Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “concept drift”

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 19 records

Adaptive Learning for Concept Drift in Application Performance Modeling

Supervised learning is a promising approach for modeling the performance of applications running on large HPC systems. A key assumption in supervised learning is that the training and testing data are obtained under the same conditions. However, in production HPC systems these conditions might not hold because the conditions of the platform can change over time as a result of hardware degradation, hardware replacement, software upgrade, and configuration updates. These changes could alter the data distribution in a way that affects the accuracy of the predictive performance models and render them less useful; this phenomenon is referred to as concept drift. Ignoring concept drift can lead to suboptimal resource usage and decreased efficiency when those performance models are deployed for tuning and job scheduling in production systems. To address this issue, we propose a concept-drift-aware predictive modeling approach that comprises two components: (1) an online Bayesian changepoint detection method that can automatically identify the location of events that lead to concept drift in near-real time and (2) a moment-matching transformation inspired by transfer learning that converts the training data collected before the drift to be useful for retraining. We use application input/output performance data collected on Cori, a production supercomputing system at the National Energy Research Scientific Computing Center, to demonstrate the effectiveness of our approach. The results show that concept-drift-aware models obtain significant improvement in accuracy; the median absolute error of the best-performing Gaussian process regression improved by 58.8% when the proposed approaches were used.

adaptive learning↗

A Data Quality-Aware Framework to Reliably Forecast Photovoltaic Generation and Consumer Load for an Improved Resilience of Microgrids

Photovoltaic (PV) power and consumer load forecasting plays a critical role to ensure operational resilience of the electric grid. Most data-driven forecasting algorithms rely heavily on the continuous availability of good quality data for periodic training and validation. When deployed at the grid’s edge, prolonged disruptions to communications during extreme events degrade data quality. Factors such as missing observations, epistemic uncertainties, data drift, and concept drift are manifestations of data quality that impact the generalization of such field-deployed forecasting models. Currently, there exists no mechanism in the literature to dynamically switch between models under varying degrees of data quality as quantified by certain metrics for each factor highlighted above. This paper addresses this shortcoming by conceptually introducing a data qualityaware framework for reliable PV generation and consumer load forecasting. The framework’s design incorporates components of missing values, divergence tests, and continuous monitoring of generalization performance to detect changes in data quality caused by communications disruptions and trigger specific classes of forecasting models grouped under three use cases (UC1- UC3). As a first step towards validating this framework, real data collected from an actual field microgrid system is used to demonstrate the viability of the three use cases. Results show that the performance is the best in UC1 with an unadjusted R-square value of 0.954, followed by 0.939 for UC2 and 0.757 for UC3.

Sundararajan, Aditya↗

Measuring, Predicting and Visualizing Short-Term Change in Word Representation and Usage in VKontakte Social Network

Language in social media is extremely dynamic: new words emerge, trend and disappear, while the meaning of existing words can fluctuate over time. This work addresses several important tasks of visualizing and predicting short term text representation shift, i.e. the change in a word’s contextual semantics. We study the relationship between short-term concept drift and representation shift on a large social media corpus – VKontakte collected during the Russia-Ukraine crisis in 2014 – 2015. We visualize short-term representation shift for example keywords and build predictive models to forecast short-term shifts in meaning from previous meaning as well as from concept drift. We show that short-term representation shift can be accurately predicted up to several weeks in advance and that visualization provides insight into meaning change. Our approach can be used to explore and characterize specific aspects of the streaming corpus during crisis events and potentially improve other downstream classification tasks including real-time event forecasting in social media.

concept drift↗

Picosecond UV single photon detectors with lateral drift field: Concept and technologies

Group III–V semiconductor materials are being considered as a Si replacement for advanced logic devices for quite some time. Advances in III–V processing technologies, such as interface and surface passivation, large area deep submicron lithography with high-aspect ratio etching primarily driven by the metal-oxide-semiconductor field-effect transistor development can also be used for other applications. In this paper we will focus on photodetectors with the drift field parallel to the surface. We compare the proposed concept to the state-of-the-art Si-based technology and discuss requirements which need to be satisfied for such detectors to be used in a single photon counting mode in blue and ultraviolet spectral region with about 10 ps photon timing resolution essential for numerous applications ranging from high-energy physics to medical imaging.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Adaptive Framework for Classification and Novel Class Detection over Evolving Data Streams with Limited Labeled Data.

Most approaches to classifying evolving data streams either divide the stream of data into fixed-size chunks or use gradual forgetting to address the problems of infinite length and concept drift. Finding the fixed size of the chunks or choosing a forgetting rate without prior knowledge about time-scale of change is not a trivial task. As a result, these approaches suffer from a trade-off between performance and sensitivity. To address this problem, we present a framework which uses change detection techniques on the classifier performance to determine chunk boundaries dynamically. Though this framework exhibits good performance, it is heavily dependent on the availability of true labels of data instances. However, labeled data instances are scarce in realistic settings and not readily available. Therefore, we present a second framework which is unsupervised in nature, and exploits change detection on classifier confidence values to determine chunk boundaries dynamically. In this way, it avoids the use of labeled data while still addressing the problems of infinite length and concept drift. Moreover, both of our proposed frameworks address the concept evolution problem by detecting outliers having similar values for the attributes. We provide theoretical proof that our change detection method works better than other state-of-the-art approaches in this particular scenario. Results from experiments on various benchmark and synthetic data sets also show the efficiency of our proposed frameworks.

97 MATHEMATICS AND COMPUTING↗

A novel data-driven learning method for radar target detection in nonstationary environments

Most existing radar algorithms are developed under the assumption that the environment (clutter) is stationary. However, in practice, the characteristics of the clutter can vary enormously depending on the radar-operational scenarios. If unaccounted for, these nonstationary variabilities may drastically hinder the radar performance. Therefore, to overcome such shortcomings, we develop a data-driven method for target detection in nonstationary environments. In this method, the radar dynamically detects changes in the environment and adapts to these changes by learning the new statistical characteristics of the environment and by intelligibly updating its statistical detection algorithm. Specifically, we employ drift detection algorithms to detect changes in the environment; incremental learning, particularly learning under concept drift algorithms, to learn the new statistical characteristics of the environment from the new radar data that become available in batches over a period of time. The newly learned environment characteristics are then integrated in the detection algorithm. Furthermore, we use Monte Carlo simulations to demonstrate that the developed method provides a significant improvement in the detection performance compared with detection techniques that are not aware of the environmental changes.

47 OTHER INSTRUMENTATION↗

Technical Design and Optimization of a HLW-Repository in the Gorleben Salt Dome including Detailed Design of the Sealing System - 13305

The preliminary safety analysis for a HLW repository at Gorleben, the potential repository site in Germany, takes into account an updated set of data on the amounts and types of expected heat generating waste, the documented results of the exploration of the Gorleben salt dome, and the 'Safety Requirements Governing the Final Disposal of Heat-Generating Radioactive Waste' as at 30 September 2010. A repository design was developed for two emplacement concepts (drift disposal and borehole disposal) mainly influenced by the thermal impact of the heat-generating waste on the host rock and taking into account mining constrains. According to the objective to create the conditions for a safe containment of the waste containers in the host rock, a closure concept consisting of backfilling and sealing measures was developed. The repository was designed in such a way that retrievability requirements can be met for all waste containers within the operating phase of the repository. In addition, it could be shown that sufficient measures for ensuring subcriticality are provided both during the operational and the post-closure phases of the repository. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Multistage depressed collector conceptual design for thin magnetically confined electron beams

The requirement of higher efficiency in high power microwave devices, such as traveling wave tubes and gyrotrons, guides scientific research to more advanced types of collector systems. First, a conceptual design approach of a multistage depressed collector for a sheet electron beam confined by a magnetic field is presented. The sorting of the electron trajectories, according to their initial kinetic energy, is based on the E × B drift concept. The optimization of the geometrical parameters is based on the analytical equations under several general assumptions. The analysis predicts very high levels of efficiency. Then, a design approach for the application of this type of collector to a gyrotron cylindrical hollow electron beam is also presented with very high levels of efficiency more than 80%.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Evaluating Model Robustness for Defect Identification and Classification in a Composite Aerostructure Material

Aircraft structures are required to have a high level of quality to satisfy their need for light weight, efficient flight, and withstanding high loads over their lifespan. These aerostructures are typically made from a composite material due to their good tensile strength and resistance to compression. To ensure their structural integrity, the composite material requires inspection for common flaws such as porosity, delaminations, voids, foreign object debris, and other defects. Ultrasonic testing (UT) is a popular non-destructive inspection (NDI) technique used for effectively evaluating the composite material. Current inspection methods rely heavily on human experience and are extremely time consuming. Therefore, there is a need for the development of techniques to reduce the manual inspection time. This work compares the performance of different deep learning-based methods in the identification and classification of defects. Deep learning has shown great promise in numerous fields, and we show its effectiveness in the evaluation of the composite aerostructure material. The methods developed here are both highly reliable with a top recall value of 98.64% as well as extremely efficient requiring an average of 4 s during the inferencing stage to evaluate new composites. Lastly, we investigate model robustness to concept drift by measuring its performance over time.

36 MATERIALS SCIENCE↗

Drifting Recovery Base Concept for GEO Derelict Object Capture

Over 250 objects hover within 6 m/sec of perfect geostationary orbit. Over half of these objects lie within 0.1 m/sec of the GEO velocity. Such items have 62% of the total velocity required to achieve Earth gravitational escape. A conceptual architecture is proposed to clean this orbit area of derelict objects while providing a demonstration mission for many facets of future asteroid mining operations. These near-GEO objects average nearly 2000kg each, consisting of (typically functioning) power systems, batteries, and large quantities of components and raw aerospace-grade refined materials. Such a demonstration collection system could capture, collect and remove all GEO derelict objects in an international effort to create a depot of components and of aerospace-grade raw materials--with a total mass greater than that of the International Space Station--as a space scrap depot ready for transfer to lunar or Mars orbit, using only two heavy-lift launches and 2-3 years of on-orbit operations.

Bacon, John B.↗

Safety Concept and Safety Assessment for a HLW Repository in Bedded Salt in Germany - 18253

The German site selection act for a repository for vitrified HLW and SNF demands the consideration of various host rock formations for deep geological disposal. One of the eligible and presently investigated host rock formations is bedded rock salt. Within the framework of the national R and D project KOSINA the scientific bases for the design and safety assessment of a repository for HLW and SNF in bedded salt formations were provided. The research findings presented in this paper include the formulation of a safety concept and safety demonstration concept as well as exemplary radiological consequences analysis for a generic repository in bedded rock salt formations. The safety concept describes in a verbal, argument-based way how the natural conditions, processes and technical measures at large yield a status of safety. Fundamental principle of the safety concept is the concentration and containment of the radioactive and other pollutants in the waste in the containment providing rock zone (CRZ), which ensures isolation from the biosphere. Based on the principle of safe containment inside the CRZ guiding principles, design objectives and strategic measures are derived. The safety demonstration concept describes the means, such as analyses and arguments, which are used in the safety case to demonstrate the safety of a repository system in bedded salt on the basis of the safety concept and its guiding principles and design objectives. The decisive elements are: the demonstration of integrity of the geological barrier, the demonstration of integrity of the geotechnical barriers, the scenario analysis, and the evaluation of the identified scenarios. These elements are complemented in the safety demonstration by concepts on how to consider criticality, non-radiological protection goals and operational safety. The radiological consequences analysis primarily consists of the calculation of the radiological safety indicator for different test cases and scenarios for the various repository design and waste emplacement options. Within the KOSINA project the disposal concepts of drift disposal, horizontal and vertical borehole disposals were investigated in detail by means of computational flow and transport modelling of the repository mine (near field) and assessment of the radiological consequences (biosphere model). The applied software package RepoTREND is introduced; the modelling concept and employed radiological indicators are explained. Examples of the modelling results for the reference cases of the drift disposal repository concept for bedded salt will be given, demonstrating the capabilities of the model and the applicability to close-to-reality repository systems for deep geological disposal. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

A LArTPC with Vertical Drift for the DUNE Far Detector

The Far Detector of the Deep Underground Neutrino Experiment (DUNE) will be a large LArdetector located at a baseline of 1300 kilometers, 1.5 km deep underground. It is planned to bemade up of four modules, each with a total mass of 17 kt of LAr, at least the first two of which willconsist in Liquid Argon Time Projection Chambers (LArTPCs). To prove the feasibility of theLArTPC technology at the kiloton scale, the ProtoDUNE Single and Dual-Phase detectors wereconstructed and operated at the CERN Neutrino facility.This document describes the Vertical Drift detector concept, which is proposed to instrument thesecond DUNE module. It consists of a TPC where the electrons drift vertically, with a cathodesuspended at mid-height, towards anodes placed at the bottom and top of the detector. The anodeswould be made out of printed PCBs instead of wires, and the new disposition would allow the topreadout electronics to be accessible during the lifetime of the experiment. An enhanced photo-detection system is also proposed, with the photo-sensors placed on the cryostat walls and thecathode, but posing a challenge in terms of power and signal transmission. Studies are ongoingboth to overcome the technical challenges of this new design and to finalize the concept.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Experiments and Modeling to Support Field Test Design

Salt repositories continue to be a primary concept for disposal of heat-generating nuclear waste (HGNW). For the past several years, efforts have focused on the in-drift disposal concept (e.g. Hansen and Leigh 2011; Stauffer et al. 2013; Jordan et al. 2015a,b,c; Bourret et al. 2016; Hansen et al. 2016; Bourret et al. 2017; Johnson et al. 2017a,b; Kuhlman et al. 2017). Ongoing research efforts aim to improve understanding of the behavior of the salt backfill surrounding HGNW sources, the responses of drift walls to heat sources, and the chemistry and chemical evolution of brine in such environments.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Uncertainty guided online ensemble for non-stationary data streams in fusion science

Machine Learning (ML) is poised to play a pivotal role in the development and operation of next-generation fusion devices. Fusion data shows non-stationary behavior with distribution drifts, resulted by both experimental evolution and machine wear-and-tear. ML models assume stationary distribution and fail to maintain performance when encountered with such non-stationary data streams. Online learning techniques have been leveraged in other domains, however it has been largely unexplored for fusion applications. In this paper, we investigate online learning for continuous adaptation to drifting data streams in the prediction of Toroidal Field (TF) coils deflection at the DIII-D fusion facility. We further address the short-term performance degradation inherent to standard online learning, which arises because ground truth is unavailable at prediction time. To mitigate this issue, we propose an uncertainty-guided online ensemble framework. The method leverages the Deep Gaussian Process Approximation (DGPA) for calibrated uncertainty estimation and uses these uncertainty measures to guide a meta-algorithm that aggregates predictions from learners trained over different historical horizons. Our results show that online learning reduces prediction error by 80% compared to a static model. The online ensemble and the proposed uncertainty-guided ensemble further reduce error by approximately 6%, and 10% respectively, relative to standard single-model online learning, while also providing calibrated uncertainty estimates to support operational decision-making.

AI↗