Search NASA⌕ Search

Engineering topics

Lee, C. S.

Publications and source records attributed to Lee, C. S..

At least 19 records

Graph neural network for neutrino physics event reconstruction

Liquid argon time projection chamber (LArTPC) detector technology offers a wealth of high-resolution information on particle interactions, and leveraging that information to its full potential requires sophisticated automated reconstruction techniques. Here, this article describes NUGRAPH 2, a graph neural network for low-level reconstruction of simulated neutrino interactions in a LArTPC detector. Simulated neutrino interactions in the MicroBooNE detector geometry are described as heterogeneous graphs, with energy depositions on each detector plane forming nodes on planar subgraphs. The network utilizes a multihead attention message-passing mechanism to perform background filtering and semantic labeling on these graph nodes, identifying those associated with the primary physics interaction with 98.0% efficiency and labeling them according to particle type with 94.9% efficiency. The network operates directly on detector observables across multiple two-dimensional representations but utilizes a three-dimensional-context-aware mechanism to encourage consistency between these representations. Model inference takes 0.12 s / event on a CPU and 0.005 s / event batched on a GPU. This architecture is designed to be a general-purpose solution for particle reconstruction in neutrino physics, with the potential for deployment across a broad range of detector technologies, and offers a core convolution engine that can be leveraged for a variety of tasks beyond the two described in this paper.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

NuGraph2: A Graph Neural Network for Neutrino Event Reconstruction

Neutrino experiments are set to probe some of the most important open questions in physics, from CP violation and the nature of dark matter. The technology of choice for many of these experiments is the liquid argon time projection chamber (LArTPC). In current LArTPC experiments, reconstruction performance often represents a limiting factor for the sensitivity. New developments are therefore needed to unlock the full potential of LArTPC experiments. NuGraph2 is a state of the art Graph Neural Network for reconstruction of data in LArTPC experiments. NuGraph2 utilizes a heterogeneous graph structure, with separate subgraphs of 2D nodes (hits in each plane) connected across planes via 3D nodes (space points). The model provides a consistent description of the neutrino interaction across all planes. NuGraph2 is a multi-purpose network, with a common message-passing attention engine connected to multiple decoders with different classification or regression tasks. These include the classification of detector hits according to the particle type that produced them (semantic segmentation) and the separation of hits from the neutrino interaction from hits due to noise or cosmic-ray background. Additional decoders are being developed, performing tasks such as the regression of the neutrino interaction vertex position. Performance results will be presented based on publicly available samples from MicroBooNE. These include both physics performance metrics, achieving 95% accuracy for semantic segmentation and 98% classification of neutrino hits, as well as computational metrics for training and for inference on CPU or GPU. The status of the NuGraph integration in the LArSoft software framework will be presented, as well as initial studies about model interpretability and injection of domain knowledge.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

NuGraph2: A Graph Neural Network for Neutrino Event Reconstruction

Neutrino experiments are set to probe some of the most important open questions in physics, from CP violation and the nature of dark matter. The technology of choice for many of these experiments is the liquid argon time projection chamber (LArTPC). In current LArTPC experiments, reconstruction performance often represents a limiting factor for the sensitivity. New developments are therefore needed to unlock the full potential of LArTPC experiments. NuGraph2 is a state of the art Graph Neural Network for reconstruction of data in LArTPC experiments. NuGraph2 utilizes a heterogeneous graph structure, with separate subgraphs of 2D nodes (hits in each plane) connected across planes via 3D nodes (space points). The model provides a consistent description of the neutrino interaction across all planes. NuGraph2 is a multi-purpose network, with a common message-passing attention engine connected to multiple decoders with different classification or regression tasks. These include the classification of detector hits according to the particle type that produced them (semantic segmentation) and the separation of hits from the neutrino interaction from hits due to noise or cosmic-ray background. Additional decoders are being developed, performing tasks such as the regression of the neutrino interaction vertex position. Performance results will be presented based on publicly available samples from MicroBooNE. These include both physics performance metrics, achieving 95% accuracy for semantic segmentation and 98% classification of neutrino hits, as well as computational metrics for training and for inference on CPU or GPU. The status of the NuGraph integration in the LArSoft software framework will be presented, as well as initial studies about model interpretability and injection of domain knowledge.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

SRBench++: Principled Benchmarking of Symbolic Regression With Domain-Expert Interpretation

Symbolic regression searches for analytic expressions that accurately describe studied phenomena. The main promise of this approach is that it may return an interpretable model that can be insightful to users, while maintaining high accuracy. The current standard for benchmarking these algorithms is SRBench, which evaluates methods on hundreds of datasets that are a mix of real-world and simulated processes spanning multiple domains. At present, the ability of SRBench to evaluate interpretability is limited to measuring the size of expressions on real-world data, and the exactness of model forms on synthetic data. In practice, model size is only one of many factors used by subject experts to determine how interpretable a model truly is. Furthermore, SRBench does not characterize algorithm performance on specific, challenging sub-tasks of regression such as feature selection and evasion of local minima. In this work, we propose and evaluate an approach to benchmarking SR algorithms that addresses these limitations of SRBench by 1) incorporating expert evaluations of interpretability on a domain-specific task, and 2) evaluating algorithms over distinct properties of data science tasks. We evaluate 12 modern symbolic regression algorithms on these benchmarks and present an in-depth analysis of the results, discuss current challenges of symbolic regression algorithms and highlight possible improvements for the benchmark itself.

97 MATHEMATICS AND COMPUTING↗

Isomeric and beta-decay spectroscopy of 173,174 Ho

β-decay spectroscopy of 173,174 Ho (Z = 67, N = 106,107) was conducted at Radioactive Isotope Beam Factory at RIKEN by using in-flight fission of a 345-MeV/u 238 U primary beam. A previously unreported isomeric state at 405 keV with half-life of 3.7(12) μs and a spin and parity of (3/2 + ) is identified in 173 Ho. Moreover, a new state with a spin and parity of 9 - was discovered in 174 Er. The experimental log ft values of 5.84(20) and 5.25(18) suggest an allowed-hindered β decay from the ground state of 174 Ho to the K π = 8 - isomeric state in 174 Er. Configuration-constrained potential energy surface (PES) calculations were performed and the predictions are in reasonable agreement with the experimental results.

150 ≤ A ≤ 189↗

Algebraic Hybridization and Static Condensation with Application to Scalable $H$(div) Preconditioning

We propose here a unified algebraic approach for the practical application and preconditioning of static condensation and hybridization, two popular techniques in finite element discretizations. We demonstrate the use of this algebraic framework for the construction of scalable solvers for problems involving $H$(div)-spaces discretized by conforming (Raviart--Thomas) elements of arbitrary order. We illustrate through numerical experiments the relative performance of the two (in some sense dual) techniques in comparison with a state-of-the-art parallel solver, ADS [T. V. Kolev and P. S. Vassilevski, SIAM J. Sci. Comput., 34 (2012), pp. A3079--A3098], available at http://www.llnl.gov/casc/hypre and http://mfem.org. Based on these results, we recommend the use of the hybridization technique in practice, due to its clearly demonstrated superior performance with increased benefit for higher-order elements.

97 MATHEMATICS AND COMPUTING↗

β - γ and isomeric decay spectroscopy of 168 Dy

Here, this contribution will report on the experimental work on the level structure of 168 Dy. The experimental data have been taken as part of the EURICA decay spectroscopy campaign at RIBF, RIKEN in November 2014. In the experiment, a 238 U primary beam is accelerated up to 345 MeV/u with an average intensity of 12 pnA. The nuclei of interest are produced by in-flight fission of 238 U impinging on Be target with a thickness of 5 mm. The excited states of 168 Dy have been populated through the decay from a newly identified isomeric state and via the beta decay from 168 Tb. In this contribution, scientific motivations, experimental procedure and some preliminary results for this study are presented.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Calculation of the rotor induced download on airfoils

Interactions between the rotors and wing of a rotary wing aircraft in hover have a significant detrimental effect on its payload performance. The reduction of payload results from the wake of lifting rotors impinging on the wing, which is at 90 deg angle of attack in hover. This vertical drag, often referred as download, can be as large as 15 percent of the total rotor thrust in hover. The rotor wake is a three-dimensional, unsteady flow with concentrated tip vortices. With the rotor tip vortices impinging on the upper surface of the wing, the flow over the wing is not only three-dimensional and unsteady, but also separated from the leading and trailing edges. A simplified two-dimensional model was developed to demonstrate the stability of the methodology. The flow model combines a panel method to represent the rotor and the wing, and a vortex method to track the wing wake. A parametric study of the download on a 20 percent thick elliptical airfoil below a rotor disk of uniform inflow was performed. Comparisons with experimental data are made where the data are available. This approach is now being extended to three-dimensional flows. Preliminary results on a wing at 90 deg angle of attack in free stream is presented.

Lee, C. S.↗

Radar cross section of an open-ended circular waveguide Calculation of second-order diffraction terms

The radar cross section of an open-ended, semi-infinite circular waveguide is calculated by the geometrical theory of diffraction and the equivalent-current (EC) method. Both single- and double-diffraction terms are included. It is found that the double-diffraction term is stronger than the single-diffraction term for the horizontal polarization at wide-angle incidence. Its inclusion is necessary in order to check with experimental data and the asymptotic expansion of the exact Wiener-Hopf solution. The EC method is used for the single diffraction. On the other hand the double-diffraction term is obtained by using the EC method for axial incidence and geometrical theory of diffraction for wide-angle incidence. These two solutions are matched with a proper matching function. This technique is computationally more efficient than the matching technique of integration over the equivalent currents.

Lee, C. S.↗

Two blowing concepts for roll and lateral control of aircraft

Two schemes to modulate aerodynamic forces for roll and lateral control of aircraft have been investigated. The first scheme, called the lateral blowing concept, consists of thin jets of air exiting spanwise, or at small angle with the spanwise direction, from slots at the tips of straight wings. For this scheme, in addition to experimental measurements, a theory was developed showing the analytical relationship between aerodynamic forces and jet and wing parameters. Experimental results confirmed the theoretically derived scaling laws. The second scheme, which was studied experimentally, is called the jet spoiler concept and consists of thin jets exiting normally to the wing surface from slots aligned with the spanwise direction.

Tavella, D. A.↗

A simple circular-polarized antenna: Circular waveguide horn coated with lossy magnetic material

A circular waveguide horn coated with a lossy material in its interior wall can be used as an alternative to a corrugated waveguide for radiating a circularly polarized (CP) field. To achieve good CP radiation, the diameter of the structure must be larger than the free-space wavelength, and the coating material must be sufficiently lossy and magnetic. This device is cheaper and lighter in weight than the corrugated one.

Lee, C. S.↗

Normal modes in an overmoded circular waveguide coated with lossy material

The normal modes in an overmoded waveguide coated with a lossy material are analyzed, particularly for their attenuation properties as a function of coating material, layer thickness, and frequency. When the coating material is not too lossy, the low-order modes are highly attenuated even with a thin layer of coating. This coated guide serves as a mode suppressor of the low-order modes, which can be particularly useful for reducing the radar cross section of a cavity structure such as a jet engine inlet. When the coating material is very lossy, low-order modes fall into two distinct groups: highly and lowly attenuated modes. However, as a/lambda (a = radius of the cylinder; lambda = the free-space wavelength) increases, the separation between these two groups becomes less distinctive. The attenuation constants of most of the low-order modes become small and decrease as a function of lambda-squared/a-cubed.

Lee, C. S.↗

Analysis of cross-hemispheric influences on the monsoon trough and tropical cyclone genesis during FGGE and diurnal subsidence differences

For a number of years our project has been studying the cross-equatorial (winter to summer hemisphere) processs that can lead to the day-to-day alterations in the strength of the monsoon trough. These processes are also related to the genesis and intensification of tropical storms. The cross-hemispheric processes that occurred during the FGGE year are currently being studied with the use of the ECMWF analysis. How the winter hemisphere can affect low level cold surge penetration across the equator following cold front passage, and how upper tropospheric anticyclones of the winter hemisphere can produce an intensification of a tropical cyclone of the opposite hemisphere is described. ECMWF analysis of the 00Z versus 12Z diurnal difference in the clear region (10 deg diameter) subsidence occurring in the subtropical Pacific Ocean during FGGE is presented. The general reliability of the FGGE ECMWF analysis with regard to the specification of the large scale structure of tropical cyclones is shown.

Gray, W. M.↗

Numerical methods for analyzing electromagnetic scattering

Attenuation properties of the normal modes in an overmoded waveguide coated with a lossy material were analyzed. It is found that the low-order modes, can be significantly attenuated even with a thin layer of coating if the coating material is not too lossy. A thinner layer of coating is required for large attenuation of the low-order modes if the coating material is magnetic rather than dielectric. The Radar Cross Section (RCS) from an uncoated circular guide terminated by a perfect electric conductor was calculated and compared with available experimental data. It is confirmed that the interior irradiation contributes to the RCS. The equivalent-current method based on the geometrical theory of diffraction (GTD) was chosen for the calculation of the contribution from the rim diffraction. The RCS reduction from a coated circular guide terminated by a PEC are planned schemes for the experiments are included. The waveguide coated with a lossy magnetic material is suggested as a substitute for the corrugated waveguide.

Lee, S. W.↗

Normal modes in an overmoded circular waveguide coated with lossy material

The normal modes in an overmoded waveguide coated with a lossy material are analyzed, particularly for their attenuation properties as a function of coating material, layer thickness, and frequency. When the coating material is not too lossy, the low-order modes are highly attenuated even with a thin layer of coating. This coated guide serves as a mode suppressor of the low-order modes, which can be particularly useful for reducing the radar cross section (RCS) of a cavity structure such as a jet inlet. When the coating material is very lossy, low-order modes fall into two distinct groups: highly and lowly attenuated modes. However, as a/lambda (a = radius of the cylinder; lambda = the free-space wavelength) increases, the separation between these two groups becomes less distinctive. The attenuation constants of most of the low-order modes become small, and decrease as a function of lambda sup 2/a sup 3.

Lee, C. S.↗

Numerical methods for analyzing electromagnetic scattering

Numerical methods to analyze electromagnetic scattering are presented. The dispersions and attenuations of the normal modes in a circular waveguide coated with lossy material were completely analyzed. The radar cross section (RCS) from a circular waveguide coated with lossy material was calculated. The following is observed: (1) the interior irradiation contributes to the RCS much more than does the rim diffraction; (2) at low frequency, the RCS from the circular waveguide terminated by a perfect electric conductor (PEC) can be reduced more than 13 dB down with a coating thickness less than 1% of the radius using the best lossy material available in a 6 radius-long cylinder; (3) at high frequency, a modal separation between the highly attenuated and the lowly attenuated modes is evident if the coating material is too lossy, however, a large RCS reduction can be achieved for a small incident angle with a thin layer of coating. It is found that the waveguide coated with a lossy magnetic material can be used as a substitute for a corrugated waveguide to produce a circularly polarized radiation yield.

Lee, S. W.↗