the RF consisted solely of the response variable itself. Each model produced similar 1-hour-ahead predictions, with root mean square errors (RMSEs) of roughly 0.006. For the 24-hour-ahead predictions, differences could be seen between LSTM, SVR, and RF, as reflected by model performances of 0.036 +- 0.014, 0.0026 +- 0, and 0.063 +- 0.004 RMSE, respectively. As big data and continuous online monitoring become more widely available, the proposed feature selection process can be used for many applications beyond the prediction of process parameters within nuclear infrastructure.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

Analytics-at-scale of Sensor Data for Digital Monitoring in Nuclear Plants (3 rd Annual Report)

Nuclear power plants collect and store large volumes of heterogeneous data from various components and systems. With recent advances in machine learning (ML) techniques, these data can be leveraged to develop diagnostic and short-term forecasting models to better predict future equipment condition. Maintenance operations can then be planned in advance whenever degraded performance is predicted, thus resulting in fewer unplanned outages and the optimization of maintenance activities. This enables lower maintenance costs and improves the overall economics of nuclear power. This report primarily focuses on developing a short-term forecasting process that leverages a feature selection process to distill large volumes of heterogeneous data and predict specific equipment parameters. A variety of feature selection methods, including Shapley Additive Explanations (SHAP) and variance inflation factor (VIF), were used to select the optimal features as inputs for three ML methods: long short-term memory (LSTM) networks, support vector regression (SVR), and random forest (RF). Each combination of model and input features was used to predict a pump bearing temperature both 1 and 24 hours in advance, based on actual plant system data. The optimal inputs for the LSTM and SVR were selected using the SHAP values, while the optimal input for the RF consisted solely of the response variable itself. Each model produced similar 1-hour-ahead predictions, with root mean square errors (RMSEs) of roughly 0.006. For the 24-hour-ahead predictions, differences could be seen between LSTM, SVR, and RF, as reflected by model performances of 0.036 ± 0.014, 0.0026 ± 0, and 0.063 ± 0.004 RMSE, respectively. As big data and continuous online monitoring become more widely available, the proposed feature selection process can be used for many applications beyond the prediction of process parameters within nuclear infrastructure. This report summarizes the Fiscal Year 2021 research progress encompassing the (1) data cleaning and feature selection necessary for ML applications; (2) development of short-term forecasting models to predict future plant process parameters for both single and multiple time steps ahead; and (3) validation of the feature selection methods and short-term forecasting models given new data from different systems.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

FLASH: FPGA-Accelerated Smart Switches with GCN Case Study

Some communication switches, e.g., the Mellanox SHArP and those in the IBM BlueGene clusters, are augmented to process packets at the application level with fixed-function collectives. This approach, however, lacks flexibility, which limits their applicability in diverse and dynamic workloads. Recently, a new type of programmable packet processor, which uses high-level languages, e.g., P4, has emerged as possible candidates. P4-based switches, however, fall short in certain applications, including machine learning, where capabilities not currently supported by P4 are needed. These include more complex calculation, such as sparse computation and fused multiply-accumulate, data-intensive floating point operations, data reuse, and significant memory. The problem addressed here is that such a switch augmentation needs to support: a large amount of state, significant flexible compute capability, and ease of programming, all while maintaining full functionality, including ensuring high throughput, and demonstrating utility. In this work, we propose a programmable look-aside-type accelerator that can be embedded into, or attached to, existing communication switch pipelines and that is capable of processing packets at line-rate. The proposed in-switch accelerator is based on mixing an ISA (subset of RISC-V instructions) with dataflow graphs (found in CGRAs). To augment performance, vector instructions are also supported. To facilitate usability, we have developed a complete toolchain to compile user-provided C/C++ codes to appropriate back-end instructions for configuring the accelerator. While this approach is flexible enough to support various workloads, in this paper, we consider Graph Convolutional Networks (GCNs) as a case study. Experimental results show that this approach considerably improves the performance of distributed GCN applications.

Haghi, Pouya↗
DOE OSTI2024

Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units

Atomic structure prediction and associated property calculations are the bedrock of chemical physics. Since high-fidelity ab initio modeling techniques for computing the structure and properties can be prohibitively expensive, this motivates the development of machine-learning (ML) models that make these predictions more efficiently. Training graph neural networks over large atomistic databases introduces unique computational challenges such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs such as social networks. We demonstrate a novel hardware-software co-design approach to scale up the training of atomistic graph neural networks (GNN) for structure and property prediction. First, to eliminate redundant computation and memory associated with alternative padding techniques and to improve throughput via minimizing communication, we formulate the effective coalescing of the batches of variable-size atomistic graphs as the bin packing problem and introduce a hardware-agnostic algorithm to pack these batches. In addition, we propose hardware-specific optimizations including a planner and vectorization for the gather-scatter operations targeted for Graphcore’s Intelligence Processing Unit (IPU), as well as model-specific optimizations such as merged communication collectives and optimized softplus. Putting these all together, we demonstrate the effectiveness of the proposed co-design approach by providing an implementation of a well-established atomistic GNN on the Graphcore IPUs. We evaluate the training performance on multiple atomistic graph databases with varying degrees of graph counts, sizes and sparsity. Here, we demonstrate that such a co-design approach can reduce the training time of atomistic GNNs and can improve the performance by up to 1.5× compared to the baseline implementation of the model on the IPUs. Additionally, we compare our IPU implementation with a Nvidia GPU-based implementation and show that our atomistic GNN implementation on the IPUs can run 1.8× faster on average compared to the execution time on the GPUs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗