SEARCH · Search NASA

Results for “gradient based optimization”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 451 records · Page 25

All sources NASA NTRS DOE OSTI

DOE OSTI2023

HPC-Enabled Optimization of High Temperature Heat Exchangers (CRADA Final Report)

This project was a collaborative effort between Lawrence Livermore National Security, LLC (LLNS) as manager and operator of Lawrence Livermore National Laboratory (LLNL) and Materials Sciences, LLC, to develop a technology for design and optimization of heat exchangers using powerful desktop and laptop computers. The project was originally designated as a 12-month project, and consisted of 3# major tasks and the following 8# major deliverables: 1) CFD models of 3D heat exchangers based on existing and new geometry. 2) Validation against experimental data provided by MSC and published in the literature. 3) CFD models of 3D unit cells based on TPMS. 4) Surrogate models capable of delivering the gradients of the homogenized properties with respect to the parametrization. 5) 3D design methodology using TO algorithms. 6) Conventional reference and topology optimized designs. 7) 3D optimized designs stored in a 3D printer build format. 8) Verification of the improved performance. All of the deliverables for this project were successfully completed with two no-cost time extensions.

36 MATERIALS SCIENCE↗

DOE OSTI2026

HPC-Enabled Optimization of High Temperature Heat Exchangers (CRADA Final Report)

13 HYDRO ENERGY↗

DOE OSTI2021

Optimizing $\mathrm{ACRT}$ to reduce inclusion formation during the $\mathrm{VGF}$ growth of cadmium zinc telluride: I. Computational approach

Theoretical computations reveal insight into parameters that govern the accelerated crucible rotation technique (ACRT) applied to the gradient freeze (VGF) growth of cadmium zinc telluride (CZT). A metric based on the classic Mullins and Sekerka instability criterion is put forth that provides a quantitative means of assessing the impact of different ACRT rotation schedules on morphological stability and inclusion formation during CZT growth. Here this metric is employed in a 2-k factorial design of experiments that identifies the maximum rotation rate as the most important ACRT parameter, followed by the acceleration period and the rest period. Subsequent optimization of rotation schedule is performed for a specific VGF growth system, and mechanistic explanations are provided for the preferred ACRT schedule, which improves interface stability via slower rotation rates and longer acceleration periods than would be applied using classical ACRT schedules.

36 MATERIALS SCIENCE↗

NASA NTRS2021

Machine Learning to Predict Joint Performance in Epoxy Composites Based on Process Parameters

Polymer matrix composites are gaining popularity in the aerospace industry due to their high specific strength, fatigue properties, and processability. However, based on current FAA certification guidelines, manufacturers utilizing current state-of-the-art composites made with adhesive bonds commonly install redundant fasteners to guarantee the strength of these adhesively bonded composite parts. The number of fasteners in a single-aisle commercial transport aircraft is typically on the order of 105, which reduces manufacturing rate, increases cost tremendously, and reduces the advantage of the specific strength composites provide. Due to this, the Adhesive Free Bonding of Composites (AERoBOND) project at NASA Langley Research Center has developed a novel assembly process to manufacture complex composite parts without the use of adhesives and fasteners. However, optimization of the process is currently challenging due to the complex and interdependent process parameters. To assist with the optimization, four machine learning algorithms utilizing gradient boosting decision trees were created to provide predictions for the mechanical and characterization properties of the composite parts. Approximately 200 random states from each algorithm were tested, and the models from each state were isolated and analyzed based on their accuracy, a validation process, and their feature importance. This analysis concluded that the models created from the machine learning algorithms could accelerate a parametric study for the AERoBOND process by rapidly optimizing process parameters to achieve desired performance characteristics.

Brennen Michael Middleton↗

NASA NTRSDate not supplied

Machine Learning to Predict Joint Performance in Epoxy Composites Based on Process Parameters

Polymer matrix composites are gaining popularity in the aerospace industry due to their high specific strength, fatigue properties, and processability. However, based on current FAA certification guidelines, manufacturers utilizing current state-of-the art composites made with adhesive bonds commonly install redundant fasteners to guarantee the strength of these adhesively bonded composite parts.1,2 The number of fasteners in a single-aisle commercial transport aircraft is typically on the order of 105, which reduces manufacturing rate, increases cost tremendously, and reduces the advantage of the specific strength composites provide. Due to this, the Adhesive Free Bonding of Composites (AERoBOND) project at NASA Langley Research Center has developed a novel assembly process to manufacture complex composite parts without the use of adhesives and fasteners.1 However, optimization of the process is currently challenging due to the complex and interdependent process parameters. To assist with the optimization, four machine learning algorithms utilizing gradient boosting decision trees were created to provide predictions for the mechanical and characterization properties of the composite parts. Approximately 200 random states from each algorithm were tested, and the models from each state were isolated and analyzed based on their accuracy, a validation process, and their feature importance. This analysis concluded that the models created from the machine learning algorithms could accelerate a parametric study for the AERoBOND process by rapidly optimizing process parameters to achieve desired performance characteristics.

Brennen M Middleton↗

DOE OSTI2023

Yet another parameter-free shape optimization method

The use of node coordinates as design variables in shape optimization offers a larger design space than computer-aided design (CAD)-based shape parameterizations. It also allows for the optimization of legacy designs, i.e., a finite element mesh from an existing design can be readily optimized to meet new performance requirements without involving a CAD model. However, it is well known that the node coordinate parameterization method is fraught with numerical difficulties, which makes it impractical to use. This has led to several of “parameter-free” shape optimization methods that seek the advantages and avoid the pitfalls of the naïve node coordinate parameterization method. These methods come in two main varieties: sensitivity filtering (or gradient smoothing) and consistent filtering. The latter is analogous to the density filter method used in topology optimization (TO). In this work, we use the PDE filter from TO and energy-based filters to implement consistent shape optimization filtering schemes easily and efficiently. Numerical experiments demonstrate that consistent methods are more robust than sensitivity filtering methods.

42 ENGINEERING↗

DOE OSTI2026

A variational framework for residual-based adaptivity in neural PDE solvers and operator learning

Residual-based adaptive strategies are widely used in scientific machine learning yet remain largely heuristic. We introduce a variational framework that formalizes these methods through convex transformations of the residual, where different transformations correspond to distinct objective functionals. For instance, exponential weights target uniform error minimization, while linear weights recover quadratic error minimization. This perspective reveals adaptive weighting as a means of selecting sampling distributions that optimize a primal objective, directly linking discretization choices to error metrics. This principled approach yields three key benefits: it enables systematic design of adaptive schemes, reduces discretization error by lowering estimator variance, and enhances learning dynamics by improving gradient signal-to-noise ratio. Extending the framework to operator learning, we demonstrate substantial performance gains across diverse optimizers and architectures. Our results provide a theoretical perspective for residual-based adaptivity and establish a foundation for principled discretization and training.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Shaping Micro-Bunched Electron Beams for Compact X-ray Free-Electron Lasers with Transverse Gradient Undulators

Laser-modulator-based micro-bunching of electron beams has been applied to many novel operating modes of X-ray free-electron lasers from harmonic generation to attosecond pulse production. Recently, it was also identified as a key enabling technology for the production of a compact XFEL driven by a relatively low-energy beam. In traditional laser modulator schemes with low-energy and high-current bunches, collective effects limit the possible working points that can be employed, and thus it is difficult to achieve optimal XFEL performance. We propose to utilize transverse longitudinal coupling in a transverse gradient undulator (TGU) to shape micro-bunched electron beams so as to optimize their performance in a compact X-ray free-electron laser. We show that a TGU added to a conventional laser modulator stage enables much more flexibility in the design, allowing one to generate longer micro-bunches less subject to slippage effects and even lower the slice emittance of the micro-bunches. We present a theoretical analysis of laser-based micro-bunching with an added TGU, simulation of compression with collective effects in such systems, and finally XFEL simulations demonstrating the gains in peak power enabled by the TGU. Although we focus on the application to compact XFELs, what we propose is a general phase space manipulation that may find utility in other applications as well.

eSASE↗

DOE OSTI2024

Co-free gradient lithium-rich cathode for high-energy batteries with optimized cyclability

Lithium-rich layered oxides (LLOs) hold the promise for high-energy battery cathodes. However, its application has been hindered by voltage decay associated with irreversible reactions at high voltages despite decades of intensive efforts. Here, we first theoretically studied the molecular orbitals of Mn-based Li-rich configurations. We found that the π-bond ring formed within the LiMn 6 structure could participate in stable redox reactions as one unit, but Co could disrupt its symmetry. We thus designed and synthesized Co-free concentration-gradient LLOs (CF-CG-LLOs) materials. The combination of concentration gradient and Co removal leads to exceptional capacity retention without any fading over 100 cycles of the pouch cell. More importantly, it exhibits an extraordinarily low voltage decay of 0.15 mV/cycle, accompanied by a high Coulombic efficiency of 99.86%. This concept and demonstration of CF-CG-LLO cathodes reveal a viable avenue toward low-cost, high-energy-density battery cathodes.

25 ENERGY STORAGE↗

DOE OSTI2022

A mixed, unified forward/inverse framework for earthquake problems: fault implementation and coseismic slip estimate

SUMMARY We introduce a new finite-element (FE) based computational framework to solve forward and inverse elastic deformation problems for earthquake faulting via the adjoint method. Based on two advanced computational libraries, FEniCS and hIPPYlib for the forward and inverse problems, respectively, this framework is flexible, transparent and easily extensible. We represent a fault discontinuity through a mixed FE elasticity formulation, which approximates the stress with higher order accuracy and exposes the prescribed slip explicitly in the variational form without using conventional split node and decomposition discrete approaches. This also allows the first order optimality condition, that is the vanishing of the gradient, to be expressed in continuous form, which leads to consistent discretizations of all field variables, including the slip. We show comparisons with the standard, pure displacement formulation and a model containing an in-plane mode II crack, whose slip is prescribed via the split node technique. We demonstrate the potential of this new computational framework by performing a linear coseismic slip inversion through adjoint-based optimization methods, without requiring computation of elastic Green’s functions. Specifically, we consider a penalized least squares formulation, which in a Bayesian setting—under the assumption of Gaussian noise and prior—reflects the negative log of the posterior distribution. The comparison of the inversion results with a standard, linear inverse theory approach based on Okada’s solutions shows analogous results. Preliminary uncertainties are estimated via eigenvalue analysis of the Hessian of the penalized least squares objective function. Our implementation is fully open-source and Jupyter notebooks to reproduce our results are provided. The extension to a fully Bayesian framework for detailed uncertainty quantification and non-linear inversions, including for heterogeneous media earthquake problems, will be analysed in a forthcoming paper.

58 GEOSCIENCES↗

DOE OSTI2023

Machine learning surrogate of physics-based building-stock simulator for end-use load forecasting

Building energy models are used to simulate heat and mass transfer and estimate end-use load in buildings. With the proliferation of solar photovoltaics on residential and commercial buildings, increasingly, buildings are expected to provide grid services, for which accurate and computationally efficient building energy simulations and end-use load prediction are imperative. Existing building energy simulation tools, however, have significant computational overhead that make them less practical in real-time deployment for optimization, design, uncertainty quantification and control in building energy management systems. Here this article presents a data-driven machine learning model based on light gradient boosting method (LightGBM) as a surrogate for a physics-based simulator for residential buildings to predict end-use load. The machine learning based surrogate model accounts for time-series related variables, seasonality and trend component of end-use load, and history of end-use load. The accuracy of the surrogate model is assessed on the prediction of the load profiles of 100 different houses in Cook County, Illinois, USA. The LightGBM surrogate model is shown to reduce the root-mean-squared error by 53% relative to a reference decision tree (DT) based model reported previously in the literature. Moreover, the model predicts the load spikes and high-ramp rate events throughout the year which are often the Achilles heel of other models in the literature. The machine learning based surrogate model is demonstrated to be computationally efficient, with a ten-fold reduction in the computational time compared to a physics-based building energy simulation, and suitable for uncertainty analysis and real-time control of building characteristics in response to uncertainty.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

NASA NTRS1999

Martian Aerocapture Terminal Point Guidance: A Reference Path Optimization Study

An effective method of terminal point guidance is to employ influence coefficients, which are solved from a set of differential equations adjoint to the linearized perturbations of the equations of motion about a reference trajectory. Hence, to optimize this type of guidance, one must first optimize the reference trajectory that the guidance is based upon. This study concentrates on various methods to optimize a reference trajectory for a Martian aerocapture maneuver, including a parametric analysis and first order gradient method. Resulting reference trajectories were tested in separate 2000 6-DOF Monte Carlo runs, using the Atmospheric Guidance Algorithm Testbed for the Mars Surveyor Program 2001 (MSP '01) Orbiter. These results were compared to an August 1998 study using the same terminal point control guidance algorithm and simulation testbed. Satisfactory improvements over the 1998 study are amply demonstrated.

Ro, Theodore U.↗

DOE OSTI2021

MCML: Combining physical constraints with experimental data for a multi-purpose meta-generalized gradient approximation

The predictive power of density functional theory for materials properties can be improved without increasing the overall computational complexity by extending the generalized gradient approximation (GGA) for electronic exchange and correlation to density functionals depending on the electronic kinetic energy density in addition to the charge density and its gradient, resulting in a meta-GGA. Here, we propose an empirical meta-GGA model that is based both on physical constraints and on experimental and quantum chemistry reference data. Overall, the resulting optimized meta-GGA MCML yields improved surface and gas phase reaction energetics without sacrificing the accuracy of bulk property predictions of existing meta-GGA approaches.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NASA NTRS1995

Performance analysis of a GPS Interferometric attitude determination system for a gravity gradient stabilized spacecraft

The performance of an unaided attitude determination system based on GPS interferometry is examined using linear covariance analysis. The modelled system includes four GPS antennae onboard a gravity gradient stabilized spacecraft, specifically the Air Force's RADCAL satellite. The principal error sources are identified and modelled. The optimal system's sensitivities to these error sources are examined through an error budget and by varying system parameters. The effects of two satellite selection algorithms, Geometric and Attitude Dilution of Precision (GDOP and ADOP, respectively) are examined. The attitude performance of two optimal-suboptimal filters is also presented. Based on this analysis, the limiting factors in attitude accuracy are the knowledge of the relative antenna locations, the electrical path lengths from the antennae to the receiver, and the multipath environment. The performance of the system is found to be fairly insensitive to torque errors, orbital inclination, and the two satellite geometry figures-of-merit tested.

Stoll, John C.↗

DOE OSTI2023

A Suite of Constitutive Promoters for Tuning Gene Expression in Plants

The need for convenient tools to express transgenes over a large dynamic range is pervasive throughout plant synthetic biology; however, current efforts are largely limited by the heavy reliance on a small set of strong promoters, precluding more nuanced and refined engineering endeavors in planta . To address this technical gap, we characterize a suite of constitutive promoters that span a wide range of transcriptional levels and develop a GoldenGate-based plasmid toolkit named PCONS, optimized for versatile cloning and rapid testing of transgene expression at varying strengths. We demonstrate how easy access to a stepwise gradient of expression levels can be used for optimizing synthetic transcriptional systems and the production of small molecules in planta . We also systematically investigate the potential of using PCONS as an internal standard in plant biology experimental design, establishing the best practices for signal normalization in experiments. Although our library has primarily been developed for optimizing expression in N. benthamiana , we demonstrate the translatability of our promoters across distantly related species using a multiplexed reporter assay with barcoded transcripts. In conclusion, our findings showcase the advantages of the PCONS library as an invaluable toolkit for plant synthetic biology.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Depth-Resolved Lithiated Gradients in Pristine and Laser-Ablated Anodes During Fast Charging

Laser ablating 3D electrode microstructures is a technique to improve Li-ion battery fast-charge performance. This technique has been theoretically proposed and electrochemically validated previously in the literature. The fundamental principle underlying laser ablation is that the ablated features reduce Li-ion transport pathways, improving access to the electrode active material near the current collector. This, in turn, promotes more homogeneous electrode utilization. The present study seeks to directly affirm the physics attributed to laser ablation using operando high-speed synchrotron X-ray diffraction. In this study, depth-resolved graphite lithiation gradients are measured operando during high-rate (15 min) charging. The depth-resolved lithiation dynamics of both ablated and non-ablated anodes are compared. The results highlight that the laser-ablated graphite electrode has notably more homogeneous utilization as compared to the non-ablated electrode. Additionally, the ablated electrode has a significant delay in reaching the maximum graphite lithiation at the separator, indicating less propensity for lithium plating. During low rate delithiation/discharge (2 hr), the two cells' lithiation gradients converge. Notably, a calibrated physics-based electrochemical model accurately reflects experimental findings, suggesting the potential to use pseudo-4D models not only to optimize laser ablation parameters in fast-charge capable electrodes but also to guide fast charging protocols that avoid lithium plating.

25 ENERGY STORAGE↗

DOE OSTI2020

Locally adaptive artificial viscosity strategies for Lagrangian hydrodynamics

To accurately model inviscid flow with shock waves using staggered-grid Lagrangian hydrodynamics, artificial viscosity is introduced to convert kinetic energy into internal energy, thereby providing a mechanism to generate the required entropy increase across shocks. In this paper, we propose a new method for constructing an adaptive, artificial viscosity in the context of one-dimensional, staggered-grid Lagrangian hydrodynamics. Our adaptive, artificial viscosity is defined in terms of two parameters that depend on density locally in a neighborhood around a shock, and hence, vary cell-by-cell. Our methodology is based on building a reference set of pre-computed optimal, globally constant artificial viscosity coefficients for a family of isolated shock test problems. For arbitrary flows, the evaluation of the unknown coefficients is automated by first estimating shock intensity locally, and second computing the corresponding adaptive parameter value via interpolation over the data from the pre-computed reference set of optimal, globally constant coefficient values. To illustrate the performance of our new approach, we compare our results against two existing methods. The first method is a limiter-based approach, which relies on estimating velocity gradients of the flow, and the second method is utilized in several commercial codes. Finally, we demonstrate that our new adaptive methodology produces more accurate results for a variety of tests with propagating shock waves, as well as for the aforementioned family of isolated shock problems.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2020

A Hybrid Gradient Method to Designing Bayesian Experiments for Implicit Models

Bayesian experimental design (BED) aims at designing an experiment to maximize the information gathering from the collected data. The optimal design is usually achieved by maximizing the mutual information (MI) between the data and the model parameters. When the analytical expression of the MI is unavailable, e.g.,having implicit models with intractable data distributions, a neural network-based lower bound of the MI was recently proposed and a gradient ascent method was used to maximize the lower bound [1]. However, the approach in [1] requires a pathwise sampling path to compute the gradient of the MI lower bound with respect to the design variables, and such a pathwise sampling path is usually inaccessible for implicit models. In this work, we propose a hybrid gradient approach that leverages recent advances in variational MI estimator and evolution strategies (ES)combined with black-box stochastic gradient ascent (SGA) to maximize the MI lower bound. This allows the design process to be achieved through a unified scalable procedure for implicit models without sampling path gradients. Several experiments demonstrate that our approach significantly improves the scalability of BED for implicit models in high-dimensional design space.

Zhang, Jiaxin↗