Search NASA⌕ Search

SEARCH · Search NASA

Results for “Soft computing”

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 379 records · Page 21
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

SU-C-213-03: Custom 3D Printed Boluses for Radiation Therapy

Purpose: To develop a clinical workflow and to commission the process of creating custom 3d printed boluses for radiation therapy. Methods: We designed a workflow to create custom boluses using a commercial 3D printer. Contours of several patients were deformably mapped to phantoms where the test bolus contours were designed. Treatment plans were created on the phantoms following our institutional planning guideline. The DICOM file of the bolus contours were then converted to stereoLithography (stl) file for the 3d printer. The boluses were printed on a commercial 3D printer using polylactic acid (PLA) material. Custom printing parameters were optimized in order to meet the requirement of bolus composition. The workflow was tested on multiple anatomical sites such as skull, nose and chest wall. The size of boluses varies from 6×9cm2 to 12×25cm2. To commission the process, basic CT and dose properties of the printing materials were measured in photon and electron beams and compared against water and soft superflab bolus. Phantoms were then scanned to confirm the placement of custom boluses. Finally dose distributions with rescanned CTs were compared with those computer-generated boluses. Results: The relative electron density(1.08±0.006) of the printed boluses resemble those of liquid tap water(1.04±0.004). The dosimetric properties resemble those of liquid tap water(1.04±0.004). The dosimetric properties were measured at dmax with an ion chamber in electron and photon open beams. Compared with solid water and soft bolus, the output difference was within 1% for the 3D printer material. The printed boluses fit well to the phantom surfaces on CT scans. The dose distribution and DVH based on the printed boluses match well with those based on TPS generated boluses. Conclusion: 3d printing provides a cost effective and convenient solution for patient-specific boluses in radiation therapy.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

Optimizing the optimizer for physics-informed neural networks and Kolmogorov-Arnold networks

Physics-Informed Neural Networks (PINNs) have revolutionized the computation of PDE solutions by integrating partial differential equations (PDEs) into the neural network’s training process as soft constraints, becoming an important component of the scientific machine learning (SciML) ecosystem. More recently, physics-informed Kolmogorv-Arnold networks (PIKANs) have also shown to be effective and comparable in accuracy with PINNs. In their current implementation, both PINNs and PIKANs are mainly optimized using first-order methods like Adam, as well as quasi-Newton methods such as BFGS and its low-memory variant, L-BFGS. However, these optimizers often struggle with highly nonlinear and non-convex loss landscapes, leading to challenges such as slow convergence, local minima entrapment, and (non)degenerate saddle points. In this study, we investigate the performance of Self- Scaled BFGS (SSBFGS), Self-Scaled Broyden (SSBroyden) methods and other advanced quasi-Newton schemes, including BFGS and L-BFGS with different line search strategies. These methods dynamically rescale updates based on historical gradient information, thus enhancing training efficiency and accuracy. We systematically compare these optimizers – using both PINNs and PIKANs – on key challenging PDEs, including the Burgers, Allen-Cahn, Kuramoto-Sivashinsky, Ginzburg-Landau, and Stokes equations. Additionally, we evaluate the performance of SSBFGS and SSBroyden for Deep Operator Network (DeepONet) architectures, demonstrating their effectiveness for data-driven operator learning. Our findings provide state-of-the-art results with orders-of-magnitude accuracy improvements without the use of adaptive weights or any other enhancements typically employed in PINNs. More broadly, our work reveal insights into the effectiveness of quasi-Newton optimization strategies in significantly improving the convergence and accurate generalization of PINNs and PIKANs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2016

Nonlinear bioheat transfer models and multi-objective numerical optimization of the cryosurgery operations

Numerical computation of the three dimensional problem of the freezing interface propagation during the cryosurgery coupled with the multi-objective optimization methods is used in order to improve the efficiency and safety of the cryosurgery operations performing. Prostate cancer treatment and cutaneous cryosurgery are considered. The heat transfer in soft tissue during the thermal exposure to low temperature is described by the Pennes bioheat model and is coupled with an enthalpy method for blurred phase change computations. The finite volume method combined with the control volume approximation of the heat fluxes is applied for the cryosurgery numerical modeling on the tumor tissue of a quite arbitrary shape. The flux relaxation approach is used for the stability improvement of the explicit finite difference schemes. The method of the additional heating elements mounting is studied as an approach to control the cellular necrosis front propagation. Whereas the undestucted tumor tissue and destucted healthy tissue volumes are considered as objective functions, the locations of additional heating elements in cutaneous cryosurgery and cryotips in prostate cancer cryotreatment are considered as objective variables in multi-objective problem. The quasi-gradient method is proposed for the searching of the Pareto front segments as the multi-objective optimization problem solutions.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2019

BSTC: A Novel Binarized-Soft-Tensor-Core Design for Accelerating Bit-Based Approximated Neural Nets

Binarized neural network (or BNN) promises tremendous performance improvement over traditional DNNs through simplified bit-level computation and significantly reduced memory access/storage cost. In addition, it has other advantages of low-cost, low-energy and high-robustness, showing great utilization potential in resources-constrained, volatile and latency-critical applications, which are critical for future HPC execution. However, the promised significant performance gain of BNN inference has never been fully demonstrated on general-purpose processors, particularly on GPUs, due to: (i) the challenge to extract and leverage sufficient fine-grained bit-level-parallelism to saturate GPU cores when batch is small; (ii) the fundamental design conflict between bit-based BNN algorithm and underlying word-based architecture; (iii) architecture \& performance unfriendly BNN network design. To address (i) and (ii), we propose binarized-soft-tensor-core as a software-hardware codesign approach to construct bit-manipulation capability for modern GPUs to effectively harvest the emerging bit-level-parallelism. To tackle (iii), we propose intra- and inter-layer fusion techniques so that the entire BNN inference can be packed into a single GPU kernel, to avoid high-cost frequent launching. Experiments demonstrate that our design can achieve over 1000x speedup for raw inference latency and 10x for inference throughput over the state-of-the-art full-precision simulated BNN inference for AlexNet on ImageNet.

Li, Ang↗
NASA NTRS1982

Computer modeling of multiple-channel input signals and intermodulation losses caused by nonlinear traveling wave tube amplifiers

The multiple channel input signal to a soft limiter amplifier as a traveling wave tube is represented as a finite, linear sum of Gaussian functions in the frequency domain. Linear regression is used to fit the channel shapes to a least squares residual error. Distortions in output signal, namely intermodulation products, are produced by the nonlinear gain characteristic of the amplifier and constitute the principal noise analyzed in this study. The signal to noise ratios are calculated for various input powers from saturation to 10 dB below saturation for two specific distributions of channels. A criterion for the truncation of the series expansion of the nonlinear transfer characteristic is given. It is found that he signal to noise ratios are very sensitive to the coefficients used in this expansion. Improper or incorrect truncation of the series leads to ambiguous results in the signal to noise ratios.

Stankiewicz, N.↗
DOE OSTI2018

Computational study of ignition behavior and hotspot dynamics of a potential class of aluminized explosives

The ignition behavior and hotspot dynamics of a potential class of aluminized energetic materials are studied computationally. The materials consist of HMX (octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine) grains embedded in an aluminum matrix and, henceforth referred to as metal–matrix explosives (MMXs). For the analysis, two different MMXs, the soft MMX with a matrix of 1100 Al alloy and the hard MMX with a matrix of 7075 T651 Al alloy are considered. The thermo-mechanical response of the MMXs are computationally analyzed by subjecting them to monotonic impact loading using a Lagrangian cohesive finite element framework, with their ignition behavior analyzed through characterization of hotspots. For comparison, a polymer-bonded explosive (PBX) consisting of HMX and Estane is also analyzed under the same conditions. Here, the results show that the MMXs have significantly lower propensity for ignition and higher structural integrity than the PBX over the loading velocity range of 200–500 m s -1 .

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗
DOE OSTI2016

Jet shapes in dijet events at the LHC in SCET

Here, we consider the class of jet shapes known as angularities in dijet production at hadron colliders. These angularities are modified from the original definitions in e + e- collisions to be boost invariant along the beam axis. These shapes apply to the constituents of jets defined with respect to either k T -type (anti-k T , C/A, and k T ) algorithms and cone-type algorithms. We present an SCET factorization formula and calculate the ingredients needed to achieve next-to-leading-log (NLL) accuracy in kinematic regions where non-global logarithms are not large. The factorization formula involves previously unstudied “unmeasured beam functions,” which are present for finite rapidity cuts around the beams. We derive relations between the jet functions and the shape-dependent part of the soft function that appear in the factorized cross section and those previously calculated for e + e - collisions, and present the calculation of the non-trivial, color-connected part of the soft-function to O(αs) . This latter part of the soft function is universal in the sense that it applies to any experimental setup with an out-of-jet p T veto and rapidity cuts together with two identified jets and it is independent of the choice of jet (sub-)structure measurement. In addition, we implement the recently introduced soft-collinear refactorization to resum logarithms of the jet size, valid in the region of non-enhanced non-global logarithm effects. While our results are valid for all 2 → 2 channels, we compute explicitly for the qq' → qq' channel the color-flow matrices and plot the NLL resummed differential dijet cross section as an explicit example, which shows that the normalization and scale uncertainty is reduced when the soft function is refactorized. For this channel, we also plot the jet size R dependence, the p$cut\atop{T}$ dependence, and the dependence on the angularity parameter a.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2026

Charge Transport in Solvated Donor–Acceptor Functionalized Peptoids: Molecular Dynamics and Rate Theory

Scalable solar-energy conversion requires photoactive materials that combine the efficiency of natural photosynthetic systems with the stability and processability needed for practical applications. Achieving reliable charge transport in soft, self-assembled organic materials remains challenging, as structural fluctuations and environmental effects strongly influence charge-transfer (CT) rates. Here, we present a broadly applicable computational framework for evaluating CT rates in the condensed phase, combining Fermi’s golden rule rate theory with inputs from all-atom molecular dynamics (MD) simulations and first-principles electronic-structure calculations. The approach does not rely on system-specific parametrization and is applicable to a wide range of soft and disordered materials. We demonstrate the applicability and usefulness of the framework on redox-active peptoids functionalized with iron–porphyrin (Fe–P) complexes, a bioinspired platform with programmable donor–acceptor units and tunable three-dimensional organization. The calculated CT rates exhibit strong sensitivity to molecular conformation, with variations spanning several orders of magnitude. This dependence is shown to arise from the pronounced variation in diabatic electronic coupling with the relative orientations and separations of the Fe–P complexes across the conformational ensemble. The framework provides a consistent route for connecting atomistic structure to CT kinetics in the condensed phase and enables analysis of structure–rate relations in organic semiconducting systems.

Charge transfer↗
DOE OSTI2020

Synchrotron CT imaging of lattice structures with engineered defects

Understanding mechanical failure, crack propagation, and compressive behavior at the micrometer scale is essential for tailoring material properties for structural performance in cellular materials. Typically, modeling of traditional polymer foam materials is clouded by the lack of control in material morphology and its inherent stochastic structure. Additive manufacturing with sub-micrometer resolution provides a direct path for experimenters to specifically tailor structures needed by modelers to explicitly probe mechanical performance. Using laboratory-based 3D X-ray computed tomographic imaging (CT), the examination of deformation and damage provides a critical path to understand how these soft materials behave. Additionally, synchrotron CT yields realistic information at higher strain rates to directly validate the robustness of our finite element modeling. For this study, nanolithographic printing was employed to generate a series of engineered lattices with increasing levels of defects through the random removal of ligaments. These structures were mechanically tested and imaged with laboratory-based microCT. Additionally, synchrotron experiments were conducted in which the structures were imaged in 3D at 14 Hz during compression at a 0.4 s –1 strain rate. These 3D images show the changes in the structure as the ligaments bend, buckle and fracture in real time. This technique provides a robust framework for developing our methodologies and future exploration of engineered structures.

36 MATERIALS SCIENCE↗
DOE OSTI2019

ESPResSo++ 2.0: Advanced methods for multiscale molecular simulation

We report that molecular simulation is a scientific tool used in many fields including material science and biology. This requires constant development and enhancement of algorithms within molecular simulation software packages. Here, we present computational tools for multiscale modeling developed and implemented within the ESPResSo++ package. These include the latest applications of the adaptive resolution scheme, the hydrodynamic interactions through a lattice Boltzmann solvent coupled to particle-based molecular dynamics, the implementation of the hierarchical strategy for equilibrating long-chained polymer melts and a heterogeneous spatial domain decomposition. The software design of ESPResSo++ has kept its highly modular C++ kernel with a Python user interface. Furthermore, it has been enhanced by automatic scripts that parse configurations from other established packages, providing scientists with the ability to rapidly set up their simulations.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Micro-imaging of Brain Cancer Radiation Therapy Using Phase-contrast Computed Tomography

Experimental neuroimaging provides a wide range of methods for the visualization of brain anatomic morphology down to subcellular detail. Still, each technique-specific detection mechanism presents compromises among the achievable field-of-view size, spatial resolution, and nervous tissue sensitivity, leading to partial sample coverage, unresolved morphologic structures, or sparse labeling of neuronal populations and often also to obligatory sample dissection or other sample invasive manipulations. X-ray phase-contrast imaging computed tomography (PCI-CT) is an experimental imaging method that simultaneously provides micrometric spatial resolution, high soft-tissue sensitivity, and ex vivo full organ rodent brain coverage without any need for sample dissection, staining or labeling, or contrast agent injection. In the present study, we explored the benefits and limitations of PCI-CT use for in vitro imaging of normal and cancerous brain neuromorphology after in vivo treatment with synchrotron-generated x-ray microbeam radiation therapy (MRT), a spatially fractionated experimental high-dose radiosurgery. The goals were visualization of the MRT effects on nervous tissue and a qualitative comparison of the results to the histologic and high-field magnetic resonance imaging findings.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2026

Development of Data-Driven Models for Performance Prediction and Chemical Dosing of a Full-Scale Controlled Phosphorus Precipitation Reactor

This study evaluated the use of data-driven models to improve control of a struvite precipitation reactor that removes phosphorus from wastewater while producing a fertilizer product. The researchers developed predictive models for influent orthophosphate concentration, effluent orthophosphate concentration, and phosphorus removal using operational data from a full-scale MagPrex™ reactor at a water resource recovery facility in Denver, Colorado. Model predictions were used to recommend magnesium chloride dosing adjustments needed to achieve a target effluent phosphorus concentration. Several machine learning approaches were tested, with ridge regression providing the best predictions for influent orthophosphate concentration and phosphorus removal, and XGBoost providing the best predictions for effluent orthophosphate concentration. Simulation results indicated that the decision-support approach could correctly identify dosing adjustments in most cases and reduce chemical use. Full-scale implementation achieved lower accuracy due to changing operating conditions and limited historical data in some operating ranges. Here, the results demonstrate the potential of data-driven tools to support phosphorus recovery process control while also identifying practical limitations that affect deployment in full-scale systems.

42 ENGINEERING↗
DOE OSTI2026

Simple algorithm for polarized parton evolution

We present an algorithm to include the correlation between the production and decay planes of gluons in a parton-shower simulation. The technique is based on identifying the charge currents responsible for the creation and annihilation of the vector field. It is applicable in both the hard-collinear and the soft wide-angle region. As a function of the number of particles, the algorithm scales linear in computing time and memory. We demonstrate agreement with fixed-order perturbative calculations in the relevant kinematical limits, and present a new observable that can be used to probe correlations beyond current-current interactions.

Höche, Stefan [Fermi National Accelerator Laborato↗
DOE OSTI2024

Inferring effective electrostatic interaction of charge-stabilized colloids from scattering using deep learning

In this article, an innovative strategy is presented that incorporates deep auto-encoder networks into a least-squares fitting framework to address the potential inversion problem in small-angle scattering. To evaluate the performance of the proposed approach, a detailed case study focusing on charged colloidal suspensions was carried out. The results clearly indicate that a deep learning solution offers a reliable and quantitative method for studying molecular interactions. The approach surpasses existing deterministic approaches with respect to both numerical accuracy and computational efficiency. Overall, this work demonstrates the potential of deep learning techniques in tackling complex problems in soft-matter structures and beyond.

36 MATERIALS SCIENCE↗
DOE OSTI2015

Low tube voltage dual source computed tomography to reduce contrast media doses in adult abdomen examinations: A phantom study

Purpose: To evaluate the potential of low tube voltage dual source (DS) single energy (SE) and dual energy (DE) computed tomography (CT) to reduce contrast media (CM) dose in adult abdominal examinations of various sizes while maintaining soft tissue and iodine contrast-to-noise ratio (CNR). Methods: Four abdominal phantoms simulating a body mass index of 16 to 35 kg/m{sup 2} with four inserted syringes of 0, 2, 4, and 8 mgI/ml CM were scanned using a 64-slice DS-CT scanner. Six imaging protocols were used; one single source (SS) reference protocol (120 kV, 180 reference mAs), four low kV SE protocols (70 and 80 kV using both SS and DS), and one DE protocol at 80/140 kV. Potential CM reduction with unchanged CNRs relative to the 120 kV protocol was calculated along with the corresponding increase in radiation dose. Results: The potential contrast media reductions were determined to be approximately 53% for DS 70 kV, 51% for SS 70 kV, 44% for DS 80 kV, 40% for SS 80 kV, and 20% for DE (all differences were significant, P < 0.05). Constant CNR could be achieved by using DS 70 kV for small to medium phantom sizes (16–26 kg/m{sup 2}) and for all sizes (16–35 kg/m{sup 2}) when using DS 80 kV and DE. Corresponding radiation doses increased by 60%–107%, 23%–83%, and 6%–12%, respectively. Conclusions: DS single energy CT can be used to reduce CM dose by 44%–53% with maintained CNR in adult abdominal examinations at the cost of an increased radiation dose. DS dual-energy CT allows reduction of CM dose by 20% at similar radiation dose as compared to a standard 120 kV single source.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

Fluctuating Hydrodynamics Methods for Dynamic Coarse-Grained Implicit-Solvent Simulations in LAMMPS

We present a software package integrated with the molecular dynamics software LAMMPS for fluctuating hydrodynamics simulations of fluid-structure interactions subject to thermal fluctuations. The package is motivated to provide dynamic thermostats to extend implicit-solvent coarse-grained (IS-CG) models by incorporating kinetic contributions from the solvent to facilitate their use in a wider range of applications. To capture the thermal and hydrodynamic contributions of the solvent to dynamics, we introduce momentum conserving thermostats and computational methods based on fluctuating hydrodynamics and the stochastic Eulerian Lagrangian method (SELM). SELM couples the coarse-grained microstructure degrees of freedom to continuum stochastic fields to capture both the relaxation of hydrodynamic modes and thermal fluctuations. Features of the SELM software include (i) numerical time-step integrators for SELM fluctuating hydrodynamics in inertial and quasi-steady regimes, (ii) Lees--Edwards-style methods for imposing shear, (iii) a Java-based graphical user interface (GUI) for setting up models and simulations, (iv) standardized XML formats for parametrization and data output, and (v) standardized formats VTK for continuum fields and microstructures. The SELM software package facilitates for pre-established models in LAMMPS easy adoption of the SELM fluctuating hydrodynamics thermostats. We offer here an overview of the SELM software package, computational methods, and applications.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2010

A Hybrid Higgs

We construct composite Higgs models admitting a weakly coupled Seiberg dual description. We focus on the possibility that only the up-type Higgs is an elementary field, while the down-type Higgs arises as a composite hadron. The model, based on a confining SQCD theory, breaks supersymmetry and electroweak symmetry dynamically and calculably. This simultaneously solves the {mu}/B{sub {mu}} problem and explains the smallness of the bottom and tau masses compared to the top mass. The proposal is then applied to a class of models where the same confining dynamics is used to generate the Standard Model flavor hierarchy by quark and lepton compositeness. This provides a unified framework for flavor, supersymmetry breaking and electroweak physics. The weakly coupled dual is used to explicitly compute the MSSM parameters in terms of a few microscopic couplings, giving interesting relations between the electroweak and soft parameters. The RG evolution down to the TeV scale is obtained and salient phenomenological predictions of this class of 'single-sector' models are discussed.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2016

Determining the vulnerability of multi-threaded program code to soft errors

The described embodiments include a program code testing system that determines the vulnerability of multi-threaded program code to soft errors. For multi-threaded program code, two to more threads from the program code may access shared architectural structures while the program code is being executed. The program code testing system determines accesses of architectural structures made by the two or more threads of the multi-threaded program code and uses the determined accesses to determine a time for which the program code is exposed to soft errors. From this time, the program code testing system determines a vulnerability of the program code to soft errors.

97 MATHEMATICS AND COMPUTING↗