Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “algorithm selection”

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 1,351 records · Page 75

Maximum a posteriori Ly α estimator (MAPLE): band power and covariance estimation of the 3D Ly α forest power spectrum

We present a novel maximum a posteriori estimator to jointly estimate band powers and the covariance of the three-dimensional power spectrum (P3D) of Ly $\alpha$ forest flux fluctuations, called MAPLE. Our Wiener-filter based algorithm reconstructs a window-deconvolved P3D in the presence of complex survey geometries typical for Ly $\alpha$ surveys that are sparsely sampled transverse to and densely sampled along the line of sight. We demonstrate our method on idealized Gaussian random fields with two selection functions: (i) a sparse sampling of 30 background sources per square degree designed to emulate the current Dark Energy Spectroscopic Instrument; (ii) a dense sampling of 900 background sources per square degree emulating the upcoming Prime Focus Spectrograph Galaxy Evolution Survey. Our proof-of-principle shows promise, especially since the algorithm can be extended to marginalize jointly over nuisance parameters and contaminants, i.e. offsets introduced by continuum fitting. Our code is implemented in JAX and is publicly available on GitHub.

79 ASTRONOMY AND ASTROPHYSICS↗

Using computer-extracted image features for modeling of error-making patterns in detection of mammographic masses among radiology residents

Purpose: Mammography is the most widely accepted and utilized screening modality for early breast cancer detection. Providing high quality mammography education to radiology trainees is essential, since excellent interpretation skills are needed to ensure the highest benefit of screening mammography for patients. The authors have previously proposed a computer-aided education system based on trainee models. Those models relate human-assessed image characteristics to trainee error. In this study, the authors propose to build trainee models that utilize features automatically extracted from images using computer vision algorithms to predict likelihood of missing each mass by the trainee. This computer vision-based approach to trainee modeling will allow for automatically searching large databases of mammograms in order to identify challenging cases for each trainee. Methods: The authors’ algorithm for predicting the likelihood of missing a mass consists of three steps. First, a mammogram is segmented into air, pectoral muscle, fatty tissue, dense tissue, and mass using automated segmentation algorithms. Second, 43 features are extracted using computer vision algorithms for each abnormality identified by experts. Third, error-making models (classifiers) are applied to predict the likelihood of trainees missing the abnormality based on the extracted features. The models are developed individually for each trainee using his/her previous reading data. The authors evaluated the predictive performance of the proposed algorithm using data from a reader study in which 10 subjects (7 residents and 3 novices) and 3 experts read 100 mammographic cases. Receiver operating characteristic (ROC) methodology was applied for the evaluation. Results: The average area under the ROC curve (AUC) of the error-making models for the task of predicting which masses will be detected and which will be missed was 0.607 (95% CI,0.564-0.650). This value was statistically significantly different from 0.5 (p < 0.0001). For the 7 residents only, the AUC performance of the models was 0.590 (95% CI,0.537-0.642) and was also significantly higher than 0.5 (p = 0.0009). Therefore, generally the authors’ models were able to predict which masses were detected and which were missed better than chance. Conclusions: The authors proposed an algorithm that was able to predict which masses will be detected and which will be missed by each individual trainee. This confirms existence of error-making patterns in the detection of masses among radiology trainees. Furthermore, the proposed methodology will allow for the optimized selection of difficult cases for the trainees in an automatic and efficient manner.

60 APPLIED LIFE SCIENCES↗

Compressing unstructured mesh data from simulations using machine learning

The amount of data output from a computer simulation has grown to terabytes and petabytes as increasingly complex simulations are being run on massively parallel systems. As we approach exaflop computing in the next decade, it is expected that the I/O subsystem will not be able to write out these large volumes of data. In this paper, we explore the use of machine learning to compress the data before it is written out. Despite the computational constraints that limit us to using very simple learning algorithms, our results show that machine learning is a viable option for compressing unstructured data. Furthermore, we demonstrate that by simply using a better sampling algorithm to generate the training set, we can obtain more accurate results compared to random sampling, but at no extra cost. Further, by carefully selecting and incorporating points with high prediction error, we can improve reconstruction accuracy without sacrificing the compression rate.

97 MATHEMATICS AND COMPUTING↗

Predicting image blur in proton radiography: comparisons between measurements and Monte Carlo simulations

Given the cost and lead-times involved in high-energy proton radiography, it is prudent to model proposed radiographic experiments to see if the images predicted would return useful information. We recently modified our raytracing transmission radiography modeling code HADES to perform simplified Monte Carlo simulations of the transport of protons in a proton radiography beamline. Beamline objects include the initial diffuser, vacuum magnetic fields, windows, angle-selecting collimators, and objects described as distorted 2D (planar or cylindrical) meshes or as distorted 3D hexahedral meshes. We present an overview of the algorithms used for the modeling and code timings for simulations through typical 2D and 3D meshes. We next calculate expected changes in image blur as scattering materials are placed upstream and downstream of a resolution test object (a 3 mm thick sheet of tantalum, into which 0.4 mm wide slits have been cut), and as the current supplied to the focusing magnets is varied. We compare and contrast the resulting simulations with the results of measurements obtained at the 800 MeV Los Alamos LANSCE Line-C proton radiography facility.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Evacuation decision behavior for no-notice emergency events

No-notice emergency events refer to unpredictable disasters such as earthquakes, chemical spills, or terrorist attacks, where it is impracticable to forewarn the public about their occurrence and design evacuation plans for them. This calls for an in-depth investigation of people’s evacuation behavior and identifying the most influential factors in their evacuation planning process to develop policy-sensitive pre-disaster plans for such events. As a response to this need, this study investigates individuals’ evacuation decision behavior in the context of no-notice emergency events. Since it is highly expected that people will have heterogenous decision behavior in such situations, we first apply a clustering algorithm to classify them into three maximally homogeneous clusters. Then, cluster-specific multivariate ordered probit models are developed to estimate the likelihood of selecting any of the three options of ignoring the situation, seeking shelter at the place, and evacuating to a safe place. The model estimation results indicate that a wide spectrum of factors affect the evacuation decision including individuals’ socio-economic attributes, disaster characteristics, built-environment factors, and government evacuation order. Further, variations of estimated coefficients across the population clusters highlight the significant dissimilarities in decision behavior of their associated members.

99 GENERAL AND MISCELLANEOUS↗

Python Library for Monte Carlo Simulations with Ab Initio and Machine-Learned Interatomic Potentials

There is a growing need in the simulation community for software that provides a transparent, reproducible, usable, and extensible (TRUE) Monte Carlo (MC) simulation framework employing energies from ab initio methods and machine-learning interatomic potentials (MLIPs). We introduce a Python library (ASE-MC) that adds Monte Carlo functionality to the Atomic Simulation Environment (ASE) package. Now, we can combine the powerful tools used to build systems and perform ab initio and MLIP in ASE with MC simulation algorithms to sample the configurational space with a concise Python script. After presenting the design philosophy, we demonstrate the flexibility of our approach using selected examples. These example simulations include liquid water described with a message-passing MLIP in the canonical and isothermal–isobaric ensembles, sampling the characteristic dihedral angle of biphenyl and comparing an MLIP to first-principles calculations, and a grand canonical Monte Carlo simulation of ammonia adsorption on Pt(111). These examples showcase the main features of the software, which include flexibility in the choice of ab initio or MLIP engine, ab initio or MLIP grand canonical MC with cavity bias insertions and deletions, the ability to add custom MC moves to the move set, and how users can condense complex MC workflows into a single Python script. Finally, this library serves as a framework for reproducible Monte Carlo simulations, facilitating easy reproduction of the work and application to new systems.

97 MATHEMATICS AND COMPUTING↗

A Kriging-Based Approach to Autonomous Experimentation with Applications to X-Ray Scattering

Abstract Modern scientific instruments are acquiring data at ever-increasing rates, leading to an exponential increase in the size of data sets. Taking full advantage of these acquisition rates will require corresponding advancements in the speed and efficiency of data analytics and experimental control. A significant step forward would come from automatic decision-making methods that enable scientific instruments to autonomously explore scientific problems—that is, to intelligently explore parameter spaces without human intervention, selecting high-value measurements to perform based on the continually growing experimental data set. Here, we develop such an autonomous decision-making algorithm that is physics-agnostic, generalizable, and operates in an abstract multi-dimensional parameter space. Our approach relies on constructing a surrogate model that fits and interpolates the available experimental data, and is continuously refined as more data is gathered. The distribution and correlation of the data is used to generate a corresponding uncertainty across the surrogate model. By suggesting follow-up measurements in regions of greatest uncertainty, the algorithm maximally increases knowledge with each added measurement. This procedure is applied repeatedly, with the algorithm iteratively reducing model error and thus efficiently sampling the parameter space with each new measurement that it requests. We validate the method using synthetic data, demonstrating that it converges to faithful replica of test functions more rapidly than competing methods, and demonstrate the viability of the approach in an experimental context by using it to direct autonomous small-angle (SAXS) and grazing-incidence small-angle (GISAXS) x-ray scattering experiments.

36 MATERIALS SCIENCE↗

Mode Selective Excitation Using Coherent Control Spectroscopy

Femtosecond time-resolved coherent anti-Stokes Raman scattering (fs-CARS) gives access to ultrafast molecular dynamics. However, femtosecond laser pulses are spectrally broad and therefore coherently excite several molecular modes. While the temporal resolution is high, usually no mode-selective excitation is possible. This paper demonstrates the feasibility of selectively exciting specific molecular vibrations in solution phase with shaped fs laser excitation using a feedback-controlled optimization technique guided by an evolutionary algorithm. This approach is also used to obtain molecule-specific CARS spectra from a mixture of different substances. The optimized phase structures of the fs pulses are characterized to get insight into the control process. Possible applications of the spectrum control are discussed.

74 ATOMIC AND MOLECULAR PHYSICS↗

The WaZP galaxy cluster sample of the dark energy survey year 1

We present a new (2+1)D galaxy cluster finder based on photometric redshifts called Wavelet Z Photometric (WaZP) applied to DES first year (Y1A1) data. The results are compared to clusters detected by the South Pole Telescope (SPT) survey and the redMaPPer cluster finder, the latter based on the same photometric data. WaZP searches for clusters in wavelet-based density maps of galaxies selected in photometric redshift space without any assumption on the cluster galaxy populations. The comparison to other cluster samples was performed with a matching algorithm based on angular proximity and redshift difference of the clusters. It led to the development of a new approach to match two optical cluster samples, following an iterative approach to minimize incorrect associations. The WaZP cluster finder applied to DES Y1A1 galaxy survey (1511.13 deg^2 up to = 23 mag) led to the detection of 60 547 galaxy clusters with redshifts 0.05 < z < 0.9 and richness N_gals ≥ 5. Considering the overlapping regions and redshift ranges between the DES Y1A1 and SPT cluster surveys, all sz based SPT clusters are recovered by the WaZP sample. The comparison between WaZP and redMaPPer cluster samples showed an excellent overall agreement for clusters with richness N_gals (λ for redMaPPer) greater than 25 (20), with 95 per cent recovery on both directions. Based on the cluster cross-match, we explore the relative fragmentation of the two cluster samples and investigate the possible signatures of unmatched clusters.

79 ASTRONOMY AND ASTROPHYSICS↗

Predictive Control to Further Reduce DC-Link Capacitor Current Stress for Segmented Inverter

Two three-phase interleaved inverters have been used in traction drive applications to reduce the current stress in a DC link capacitor bank. In such applications, either carrier-based or space vector modulation is used to select the optimum switching sequences, and the results show 50% less capacitor current than that of a single three-phase inverter. The switching state selection process for this inverter is tedious, and there has been no research to find the optimal switching state. To overcome this challenge, this research employed a simple finite set model predictive control to select the optimum switching sequence for a dual three-phase interleaved topology, called a segmented inverter. The results show that the predictive control algorithm can provide a simple solution and can reduce the current stress by 27% compared with traditional modulation techniques for the segmented inverters.

Ribeiro, Pedro↗

SU-E-T-201: Impact of Gantry and Table Angles On Proton Range in XiO Treatment Planning System

Purpose: To investigate if the selection of gantry and table angle has an impact on the calculated proton range (defined at the 90% distal falloff of the spread out Bragg peak) in XiO treatment planning system (TPS). Methods: A homogenous water phantom (40x40x40 cm{sup 3}) was constructed in XiO TPS. Ten treatment plans were generated using identical beam conditions (proton range = 25cm, modulation = 10cm) and device (10 cm diameter aperture) for various combinations of gantry (G) and table (T) angles (G0T90, G90T90, G180T90, G270T90, G90T0, G90T180, G90T270, G270T0, G270T180, and G270T270). The isocenter in each plan was selected at the center of the cubic target volume (5x5x5 cm{sup 3}) inside the phantom. All ten plans were calculated with pencil beam algorithm using a grid size of 3mm. The calculated proton ranges in all 10 plans were then compared against the prescribed (or expected) proton range, which was 25 cm in this study. Results: For all 10 treatment plans, calculated proton range was smaller when compared to the expected value (25cm), with XiO predicting the proton range from 24.62cm to 24.92cm. The difference between the prescribed and calculated proton range was highest (3.8mm) for plans with G90T270 and G270T90. Conclusion: Treatment plans generated in XiO TPS using G90T270 and G270T90 could produce a difference of more than 3 mm between the prescribed and calculated proton ranges. Further study using smaller dose calculation grid size (e.g., 1mm) is under progress to determine if the dose calculation grid size has any impact on the proton range. What causes the angle dependence of proton range calculation with the XiO TPS is also under investigation.

60 APPLIED LIFE SCIENCES↗

Using GANs with adaptive training data to search for new molecules

The process of drug discovery involves a search over the space of all possible chemical compounds. Generative Adversarial Networks (GANs) provide a valuable tool towards exploring chemical space and optimizing known compounds for a desired functionality. Standard approaches to training GANs, however, can result in mode collapse, in which the generator primarily produces samples closely related to a small subset of the training data. In contrast, the search for novel compounds necessitates exploration beyond the original data. In this work, we present an approach to training GANs that promotes incremental exploration and limits the impacts of mode collapse using concepts from Genetic Algorithms. In our approach, valid samples from the generator are used to replace samples from the training data. We consider both random and guided selection along with recombination during replacement. By tracking the number of novel compounds produced during training, we show that updates to the training data drastically outperform the traditional approach, increasing potential applications for GANs in drug discovery.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Measurement of $Z/\gamma^* + b$-jet Production Cross section in $p\bar{p}$ collisions at $\sqrt{s}= 1.96$ TeV with the CDF detector

Processes at hadron colliders, such as the production of jets, are described by the Quantum Chromodynamics theory (QCD). Precise descriptions of processes involving jets in association with a vector boson have nowadays large relevance as they represent irreducible background to other Standard Model (SM) processes and searches for new physics. The experimental study and understanding of the b-jet production in association with a Z boson are crucial for many reasons. For one side, it is the most important background for a light Higgs boson decaying into a bottom-antibottom quark pair and produced in the ZH mode.This is one of the most promising channels for the Higgs search at Tevatron in particular since the latest results have excluded the high mass region (MH > 127 GeV/c 2 ). For another side the signature of b-jets and a Z boson is also background to new physics searches, such as supersymmetry, where a large coupling of the Higgs boson to bottom quarks is allowed. The produ ction cross section measurement of b-jets in events with a Z boson has already been performed at hadron colliders, at the Tevatron by CDF and D0 experiments and are now pursued at the LHC by ATLAS and CMS. In particular the CDF measurement was performed with only 2 fb -1 and was limited by the statistical uncertainty. This PhD thesis presents a new measurement of the $Z/\gamma^* + b$-jet production cross section using the complete dataset collected by CDF during the Run II. $Z/\gamma^*$ bosons are selected in the electron and muon decay modes and are required to have 66 < MZ < 116 GeV/c 2 while jets, reconstructed with the MidPoint algorithm, have to be central (|Y| < 1.5) with pT > 20 GeV/c . The per jet cross section is measured with respect to the $Z/\gamma^*$ inclusive and the $Z/\gamma^* +$ jets cross sections. Results are compared to leading order (LO) event generator plus parton shower and next-to-leading order (NLO) predictions corrected for non perturbative effects such as hadronizat ion and underlying event. Differential distributions as a fu! nction of jet transverse moment and jet rapidity are also presented together with the comparison to NLO pQCD predictions for different renormalization and factorization scales and various PDF sets.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Evaluation and Comparison of Machine Learning Techniques for Rapid QSTS Simulations

Rapid and accurate quasi-static time series (QSTS) analysis is becoming increasingly important for distribution system analysis as the complexity of the distribution system intensifies with the addition of new types, and quantities, of distributed energy resources (DER). The expanding need for hosting capacity analysis, control systems analysis, photovoltaic (PV) and DER impact analysis, and maintenance cost estimations are just a few reasons that QSTS is necessary. Historically, QSTS analysis has been prohibitively slow due to the number of computations required for a full-year analysis. Therefore, new techniques are required that allow QSTS analysis to rapidly be performed for many different use cases. This research demonstrates a novel approach to doing rapid QSTS analysis for analyzing the number of voltage regulator tap changes in a distribution system with PV components. A representative portion of a yearlong dataset is selected and QSTS analysis is performed to determine the number of tap changes, and this is used as training data for a machine learning algorithm. The machine learning algorithm is then used to predict the number of tap changes in the remaining portion of the year not analyzed directly with QSTS. The predictions from the machine learning algorithms are combined with the results of the partial year simulation for a final prediction for the entire year, with the goal of maintaining an error <10% on the full-year prediction. Five different machine learning techniques were evaluated and compared with each other; a neural network ensemble, a random forest decision tree ensemble, a boosted decision tree ensemble, support vector machines, and a convolutional neural network deep learning technique. A combination of the neural network ensemble together with the random forest produced the best results. Using 20% of the year as training data, analyzed with QSTS, the average performance of the technique resulted in ~2.5% error in the yearly tap changes, while maintaining a <10% 99.9 th percentile error bound on the results. This is a 5x speedup compared to a standard, full-length QSTS simulation. These results demonstrate the potential for applying machine learning techniques to facilitate modern distribution system analysis and further integration of distributed energy resources into the power grid.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Model for mapping settlements

A programmable media includes a graphical processing unit in communication with a memory element. The graphical processing unit is configured to detect one or more settlement regions from a high resolution remote sensed image based on the execution of programming code. The graphical processing unit identifies one or more settlements through the execution of the programming code that executes a multi-instance learning algorithm that models portions of the high resolution remote sensed image. The identification is based on spectral bands transmitted by a satellite and on selected designations of the image patches.

97 MATHEMATICS AND COMPUTING↗

Autoencoders on FPGAs for real-time, unsupervised new physics detection at 40 MHz at the Large Hadron Collider

In this paper, we show how to adapt and deploy anomaly detection algorithms based on deep autoencoders, for the unsupervised detection of new physics signatures in the extremely challenging environment of a real-time event selection system at the Large Hadron Collider (LHC). We demonstrate that new physics signatures can be enhanced by three orders of magnitude, while staying within the strict latency and resource constraints of a typical LHC event filtering system. This would allow for collecting datasets potentially enriched with high-purity contributions from new physics processes. Through per-layer, highly parallel implementations of network layers, support for autoencoder-specific losses on FPGAs and latent space based inference, we demonstrate that anomaly detection can be performed in as little as $80\,$ns using less than 3% of the logic resources in the Xilinx Virtex VU9P FPGA. Opening the way to real-life applications of this idea during the next data-taking campaign of the LHC.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Two-Dimensional Computational Fluid Dynamics and Conduction Simulations of Heat Transfer in Horizontal Window Frames with Internal Cavities

This paper assesses the accuracy of the simplified frame cavity conduction/convection and radiation models presented in ISO 15099 and used in software for rating and labeling window products. Temperatures and U-factors for typical horizontal window frames with internal cavities are compared; results from Computational Fluid Dynamics (CFD) simulations with detailed radiation modeling are used as a reference. Four different frames were studied. Two were made of polyvinyl chloride (PVC) and two of aluminum. For each frame, six different simulations were performed, two with a CFD code and four with a building-component thermal-simulation tool using the Finite Element Method (FEM). This FEM tool addresses convection using correlations from ISO 15099; it addressed radiation with either correlations from ISO 15099 or with a detailed, view-factor-based radiation model. Calculations were performed using the CFD code with and without fluid flow in the window frame cavities; the calculations without fluid flow were performed to verify that the CFD code and the building-component thermal-simulation tool produced consistent results. With the FEM-code, the practice of subdividing small frame cavities was examined, in some cases not subdividing, in some cases subdividing cavities with interconnections smaller than five millimeters (mm) (ISO 15099) and in some cases subdividing cavities with interconnections smaller than seven mm (a breakpoint that has been suggested in other studies). For the various frames, the calculated U-factors were found to be quite comparable (the maximum difference between the reference CFD simulation and the other simulations was found to be 13.2 percent). A maximum difference of 8.5 percent was found between the CFD simulation and the FEM simulation using ISO 15099 procedures. The ISO 15099 correlation works best for frames with high U-factors. For more efficient frames, the relative differences among various simulations are larger. Temperature was also compared, at selected locations on the frames. Small differences was found in the results from model to model. Finally, the effectiveness of the ISO cavity radiation algorithms was examined by comparing results from these algorithms to detailed radiation calculations (from both programs). Our results suggest that improvements in cavity heat transfer calculations can be obtained by using detailed radiation modeling (i.e. view-factor or ray-tracing models), and that incorporation of these strategies may be more important for improving the accuracy of results than the use of CFD modeling for horizontal cavities.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

ReSCo: A Middleware Component for Reliable Service Composition in Pervasive Systems

Service composition schemes create high-level application services by combining several basic services. Service composition schemes for dynamic, open systems, such as those found in pervasive environments, must be cognizant of the possibility of failures and attacks. In open systems, it is seldom feasible to guarantee the reliability of each node prior to access; however, there may be several possible ways to compose the same high-level service, each having a different (though possibly overlapping) set of nodes that can satisfy the composition. We approach this problem with a Reliable Service Composition middleware component, ReSCo, to determine trustworthy compositions and nodes for service composition in dynamic, open systems. ReSCo is a modular, adaptive middleware component that selects from possible composition paths and nodes to enhance reliability of service compositions. ReSCo can work with a broad range of both service composition algorithms and trust establishment mechanisms.

97 MATHEMATICS AND COMPUTING↗