Search NASA⌕ Search

SEARCH · Search NASA

Results for “unsupervised method”

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 181 records · Page 10

Learning Stochastic Parametric Differentiable Predictive Control Policies

We present a scalable unsupervised learning-based method for obtaining explicit control policies for model predictive control problems for stochastic linear systems with additive uncertainties subject to nonlinear chance constraints. We call the proposed method stochastic parametric differentiable predictive control (SP-DPC), which extends the recently proposed deterministic DPC policy optimization algorithm. We formulate the SP-DPC as a deterministic approximation to the stochastic parametric constrained optimal control problem via independent sampling of the problem's parameters and uncertainties. This formulation allows us to directly compute the policy gradients via automatic differentiation of the problem's value function, evaluated over sampled parameters and uncertainties. In particular, the computed expectation of the problem's value function is backpropagated through the finite-time closed-loop system rollouts parametrized by a known nominal system dynamics model and neural control policy. We also provide theoretical probabilistic guarantees on closed-loop stability and chance constraints satisfaction for systems controlled by learned neural policies. We demonstrate the computational efficiency and scalability of the proposed policy optimization algorithm in three numerical examples, including systems with a large number of states or subject to nonlinear constraints.

Drgona, Jan↗

Unsupervised Clustering-Based Short-Term Solar Forecasting

Solar forecasting accuracy is highly affected by weather conditions, therefore, weather awareness forecasting models are expected to improve the forecasting performance. However, it may not be available or reliable to classify different forecasting tasks by only using predefined meteorological weather categorization. In this paper, an unsupervised clustering-based (UC-based) solar forecasting method is developed for short-term (1-h-ahead) global horizontal irradiance (GHI) forecasting. This UC-based method consists of three parts: GHI time series unsupervised clustering, pattern recognition, and UC-based forecasting. The daily GHI time series is first clustered by an Optimized Cross-validated ClUsteRing (OCCUR) method, which determines the optimal number of clusters and best clustering results. Then, support vector machine pattern recognition is adopted to recognize the category of a certain day using the first four hours' data in the forecasting stage. GHI forecasts are generated by the most suitable models in different clusters, which are built by a two-layer machine learning based multi-model (M3) forecasting framework. The developed UC-M3 method is validated by using 1-year of data with 13 solar features from three information sources. Numerical results show that 1) UC-based models outperform non-UC (all-in-one) models with the same M3 architecture by approximately 20%; and 2) M3-based models also outperform the single-algorithm machine learning models by approximately 20%.

14 SOLAR ENERGY↗

Second-order statistical bootstrap for the uncertainty quantification of time-temperature-superposition analysis

Time-temperature superposition (TTS), which for decades has been a powerful method for long-term prediction from accelerated aging data, involves rigid-shifting isotherms in logarithmic time to produce a single master prediction curve. For simple thermo-rheological properties that accurately follow the TTS principle, the shifts can be easily determined, even manually by the eye. However, for many properties of interest, where the principle is obeyed only approximately, or the data is noisy, it is imperative to develop objective shifting techniques along with reliable uncertainty bounds. This work analyzes in detail the method of arclength-minimization as an unsupervised algorithm to determining optimum shifts and demonstrates that the method is nearly unbiased for all practical datasets with a variety of noise distributions. Moreover, if averaged over with-replacement (bootstrap) resamples, the predicted shifts follow a normal distribution, a fact that can be used to construct confidence interval for the master curve through a second-order bootstrap procedure.

36 MATERIALS SCIENCE↗

Binning sequences using very sparse labels within a metagenome

Background: In metagenomic studies, a process called binning is necessary to assign contigs that belong to multiple species to their respective phylogenetic groups. Most of the current methods of binning, such as BLAST, k-mer and PhyloPythia, involve assigning sequence fragments by comparing sequence similarity or sequence composition with already-sequenced genomes that are still far from comprehensive. We propose a semi-supervised seeding method for binning that does not depend on knowledge of completed genomes. Instead, it extracts the flanking sequences of highly conserved 16S rRNA from the metagenome and uses them as seeds (labels) to assign other reads based on their compositional similarity. Results: The proposed seeding method is implemented on an unsupervised Growing Self-Organising Map (GSOM), and called Seeded GSOM (S-GSOM). We compared it with four well-known semi-supervised learning methods in a preliminary test, separating random-length prokaryotic sequence fragments sampled from the NCBI genome database. We identified the flanking sequences of the highly conserved 16S rRNA as suitable seeds that could be used to group the sequence fragments according to their species. S-GSOM showed superior performance compared to the semi-supervised methods tested. Additionally, S-GSOM may also be used to visually identify some species that do not have seeds. The proposed method was then applied to simulated metagenomic datasets using two different confidence threshold settings and compared with PhyloPythia, k-mer and BLAST. At the reference taxonomic level Order, S-GSOM outperformed all k-mer and BLAST results and showed comparable results with PhyloPythia for each of the corresponding confidence settings, where S-GSOM performed better than PhyloPythia in the ≥ 10 reads datasets and comparable in the ≥ 8 kb benchmark tests. Conclusion: In the task of binning using semi-supervised learning methods, results indicate S-GSOM to be the best of the methods tested. Most importantly, the proposed method does not require knowledge from known genomes and uses only very few labels (one per species is sufficient in most cases), which are extracted from the metagenome itself. These advantages make it a very attractive binning method. SGSOM outperformed the binning methods that depend on already-sequenced genomes, and compares well to the current most advanced binning method, PhyloPythia.

59 BASIC BIOLOGICAL SCIENCES↗

Fuzzy self-learning control for magnetic servo system

It is known that an effective control system is the key condition for successful implementation of high-performance magnetic servo systems. Major issues to design such control systems are nonlinearity; unmodeled dynamics, such as secondary effects for copper resistance, stray fields, and saturation; and that disturbance rejection for the load effect reacts directly on the servo system without transmission elements. One typical approach to design control systems under these conditions is a special type of nonlinear feedback called gain scheduling. It accommodates linear regulators whose parameters are changed as a function of operating conditions in a preprogrammed way. In this paper, an on-line learning fuzzy control strategy is proposed. To inherit the wealth of linear control design, the relations between linear feedback and fuzzy logic controllers have been established. The exercise of engineering axioms of linear control design is thus transformed into tuning of appropriate fuzzy parameters. Furthermore, fuzzy logic control brings the domain of candidate control laws from linear into nonlinear, and brings new prospects into design of the local controllers. On the other hand, a self-learning scheme is utilized to automatically tune the fuzzy rule base. It is based on network learning infrastructure; statistical approximation to assign credit; animal learning method to update the reinforcement map with a fast learning rate; and temporal difference predictive scheme to optimize the control laws. Different from supervised and statistical unsupervised learning schemes, the proposed method learns on-line from past experience and information from the process and forms a rule base of an FLC system from randomly assigned initial control rules.

Tarn, J. H.↗

A review of computing-based automated fault detection and diagnosis of heating, ventilation and air conditioning systems

We report faults in Heating, Ventilation, and Air Conditioning (HVAC) systems of buildings result in significant energy waste in building operation. With fast-growing sensing data availability and advancement in computing, computational modeling has demonstrated strong capability to detect and diagnose HVAC system faults, hence, ensuring efficient building operation. This paper comprehensively reviews the state-of-the-art computing-based fault detection and diagnosis (FDD) for HVAC systems. Overall, the reviewed computing-based FDD methods are classified as two major approaches: knowledge-based and data-driven approaches. We then identify multiple important topics, including data availability, training data size, data quality, approach generality, capability, interpretability, and required modeling efforts, along with corresponding metrics to summarize the most updated FDD development. Generally, the knowledge-based approaches are further divided as physics-based modeling, Diagnostic Bayesian Network, and performance indicator-based methods while data-driven approaches include supervised learning, unsupervised learning, and regression and statistics-based methods. State-of-the-art FDD development, remaining challenges, and future research directions are further discussed to push forward FDD in practice. Availability of fault data, capability of existing methods to deal with complex fault situations (such as simultaneous faults), modeling interpretability for data-driven methods, and required engineering efforts for physics-based methods are identified as remaining challenges in FDD development. Improving modeling fidelity and reducing modeling efforts are essential for applying physics-based methods in real buildings. Meanwhile, addressing fault data availability, increasing algorithm adaptability, and handling multiple faults are essential to further enhance the applicability of data-driven FDD approaches.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

An interpretation of a geologic map of Fannin County, Texas, prepared by ADP techniques from ERTS MSS data

Multispectral scanner data, collected by ERTS-1 from an area in northeast Texas underlain by structurally simple cretaceous sedimentary rocks, were analyzed using several automatic data processing techniques. Training areas for the computer analysis were selected in three ways using: (1) an unsupervised classifier, (2) the unsupervised classifier to further refine training areas, and (3) available ground information. The first two methods utilize the capability of the unsupervised classifier to the unsupervised classifier to group resolution elements with similar spectral properties. With the imagery produced by these procedures, several geologic features can be identified. The most easily recognizable features are streams and other water bodies. The approximate location of contacts between several rock units can be mapped.

Henderson, J. A., Jr.↗

Signature extension: An approach to operational multispectral surveys

Two data processing techniques were suggested as applicable to the large area survey problem. One approach was to use unsupervised classification (clustering) techniques. Investigation of this method showed that since the method did nothing to reduce the signal variability, the use of this method would be very time consuming and possibly inaccurate as well. The conclusion is that unsupervised classification techniques of themselves are not a solution to the large area survey problem. The other method investigated was the use of signature extension techniques. Such techniques function by normalizing the data to some reference condition. Thus signatures from an isolated area could be used to process large quantities of data. In this manner, ground information requirements and computer training are minimized. Several signature extension techniques were tested. The best of these allowed signatures to be extended between data sets collected four days and 80 miles apart with an average accuracy of better than 90%.

Nalepka, R. F.↗

A Method for Landsat and Sentinel 2 (HLS) BRDF Normalization

The Harmonized Landsat/Sentinel-2 (HLS) project aims to generate a seamless surface reflectance product by combining observations from USGS/NASA Landsat-8 and ESA Sentinel-2 remote sensing satellites. These satellites’ sampling characteristics provide nearly constant observation geometry and low illumination variation through the scene. However, the illumination variation throughout the year impacts the surface reflectance by producing higher values for low solar zenith angles and lower reflectance for large zenith angles. In this work, we present a model to derive the bidirectional reflectance distribution function (BRDF) normalization and apply it to the HLS product at 30m spatial resolution. It is based on the BRDF parameters estimated from the MODerate Resolution Imaging Spectroradiometer (MODIS) surface reflectance product (M{O,Y}D09) at 1 km spatial resolution using the VJB method (Vermote et al., 2009). Unsupervised classification (segmentation) of HLS images is used to disaggregate the BRDF parameters to the HLS spatial resolution and to build a BRDF parameters database at HLS scale. We first test the proposed BRDF normalization for different solar zenith angles over two homogeneous sites, in particular one desert and one Peruvian Amazon forest. The proposed method reduces both the correlation with the solar zenith angle and the coefficient of variation (CV) of the reflectance time series in the red and near infrared bands to 4% in forest and keeps a low CV of 3% to 4% for the deserts. Additionally, we assess the impact of the view zenith angle (VZA) in an area of the Brazilian Amazon forest close to the equator, where impact of the angular variation is stronger because it occurs in the principal plane. The directional reflectance shows a strong dependency with the VZA. The current HLS BRDF correction reduces this dependency but still shows an under-correction, especially in the near infrared, while the proposed method shows no dependency with the view angles. We also evaluate the BRDF parameters using field surface albedo measurements as a reference over seven different sites of the US surface radiation budget observing network (SURFRAD) and five sites of the Australian OzFlux network.

Belen Franch↗

Unsupervised Deep Learning Can Identify Physical Parameters

Parametrizing a problem is one of the biggest obstacles toward the application of physical methods to novel systems. In this research, we show that deep learning methods can be used to accelerate this parametrization process. We use an unsupervised learning algorithm, a variational autoencoder, to extract meaningful physical parameters from unlabeled time-series generated from classical physical systems. For example, given time series data of projectile motion and the instruction to represent the data in a small number of unspecified parameters, the model can discover, without further direction, the concepts of velocity and gravity. Using this method, one can uncover information about the physics of a system, even when the relevant physical parameters are unknown to the experimenter. This has obvious applications to novel systems in which parameters are not known a priori. Moreover, errors in the reconstruction of the parameters can be used to reveal meaningful info rmation about state transitions, which can be conceptualized as a generalization of similar unsupervised learning methods used to detect phase transitions in the thermodynamics of particle systems to systems that evolve in the time domain. While our proof-of-concept applications are modest, this method shows potential for making progress in a wide array of problems both in physics and outside of the field.

Veillette, John↗

Data-driven analysis of ultrasonic pressure tube inspection data - 220

Pressure tubes are critical components of the CANDU reactors and other pressurized heavy water type reactors, as they contain the nuclear fuel and the coolant. Manufacturing flaws, as well as defects developed during the in-service operation, can lead to coolant leakage and can potentially damage the reactor. The current inspection process of these flaws is based on manually analyzing ultrasonic data received from multiple probes during planned, statutory outages. Recent advances on ultrasonic inspection tools enable the provision of high resolution data of significantly large volumes. This is highlighting the need for an efficient autonomous signal analysis process. Typically, the automation of ultrasonic inspection data analysis is approached by knowledge-based or supervised data-driven methods. This work proposes an unsupervised data-driven framework that requires no explicit rules, nor individually labeled signals. The framework follows a two-stage clustering procedure that utilizes the DBSCAN density-based clustering algorithm and aims to provide decision support for the assessment of potential defects in a robust and consistent way. Nevertheless, verified defect dimensions are essential in order to assess the results and train the framework for unseen defects. Initial results of the implementation are presented and discussed, with the method showing promise as a means of assessing ultrasonic inspection data. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Characterization of Extreme Hydroclimate Events in Earth System Models using ML/AI

Focal Area(s): (1) We put forward the concepts of data assimilation enabled by machine learning, AI, and advanced methods including experimental/network design/optimization and unsupervised learning applied to downscale information within Earth System Models (ESMs). (2) We discuss predictive modeling through the use of AI techniques and other tools to design a prediction system comprising of a hierarchy of models (e.g., AI-driven model/component/parameterization selection) to improve the characterization of extreme hydroclimate events in ESMs. The AI-based models will run five-six order of magnitude times faster, yet will provide similar accuracy, allowing us to provide range bounds on uncertainty faster and thus enabling faster extreme event identification. (3) Further, we consider the insight gleaned from complex data (observed/simulated) using AI, big data analytics, and other advanced methods, including explainable AI and physics- or knowledge-guided AI to improve the characterization of extreme hydroclimate events in ESMs.

54 ENVIRONMENTAL SCIENCES↗

Unsupervised Framework to Monitor Lake Dynamics

A method of reducing processing time when assigning geographic areas to land cover labels using satellite sensor values includes a processor receiving a feature value for each pixel in a time series of frames of satellite sensor values, each frame containing multiple pixels and each frame covering a same geographic location. For each sub-area of the geographic location, the sub-area is assigned to one of at least three land cover labels. The processor determines a fraction function for a first sub-area assigned to a first land cover label. The sub-areas that were assigned to the first land cover label are reassigned to one of the second land cover label and the third land cover label based on the fraction functions of the sub-areas.

Boriah, Shyam↗

Ultrafast Bragg coherent diffraction imaging of epitaxial thin films using deep complex-valued neural networks

Abstract Domain wall structures form spontaneously due to epitaxial misfit during thin film growth. Imaging the dynamics of domains and domain walls at ultrafast timescales can provide fundamental clues to features that impact electrical transport in electronic devices. Recently, deep learning based methods showed promising phase retrieval (PR) performance, allowing intensity-only measurements to be transformed into snapshot real space images. While the Fourier imaging model involves complex-valued quantities, most existing deep learning based methods solve the PR problem with real-valued based models, where the connection between amplitude and phase is ignored. To this end, we involve complex numbers operation in the neural network to preserve the amplitude and phase connection. Therefore, we employ the complex-valued neural network for solving the PR problem and evaluate it on Bragg coherent diffraction data streams collected from an epitaxial La 2-x Sr x CuO 4 (LSCO) thin film using an X-ray Free Electron Laser (XFEL). Our proposed complex-valued neural network based approach outperforms the traditional real-valued neural network methods in both supervised and unsupervised learning manner. Phase domains are also observed from the LSCO thin film at an ultrafast timescale using the complex-valued neural network.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Unsupervised learning of ferroic variants from atomically resolved STEM images

An approach for the analysis of atomically resolved scanning transmission electron microscopy data with multiple ferroic variants in the presence of imaging non-idealities and chemical variabilities based on a rotationally invariant variational autoencoder (rVAE) is presented. We show that an optimal local descriptor for the analysis is a sub-image centered at specific atomic units, since materials and microscope distortions preclude the use of an ideal lattice as a reference point. The applicability of unsupervised clustering and dimensionality reduction methods is explored and is shown to produce clusters dominated by chemical and microscope effects, with a large number of classes required to establish the presence of rotational variants. Comparatively, the rVAE allows extraction of the angle corresponding to the orientation of ferroic variants explicitly, enabling straightforward identification of the ferroic variants as regions with constant or smoothly changing latent variables and sharp orientational changes. This approach allows further exploration of the chemical variability by separating the rotational degrees of freedom via rVAE and searching for remaining variability in the system. The code used in this article is available at https://github.com/saimani5/ferroelectric_domains_rVAE.

30 DIRECT ENERGY CONVERSION↗

On-line object feature extraction for multispectral scene representation

A new on-line unsupervised object-feature extraction method is presented that reduces the complexity and costs associated with the analysis of the multispectral image data and data transmission, storage, archival and distribution. The ambiguity in the object detection process can be reduced if the spatial dependencies, which exist among the adjacent pixels, are intelligently incorporated into the decision making process. The unity relation was defined that must exist among the pixels of an object. Automatic Multispectral Image Compaction Algorithm (AMICA) uses the within object pixel-feature gradient vector as a valuable contextual information to construct the object's features, which preserve the class separability information within the data. For on-line object extraction the path-hypothesis and the basic mathematical tools for its realization are introduced in terms of a specific similarity measure and adjacency relation. AMICA is applied to several sets of real image data, and the performance and reliability of features is evaluated.

Ghassemian, Hassan↗

Clonal Selection Based Artificial Immune System for Generalized Pattern Recognition

The last two decades has seen a rapid increase in the application of AIS (Artificial Immune Systems) modeled after the human immune system to a wide range of areas including network intrusion detection, job shop scheduling, classification, pattern recognition, and robot control. JPL (Jet Propulsion Laboratory) has developed an integrated pattern recognition/classification system called AISLE (Artificial Immune System for Learning and Exploration) based on biologically inspired models of B-cell dynamics in the immune system. When used for unsupervised or supervised classification, the method scales linearly with the number of dimensions, has performance that is relatively independent of the total size of the dataset, and has been shown to perform as well as traditional clustering methods. When used for pattern recognition, the method efficiently isolates the appropriate matches in the data set. The paper presents the underlying structure of AISLE and the results from a number of experimental studies.

pattern recognition↗

GeoThermalCloud: Machine Learning for Geothermal Resource Exploration

Geothermal is a renewable energy source that can provide reliable and flexible electricity generation for the world. In the past decade, the U.S. Geological Survey's resource assessments, Play Fairway Analyses (PFA), and GeoVision report by the U.S. Department of Energy's Geothermal Technologies Office provided insights on enormous untapped potential for geothermal energy to contribute to the U.S. domestic energy needs. The past studies identified that geothermal resources without surface expression (e.g., blind/hidden hydrothermal systems) comprise a huge potential. These blind systems can significantly increase power generation. But a primary challenge is locating and quantifying these hidden resources, which do not have any thermal manifestations on the surface. PFA has successfully identified some blind systems in the western USA (e.g., specific locations in the Great Basin region within Nevada). However, a comprehensive search for these blind systems can be time-consuming, expensive, and resource-intensive with a low probability of success. Accelerated discovery of these blind resources is needed with growing energy needs and higher chances of exploration success. Recent advances in machine learning (ML) have shown promise in shortening the timeline for this discovery. This paper presents a novel ML-based methodology for geothermal exploration towards PFA applications. Our methodology is provided through our open-source ML framework called GeoThermalCloud \url{https://github.com/SmartTensors/GeoThermalCloud.jl}. GeoThermalCloud uses a series of unsupervised, supervised, and physics-informed ML methods available in SmartTensors AI platform \url{https://github.com/SmartTensors}. Here, the presented analyses are performed using our unsupervised ML algorithm called NMF$k$, which is available in the SmartTensors AI platform. Our ML algorithm facilitates the discovery of new phenomena, hidden patterns, and mechanisms that helps us to make informed decisions. Moreover, the GeoThermalCloud enhances the collected PFA data and discovers signatures representative of geothermal resources. Through GeoThermalCloud, we were able to identify hidden patterns in the geothermal field data needed for the efficient discovery of blind systems. Crucial geothermal signatures often overlooked in traditional PFA are extracted using GeoThermalCloud and analyzed by the subject matter experts to provide ML-enhanced PFA, which is informative for efficient exploration. We applied our ML methodology on various open-source geothermal datasets within the U.S. (some of these are collected by past PFA work), and the results provide valuable insights on resource types within those explored regions. This ML-enhanced workflow makes GeoThermalCloud attractive for the geothermal community to improve existing datasets and extract valuable information often unnoticed during geothermal exploration.

machine learning (ML), geothermal energy↗