Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “optimal accuracy”

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 469 records · Page 26

A Generalized Variational Principle with Applications to Excited State Mean Field Theory

We present a generalization of the variational principle that is compatible with any Hamiltonian eigenstate that can be specified uniquely by a list of properties. This variational principle appears to be compatible with a wide range of electronic structure methods, including mean field theory, density functional theory, multireference theory, and quantum Monte Carlo. Like the standard variational principle, this generalized variational principle amounts to the optimization of a nonlinear function that, in the limit of an arbitrarily flexible wave function, has the desired Hamiltonian eigenstate as its global minimum. Unlike the standard variational principle, it can target excited states and select individual states in cases of degeneracy or near-degeneracy. As an initial demonstration of how this approach can be useful in practice, we employ it to improve the optimization efficiency of excited state mean field theory by an order of magnitude. With this improved optimization, we are able to demonstrate that the accuracy of the corresponding second-order perturbation theory rivals that of singles-and-doubles equation-of-motion coupled cluster in a substantially broader set of molecules than could be explored by our previous optimization methodology.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Photo-z optimization for measurements of the BAO radial scale

Baryon Acoustic Oscillations (BAO) in the radial direction offer a method to directly measure the Universe expansion history, and to set limits to space curvature when combined to the angular BAO signal. In addition to spectroscopic surveys, radial BAO might be measured from accurate enough photometric redshifts obtained with narrow-band filters. We explore the requirements for a photometric survey using Luminous Red Galaxies (LRG) to competitively measure the radial BAO signal and discuss the possible systematic errors of this approach. Using the LRG spectra obtained from SDSS, we find that the optimal system consists of {approx} 30 filters of width {Delta}{lambda}/{lambda} {approx} 0.02. A S/N>20 is generally necessary at the filters on the red side of the Balmer-break to reach the target photometric accuracy {sigma}{sub z} = 0.003(1+z) for measuring radial BAO. The presence of outliers can significantly degrade this: in any practical application, being able to efficiently remove outliers may require ancillary observations. We estimate that a 5-year survey in a dedicated telescope with etendue in excess of 60 m{sup 2} deg{sup 2} would be necessary to obtain a high enough density of galaxies to measure the radial BAO with shot noise smaller than cosmic variance up to z = 0.85. We conclude that spectroscopic surveys have a superior performance than photometric ones for measuring BAO in the radial direction.

79 ASTRONOMY AND ASTROPHYSICS↗

Electron-optical systems for planar gyrotrons

The methodology of designing an electron-optical system (EOS) that forms sheet helical electron beams (HEBs) for high-power gyrotrons is developed. As an example, we consider the EOS for a 140-GHz gyrotron operated at the first harmonic of the cyclotron frequency with an accelerating voltage of 50 kV, a beam current of 30 A, and a magnetic field compression of 36. A planar geometry of the magnetron-injection gun (MIG) is suggested. The adiabatic theory of MIGs modified for the planar geometry of EOS is used for preliminary estimations of MIG parameters. Numerical simulation of the HEB properties based on the CST STUDIO SUITE 3D code is performed to find the optimal configuration of a planar MIG. The accuracy of the calculated data is discussed. The main factors that affect the HEB quality are considered. It is shown that a sheet HEB with a pitch-factor of 1.3 and velocity spread not exceeding 25%–30% can be formed; this is quite acceptable for high-efficiency operation of modern gyrotrons. Calculation of the beam-wave interaction with the obtained HEB parameters proved that a high output power with a sufficiently good efficiency of about 20% can be reached. Simulations show the feasibility of the experimental implementation of a novel planar EOS and its use in short-wave planar gyrotrons. The developed technique can be used for the study and optimization of planar gyrotrons of different frequency bands and power levels.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

TU-EF-304-07: Monte Carlo-Based Inverse Treatment Plan Optimization for Intensity Modulated Proton Therapy

Purpose: Intensity-modulated proton therapy (IMPT) is increasingly used in proton therapy. For IMPT optimization, Monte Carlo (MC) is desired for spots dose calculations because of its high accuracy, especially in cases with a high level of heterogeneity. It is also preferred in biological optimization problems due to the capability of computing quantities related to biological effects. However, MC simulation is typically too slow to be used for this purpose. Although GPU-based MC engines have become available, the achieved efficiency is still not ideal. The purpose of this work is to develop a new optimization scheme to include GPU-based MC into IMPT. Methods: A conventional approach using MC in IMPT simply calls the MC dose engine repeatedly for each spot dose calculations. However, this is not the optimal approach, because of the unnecessary computations on some spots that turned out to have very small weights after solving the optimization problem. GPU-memory writing conflict occurring at a small beam size also reduces computational efficiency. To solve these problems, we developed a new framework that iteratively performs MC dose calculations and plan optimizations. At each dose calculation step, the particles were sampled from different spots altogether with Metropolis algorithm, such that the particle number is proportional to the latest optimized spot intensity. Simultaneously transporting particles from multiple spots also mitigated the memory writing conflict problem. Results: We have validated the proposed MC-based optimization schemes in one prostate case. The total computation time of our method was ∼5–6 min on one NVIDIA GPU card, including both spot dose calculation and plan optimization, whereas a conventional method naively using the same GPU-based MC engine were ∼3 times slower. Conclusion: A fast GPU-based MC dose calculation method along with a novel optimization workflow is developed. The high efficiency makes it attractive for clinical usages.

60 APPLIED LIFE SCIENCES↗

Machine Learning-Based Temperature Prediction for Runtime Thermal Management Across System Components

Elevated temperatures limit the peak performance of systems because of frequent interventions by thermal throttling. Non-uniform thermal states across system nodes also cause performance variation within seemingly equivalent nodes leading to significant degradation of overall performance. In this paper we present a framework for creating a lightweight thermal prediction system suitable for run-time management decisions. We pursue two avenues to explore optimized lightweight thermal predictors. First, we use feature selection algorithms to improve the performance of previously designed machine learning methods. Second, we develop alternative methods using neural network and linear regression-based methods to perform a comprehensive comparative study of prediction methods. We show that our optimized models achieve improved performance with better prediction accuracy and lower overhead as compared with the Gaussian process model proposed previously. Specifically we present a reduced version of the Gaussian process model, a neural network-based model, and a linear regression-based model. Using the optimization methods, we are able to reduce the average prediction errors in the Gaussian process from 4: 2 degrees C to 2: 9 degrees C. We also show that the newly developed models using neural network and Lasso linear regression have average prediction errors of 2: 9 degrees C and 3: 8 degrees C respectively. The prediction overheads are 0.22, 0.097, and 0.026 ms per prediction for reduced Gaussian process, neural network, and Lasso linear regression models, respectively, compared with 0.57 ms per prediction for the previous Gaussian process model. We have implemented our proposed thermal prediction models on a two-node system configuration to help identify the optimal task placement. The task placement identified by the models reduces the average system temperature by up to 11: 9 degrees C without any performance degradation. Furthermore, these models respectively achieve 75, 82.5, and 74.17 percent success rates in correctly pointing to those task placements with better thermal response, compared with 72.5 percent success for the original model in achieving the same objective. Lastly, we extended our analysis to a 16-node system and we were able to train models and execute them in real time to guide task migration and achieve on average 17 percent reduction in the overall system cooling power.

97 MATHEMATICS AND COMPUTING↗

Dynamical Sketching for Enhanced Communication Efficiency in Federated Learning

Federated learning (FL) has revolutionized distributed machine learning by enabling collaborative model training without sharing local data. However, communication efficiency and privacy guarantees remain significant challenges. This paper introduces a dynamic sketching mechanism in FL, optimizing the trade-off between communication efficiency and model accuracy. By dynamically selecting the sketch matrix size, our approach adapts to the evolving characteristics of the data and the model, ensuring optimal performance across diverse scenarios. We leverage Bayesian optimization to systematically tune the sketch parameters, achieving an effective balance between resource efficiency and model performance. Experimental results on the MNIST dataset using a convolutional neural network (CNN) architecture validate the proposed method's efficiency and scalability. Our dynamic sketching approach significantly outperforms fixed-size sketching techniques, achieving higher compression ratios (up to 62x) and providing better privacy guarantees while maintaining high model accuracy. These findings highlight the robustness and versatility of our approach and make it a valuable solution for privacy-preserving, communication-efficient federated learning.

Afrose, Sharmin [ORNL]↗

Robustness, generality and efficiency of optimization algorithms in practical applications

The theoretical foundations of two approaches, sequential quadratic programming (SQP) and optimality criteria (OC), are analyzed and compared, with emphasis on the critical importance of parameters such as accuracy, generality, robustness, efficiency, and ease of use in large scale structural optimization. A simplified fighter wing and active control of space structures are considered with other example problems. When applied to general system identification problems, the OC methods are shown to lose simplicity and demonstrate lack of generality, accuracy and robustness. It is concluded that the SQP method with a potential constraint strategy is a better choice as compared to the currently prevalent mathematical programming and OC approaches.

Thanedar, P. B.↗

SU-E-T-21: A Novel Sampling Algorithm to Reduce Intensity-Modulated Radiation Therapy (IMRT) Optimization Time

Purpose: The IMRT optimization problem requires substantial computer time to find optimal dose distributions because of the large number of variables and constraints. Voxel sampling reduces the number of constraints and accelerates the optimization process, but usually deteriorates the quality of the dose distributions to the organs. We propose a novel sampling algorithm that accelerates the IMRT optimization process without significantly deteriorating the quality of the dose distribution. Methods: We included all boundary voxels, as well as a sampled fraction of interior voxels of organs in the optimization. We selected a fraction of interior voxels using a clustering algorithm, that creates clusters of voxels that have similar influence matrix signatures. A few voxels are selected from each cluster based on the pre-set sampling rate. Results: We ran sampling and no-sampling IMRT plans for de-identified head and neck treatment plans. Testing with the different sampling rates, we found that including 10% of inner voxels produced the good dose distributions. For this optimal sampling rate, the algorithm accelerated IMRT optimization by a factor of 2–3 times with a negligible loss of accuracy that was, on average, 0.3% for common dosimetric planning criteria. Conclusion: We demonstrated that a sampling could be developed that reduces optimization time by more than a factor of 2, without significantly degrading the dose quality.

60 APPLIED LIFE SCIENCES↗

Towards computational materials design from first principles using alchemical changes and derivatives.

The design of new materials with specific physical, chemical, or biological properties is a central goal of much research in materials and medicinal sciences. Except for the simplest and most restricted cases brute-force computational screening of all possible compounds for interesting properties is beyond any current capacity due to the combinatorial nature of chemical compound space (set of stoichiometries and configurations). Consequently, when it comes to computationally optimizing more complex systems, reliable optimization algorithms must not only trade-off sufficient accuracy and computational speed of the models involved, they must also aim for rapid convergence in terms of number of compounds 'visited'. I will give an overview on recent progress on alchemical first principles paths and gradients in compound space that appear to be promising ingredients for more efficient property optimizations. Specifically, based on molecular grand canonical density functional theory an approach will be presented for the construction of high-dimensional yet analytical property gradients in chemical compound space. Thereafter, applications to molecular HOMO eigenvalues, catalyst design, and other problems and systems shall be discussed.

99 GENERAL AND MISCELLANEOUS↗

Active operator learning with predictive uncertainty quantification for partial differential equations

With the increased prevalence of neural operators being used to provide rapid solutions to partial differential equations (PDEs), understanding the accuracy of model predictions and the associated error levels is necessary for deploying reliable surrogate models in scientific applications. Existing uncertainty quantification (UQ) frameworks employ ensembles or Bayesian methods, which can incur substantial computational costs during both training and inference. Here, we propose a lightweight predictive UQ method tailored for Deep operator networks (DeepONets) that also generalizes to other operator networks. Numerical experiments on linear and nonlinear PDEs demonstrate that the framework’s uncertainty estimates are unbiased and provide accurate out-of-distribution uncertainty predictions with a sufficiently large training dataset. Our framework provides fast inference and uncertainty estimates that can efficiently drive outer-loop analyses that would be prohibitively expensive with conventional solvers. We demonstrate how predictive uncertainties can be used in the context of Bayesian optimization and active learning problems to yield improvements in accuracy and data-efficiency for outer-loop optimization procedures. In the active learning setup, we extend the framework to Fourier Neural Operators (FNO) and describe a generalized method for other operator networks. To enable real-time deployment, we introduce an inference strategy based on precomputed trunk outputs and a sparse placement matrix, reducing evaluation time by more than a factor of five. Our method provides a practical route to uncertainty-aware operator learning in time-sensitive settings.

97 MATHEMATICS AND COMPUTING↗

Sloan Digital Sky Survey III photometric quasar clustering: Probing the initial conditions of the Universe

Here, the Sloan Digital Sky Survey has surveyed 14,555 square degrees of the sky, and delivered over a trillion pixels of imaging data. We present the large-scale clustering of 1.6 million quasars between z=0.5 and z=2.5 that have been classified from this imaging, representing the highest density of quasars ever studied for clustering measurements. This data set spans 0~ 11,00 square degrees and probes a volume of 80 h –3 Gpc 3 . In principle, such a large volume and medium density of tracers should facilitate high-precision cosmological constraints. We measure the angular clustering of photometrically classified quasars using an optimal quadratic estimator in four redshift slices with an accuracy of ~ 25% over a bin width of δ l ~ 10–15 on scales corresponding to matter-radiation equality and larger (0ℓ ~ 2–3).

79 ASTRONOMY AND ASTROPHYSICS↗

Even Higher-Level Synthesis: An Exploration of AI Hardware Accelerators using HLS4ML

With the rise of artificial intelligence, the popularization of deep learning, and a constantly evolving industry, the demand for flexible and efficient tools has never been greater. As algorithms grow more complex, their runtime and energy consumption increase exponentially. Customized hardware accelerators, long used for specific mathematical operations, remain essential for managing modern applications' computational and power demands. Hardware accelerators can speed up complex computations by orders of magnitude, but their manual design and verification processes are often challenging and time-consuming. High-Level Synthesis (HLS) provides a solution by transforming high-level algorithm descriptions, typically written in C++ or SystemC, into synthesizable RTL suitable for hardware implementation. This approach reduces development time for RTL engineers while offering flexibility beyond what traditional handwritten RTL can provide. We extended this capability to the machine-learning domain with the open-source framework hls4ml, which allows neural networks trained in Python frameworks like Tensorflow or PyTorch to be synthesized into efficient hardware representations for the traditional FPGA and ASIC flows. This breakthrough addresses the growing need for reduced design turnaround and easy verification of ML hardware accelerators with low latency and power efficiency constraints. During this tutorial, we will demonstrate how Python complements HLS by simplifying the ML design process, bridging the gap between software and hardware development. Attendees will explore how we translate neural networks modeled in Python into fixed-point C++ models suitable for HLS workflows. We will dive into strategies like Value-Range Analysis and Quantization-Aware Training, which optimize these designs for deployment and evaluate their accuracy, power consumption, and energy efficiency. To exemplify these concepts, experts from Fermilab will share their experiences applying this technology to high-energy physics experiments, where real-time, low-latency processing is critical. Over the years, Fermilab engineers have demonstrated how deep neural networks, optimized for hardware using hls4ml, can meet the stringent requirements of trigger systems at the CERN Large Hadron Collider. These systems rely on rapid decision-making to process immense data volumes while retaining only the most relevant events for further analysis. The application of hls4ml has also been extended to innovative technologies like smart pixel arrays. These smart pixels integrate ML inference capabilities directly into sensor devices, enabling localized data processing at the pixel level. This approach drastically reduces the need to transmit raw data to external processing units, significantly decreasing power consumption and latency. By embedding neural networks within the pixel architecture, the smart pixels can identify and prioritize relevant data in real time, providing a highly efficient solution for edge computing in scenarios such as particle detectors and imaging systems. Fermilab's work highlights the potential of hardware-accelerated ML in scenarios where both speed and power efficiency are mission-critical. Through this tutorial, attendees will gain valuable insights into the challenges and solutions of deploying ML in hardware. Understanding how HLS and hls4ml streamline the development of neural network-based hardware accelerators is fundamental for the industry's future. Participants will learn how these technologies are shaping the future of AI and scientific computing.

Di Guglielmo, Giuseppe [Fermilab]↗

Designing Ti-6Al-4V microstructure for strain delocalization using neural networks

Abstract The deformation behavior of Ti-6Al-4V titanium alloy is significantly influenced by slip localized within crystallographic slip bands. Experimental observations reveal that intense slip bands in Ti-6Al-4V form at strains well below the macroscopic yield strain and may serially propagate across grain boundaries, resulting in long-range localization that percolates through the microstructure. These connected, localized slip bands serve as potential sites for crack initiation. Although slip localization in Ti-6Al-4V is known to be influenced by various factors, an investigation of optimal microstructures that limit localization remains lacking. In this work, we develop a novel strategy that integrates an explicit slip band crystal plasticity technique, graph networks, and neural network models to identify Ti-6Al-4V microstructures that reduce the propensity for strain localization. Simulations are conducted on a dataset of 3D polycrystals, each represented as a graph to account for grain neighborhood and connectivity. The results are then used to train neural network surrogate models that accurately predict localization-based properties of a polycrystal, given its microstructure. These properties include the ratio of slip accumulated in the band to that in the matrix, fraction of total applied strain accommodated by slip bands, and spatial connectivity of slip bands throughout the microstructure. The initial dataset is enriched by synthetic data generated by the surrogate models, and a grid search optimization is subsequently performed to find optimal microstructures. Describing a 3D polycrystal with only a few features and a combination of graph and neural network models offer robustness compared to the alternative approaches without compromising accuracy. We show that while each material property is optimized through a unique microstructure solution, elongated grain shape emerges as a recurring feature among all optimal microstructures. This finding suggests that designing microstructures with elongated grains could potentially mitigate strain localization without compromising strength.

Ahmadikia, Behnam↗

Development of GPU-based Monte Carlo code for fast CT imaging dose calculation on CUDA Fermi architecture

This paper describes the development of a Graphics Processing Unit (GPU) accelerated Monte Carlo photon transport code, ARCHER{sub GPU}, to perform CT imaging dose calculations with good accuracy and performance. The code simulates interactions of photons with heterogeneous materials. It contains a detailed CT scanner model and a family of patient phantoms. Several techniques are used to optimize the code for the GPU architecture. In the accuracy and performance test, a 142 kg adult male phantom was selected, and the CT scan protocol involved a whole-body axial scan, 20-mm x-ray beam collimation, 120 kVp and a pitch of 1. A total of 9 x 108 photons were simulated and the absorbed doses to 28 radiosensitive organs/tissues were calculated. The average percentage difference of the results obtained by the general-purpose production code MCNPX and ARCHER{sub GPU} was found to be less than 0.38%, indicating an excellent agreement. The total computation time was found to be 8,689, 139 and 56 minutes for MCNPX, ARCHER{sub CPU} (6-core) and ARCHER{sub GPU}, respectively, indicating a decent speedup. Under a recent grant funding from the NIH, the project aims at developing a Monte Carlo code with the capability of sub-minute CT organ dose calculations. (authors)

61 RADIATION PROTECTION AND DOSIMETRY↗

Impact of leaf motion constraints on IMAT plan quality, deliver accuracy, and efficiency

Purpose: Intensity modulated arc therapy (IMAT) is a radiation therapy delivery technique that combines the efficiency of arc based delivery with the dose painting capabilities of intensity modulated radiation therapy (IMRT). A key challenge in developing robust inverse planning solutions for IMAT is the need to account for the connectivity of the beam shapes as the gantry rotates from one beam angle to the next. To overcome this challenge, inverse planning solutions typically impose a leaf motion constraint that defines the maximum distance a multileaf collimator (MLC) leaf can travel between adjacent control points. The leaf motion constraint ensures the deliverability of the optimized plan, but it also impacts the plan quality, the delivery accuracy, and the delivery efficiency. In this work, the authors have studied leaf motion constraints in detail and have developed recommendations for optimizing the balance between plan quality and delivery efficiency. Methods: Two steps were used to generate optimized IMAT treatment plans. The first was the direct machine parameter optimization (DMPO) inverse planning module in the Pinnacle{sup 3} planning system. Then, a home-grown arc sequencer was applied to convert the optimized intensity maps into deliverable IMAT arcs. IMAT leaf motion constraints were imposed using limits of between 1 and 30 mm/deg. Dose distributions were calculated using the convolution/superposition algorithm in the Pinnacle{sup 3} planning system. The IMAT plan dose calculation accuracy was examined using a finer sampling calculation and the quality assurance verification. All plans were delivered on an Elekta Synergy with an 80-leaf MLC and were verified using an IBA MatriXX 2D ion chamber array inserted in a MultiCube solid water phantom. Results: The use of a more restrictive leaf motion constraint (less than 1-2 mm/deg) results in inferior plan quality. A less restrictive leaf motion constraint (greater than 5 mm/deg) results in improved plan quality but can lead to less accurate dose distribution as evidenced by increasing discrepancies between the planned and the delivered doses. For example, the results from our patient-specific quality assurance measurements demonstrated that the average gamma analysis passing rate decreased from 98% to 80% when the allowable leaf motion increased from 3 to 20 mm/deg. Larger leaf motion constraints also led to longer treatment delivery times (2 to 4 folds) due to the additional MLC leaf motion. Conclusions: Leaf motion constraints significantly impact IMAT plans in terms of plan quality, delivery accuracy, and delivery efficiency with the impact magnified for more complex cases. Our studies indicate that a leaf motion constraint of 2 to 3 mm/deg of gantry rotation can provide an optimal balance between plan quality, delivery accuracy, and efficiency.

61 RADIATION PROTECTION AND DOSIMETRY↗

Application of numerical optimization to rotor aerodynamic design

Based on initial results obtained from the performance optimization code, a number of observations can be made regarding the utility of optimization codes in supporting design of rotors for improved performance. (1) The primary objective of improving the productivity and responsiveness of current design methods can be met. (2) The use of optimization allows the designer to consider a wider range of design variables in a greatly compressed time period. (3) Optimization requires the user to carefully define his problem to avoid unproductive use of computer resources. (4) Optimization will increase the burden on the analyst to validate designs and to improve the accuracy of analysis methods. (5) Direct calculation of finite difference derivatives by the optimizer was not prohibitive for this application but was expensive. Approximate analysis in some form would be considered to improve program response time. (6) Program developement is not complete and will continue to evolve to integrate new analysis methods, design problems, and alternate optimizer options.

Pleasants, W. A., III↗

Proton beam scattering system optimization for clinical and research applications

Purpose: To develop and test a method for optimizing and constructing a dual scattering system in passively scattered proton therapy. Methods: A beam optics optimization algorithm was developed to optimize the thickness of the first scatterer (S1) and the profile (of both the high-Z material and Lexan) of the second scatterer (S2) to deliver a proton beam matching a given set of parameters, including field diameter, fluence, flatness, and symmetry. A new manufacturing process was also tested that allows the contoured second scattering foil to be created much more economically and quickly using Cerrobend casting. Two application-specific scattering systems were developed and tested using both experimental and Monte Carlo techniques to validate the optimization process described. Results: A scattering system was optimized and constructed to deliver large uniform irradiations of radiobiology samples at low dose rates. This system was successfully built and tested using film and ionization chambers. The system delivered a uniform radiation field of 50 cm diameter (to a dose of {+-}7% of the central axis) while the depth dose profile could be tuned to match the specifications of the particular investigator using modulator wheels and range shifters. A second scattering system for intermediate field size (4 cm < diameter < 10 cm) stereotactic radiosurgery and radiation therapy (SRS and SRT) treatments was also developed and tested using GEANT4. This system improved beam efficiency by over 70% compared with existing scattering systems while maintaining field flatness and depth dose profile. In both cases the proton range uniformity across the radiation field was maintained, further indicating the accuracy of the energy loss formalism in the optimization algorithm. Conclusions: The methods described allow for rapid prototyping of scattering foils to meet the demands of both research and clinical beam delivery applications in proton therapy.

60 APPLIED LIFE SCIENCES↗