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 289 records · Page 16

TU-FG-BRB-04: A New Optimization Method for Pre-Treatment Patient-Specific Stopping-Power by Combining Proton Radiography and X-Ray CT

Purpose: The relative stopping power (RSP) uncertainty is the largest contributor to the range uncertainty in proton therapy. The purpose of this work is to develop a robust and systematic method that yields accurate patient specific RSPs by combining pre-treatment X-ray CT and daily proton radiography. Methods: The method is formulated as a penalized least squares optimization (PLSO) problem min(|Ax-B|). The matrix A represents the cumulative path-length crossed in each material computed by calculating proton trajectories through the X-ray CT. The material RSPs are denoted by x and B is the pRad, expressed as water equivalent thickness. The equation is solved using a convex-conic optimizer. Geant4 simulations were made to assess the feasibility of the method. RSP extracted from the Geant4 materials were used as a reference and the clinical HU-RSP curve as a comparison. The PLSO was first tested on a Gammex RMI-467 phantom. Then, anthropomorphic phantoms of the head, pelvis and lung were studied and resulting RSPs were evaluated. A pencil beam was generated in each phantom to evaluate the proton range accuracy achievable by using the optimized RSPs. Finally, experimental data of a pediatric head phantom (CIRS) were acquired using a recently completed experimental pCT scanner. Results: Numerical simulations showed precise RSP (<0.75%) for Gammex materials except low-density lung (LN-300) (1.2%). Accurate RSP have been obtained for the head (µ=−0.10%, 1.5σ=1.12%), lung (µ=−0.33, 1.5σ=1.02%) and pelvis anthropomorphic phantoms (µ=0.12, 1.5σ=0,99%). The range precision has been improved with an average R80 difference to the reference (µ±1.5σ) of −0.20±0.35%, −0.47±0.92% and −0.06±0.17% in the head, lung and pelvis phantoms respectively, compared to the 3.5% clinical margin. Experimental HU-RSP curve have been produced on the CIRS pediatric head. Conclusion: The proposed PLSO with prior knowledge X-ray CT shows promising potential (R80 σ<1.0% over all sites) to decrease the range uncertainty.

60 APPLIED LIFE SCIENCES↗

A procedure based on the Euler equations for correcting transonic wind tunnel wall interference

Based on an optimization formulation, a procedure has been developed to evaluate Mach number and angle-of-attack corrections. The Euler equations are assumed to be the flow governing equations. To obtain efficient solutions for the optimization problem, the iterative solutions for the flow variables and the design parameters are simultaneously updated. In addition to the model lift and geometry, the procedure requires pressure measurements near the tunnel walls. The accuracy and efficiency of several optimization techniques are investigated. The effect of perturbing certain test conditions on the residual interference is investigated.

Rizk, Magdi H.↗

Optimization-based scatter estimation using primary modulation for computed tomography

Purpose: Scatter reduces the image quality in computed tomography (CT), but scatter correction remains a challenge. A previously proposed primary modulation method simultaneously obtains the primary and scatter in a single scan. However, separating the scatter and primary in primary modulation is challenging because it is an underdetermined problem. In this study, an optimization-based scatter estimation (OSE) algorithm is proposed to estimate and correct scatter. Methods: In the concept of primary modulation, the primary is modulated, but the scatter remains smooth by inserting a modulator between the x-ray source and the object. In the proposed algorithm, an objective function is designed for separating the scatter and primary. Prior knowledge is incorporated in the optimization-based framework to improve the accuracy of the estimation: (1) the primary is always positive; (2) the primary is locally smooth and the scatter is smooth; (3) the location of penumbra can be determined; and (4) the scatter-contaminated data provide knowledge about which part is smooth. Results: The simulation study shows that the edge-preserving weighting in OSE improves the estimation accuracy near the object boundary. Simulation study also demonstrates that OSE outperforms the two existing primary modulation algorithms for most regions of interest in terms of the CT number accuracy and noise. The proposed method was tested on a clinical cone beam CT, demonstrating that OSE corrects the scatter even when the modulator is not accurately registered. Conclusions: The proposed OSE algorithm improves the robustness and accuracy in scatter estimation and correction. This method is promising for scatter correction of various kinds of x-ray imaging modalities, such as x-ray radiography, cone beam CT, and the fourth-generation CT.

60 APPLIED LIFE SCIENCES↗

On the determination of optimal costly measurement strategies for linear stochastic systems.

This paper presents the formulation of a class of optimization problems dealing with selecting, at each instant of time, one measurement provided by one out of many sensors. Each measurement has an associated measurement cost. The basic problem is then to select an optimal measurement policy, during a specified observation time interval, so that a weighted combination of prediction accuracy and accumulated observation cost is optimized. The current analysis is limited to the class of linear stochastic dynamic systems and measurement subsystems. The problem of selecting the optimal measurement strategy can be transformed into a deterministic optimal control problem. It is shown that the optimal measurement policy and the associated matched Kalman-type filter can be precomputed.

Athans, M.↗

OptiBench: An Optimization Benchmark Tool for Renewable Energy Problems

We propose a benchmark framework and visualization tool, OptiBench, for analyzing the performance of state-of-the-art optimization solvers across a variety of optimization problems in renewable energy research. Our framework is designed from the ground up in the Julia programming language and enables analysis at scale on high performance computing (HPC) systems. Our visualization tool allows effortless evaluation of optimization solver performance, robustness, and accuracy through intuitive plots, e.g., performance profiles, heat maps, and distribution plots. We have tested three benchmark suites relevant to the modeling of renewable energy systems, viz., CUTEst, PGLib-OPF, and WaterTAP water treatment optimization problems. We illustrate benchmarking of CUTEst using OptiBench on the National Renewable Energy Laboratory's (NREL) HPC Kestrel. Our findings indicate that MA57 HSL linear solver demonstrated the best overall performance for an experimental IPOPT implementation. Our work is ongoing and we intend to add support for more optimization solvers and benchmark test suites in the future.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

OptiBench: An Optimization Benchmark Tool for Renewable Energy Problems

We propose a benchmark framework and visualization tool, OptiBench, for analyzing the performance of state-of-the-art optimization solvers across a variety of optimization problems in renewable energy research. Our framework is designed from the ground up in the Julia programming language and enables analysis at scale on high performance computing (HPC) systems. Our visualization tool allows effortless evaluation of optimization solver performance, robustness, and accuracy through intuitive plots, e.g., performance profiles, heat maps, and distribution plots. We have tested three benchmark suites relevant to the modeling of renewable energy systems, viz., CUTEst, PGLib-OPF, and WaterTAP water treatment optimization problems. We illustrate benchmarking of CUTEst using OptiBench on the National Laboratory of the Rockies's (NLR) HPC Kestrel. Our findings indicate that MA57 HSL linear solver demonstrated the best overall performance for an experimental IPOPT implementation. Our work is ongoing and we intend to add support for more optimization solvers and benchmark test suites in the future.

97 MATHEMATICS AND COMPUTING↗

TPSAS-NF1676L-33512-DND

INTRODUCTION: In an effort to quantify the predictive capability of the Livermore Software Technology Corporation (LSTC)-provided Hybrid III (small female, mid-size male, and large male) anthropomorphic test device (ATD) finite element models (FEMs) within aerospace loading environments, the National Aeronautics and Space Administration (NASA) performed a series of sled tests with the isolated head-neck complex of each ATD. These tests and subsequent modeling efforts revealed significant deficiencies in the large male head neck FEM. Differences in both the shape and material properties between the FEM and the physical ATD were identified. In order to improve the predictive response of the FEM, a model development effort was undertaken to update the head-neck complex to better represent the geometric and material properties of the physical ATD.METHODS: To better match the physical geometry, a new neck component model was meshed and implemented into the Hybrid III large male head-neck complex. The updated geometry and weight were verified against measurements of the physical ATD to insure accuracy. The primary source of material discrepancy between the physical ATD and FEM was found to be in the material definition of the rubber neck pucks. Using LS-Opt the material properties of this part were calibrated to match physical response of the head-neck complex in the isolated head-neck testing performed. RESULTS AND DISCUSSION: The updated geometry and calibrated material properties were shown to significantly improve the FEM predictive accuracy under lateral, rear, and frontal impacts with combined horizontal and vertical loading. The results of this study demonstrate an effective means for improving ATD FEM response through isolated component level testing and material parameter optimization. The improved and quantified accuracy of the Hybrid III large male FEM head-neck complex stemming from with work lends to confidence in its future use for occupant protection evaluation of large male occupants

J Putnam↗

Integrating Natural Language Processing and Machine Learning Algorithms to Categorize Oncologic Response in Radiology Reports

A significant volume of medical data remains unstructured. Natural language processing (NLP) and machine learning (ML) techniques have shown to successfully extract insights from radiology reports. However, the codependent effects of NLP and ML in this context have not been well-studied. Between April 1, 2015 and November 1, 2016, 9418 cross-sectional abdomen/pelvis CT and MR examinations containing our internal structured reporting element for cancer were separated into four categories: Progression, Stable Disease, Improvement, or No Cancer. We combined each of three NLP techniques with five ML algorithms to predict the assigned label using the unstructured report text and compared the performance of each combination. The three NLP algorithms included term frequency-inverse document frequency (TF-IDF), term frequency weighting (TF), and 16-bit feature hashing. The ML algorithms included logistic regression (LR), random decision forest (RDF), one-vs-all support vector machine (SVM), one-vs-all Bayes point machine (BPM), and fully connected neural network (NN). The best-performing NLP model consisted of tokenized unigrams and bigrams with TF-IDF. Increasing N-gram length yielded little to no added benefit for most ML algorithms. With all parameters optimized, SVM had the best performance on the test dataset, with 90.6 average accuracy and F score of 0.813. The interplay between ML and NLP algorithms and their effect on interpretation accuracy is complex. The best accuracy is achieved when both algorithms are optimized concurrently.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

High-throughput generation, optimization and analysis of genome-scale metabolic models.

Genome-scale metabolic models have proven to be valuable for predicting organism phenotypes from genotypes. Yet efforts to develop new models are failing to keep pace with genome sequencing. To address this problem, we introduce the Model SEED, a web-based resource for high-throughput generation, optimization and analysis of genome-scale metabolic models. The Model SEED integrates existing methods and introduces techniques to automate nearly every step of this process, taking {approx}48 h to reconstruct a metabolic model from an assembled genome sequence. We apply this resource to generate 130 genome-scale metabolic models representing a taxonomically diverse set of bacteria. Twenty-two of the models were validated against available gene essentiality and Biolog data, with the average model accuracy determined to be 66% before optimization and 87% after optimization.

59 BASIC BIOLOGICAL SCIENCES↗

Fabrication and characterization of a full-size ultra-precise lamellar grating for the Cosmic beamline at ALS-U

We have developed a new process for the production of ultra-precise variable line spacing (VLS) lamellar diffraction gratings through nanofabrication. The process enables the fabrication of full-size X-ray gratings with sub-nanometre accuracy in groove depth, an optimal land-to-groove ratio, and uniform groove depth across the entire grating area. We also established a method for evaluating VLS groove density variation using stitched Fizeau interferometry. The measurements confirmed the exceptionally high accuracy of the VLS groove density in the fabricated gratings, which is well within the specification tolerances while the residual groove density errors are vanishingly small. The gold-coated grating demonstrated near-theoretical diffraction efficiency across the energy range of 100–1200 eV.

36 MATERIALS SCIENCE↗

Automation of POST Cases via External Optimizer and "Artificial p2" Calculation

During conceptual design speed and accuracy are often at odds. Specifically in the realm of launch vehicles, optimizing the ascent trajectory requires a larger pool of analytical power and expertise. Experienced analysts working on familiar vehicles can produce optimal trajectories in a short time frame, however whenever either "experienced" or "familiar " is not applicable the optimization process can become quite lengthy. In order to construct a vehicle agnostic method an established global optimization algorithm is needed. In this work the authors develop an "artificial" error term to map arbitrary control vectors to non-zero error by which a global method can operate. Two global methods are compared alongside Design of Experiments and random sampling and are shown to produce comparable results to analysis done by a human expert.

Dees, Patrick D.↗

Throughput-Oriented and Accuracy-Aware DNN Training with BFloat16 on GPU

Deep Neural Networks (DNNs) have transformed the field of artificial intelligence and achieved extraordinary success in many areas. The training of DNNs is commonly compute and memory-intensive, which has resulted in several optimizations in the training phase. Among them, reduced precision is a typical and widely used technique to accelerate DNN training and reduce memory requirements. However, applying a widely adopted reduced precision format such as Float16 to all involved operations in DNN training is not optimal as the use of Float16 in some operations can hurt model accuracy. Meanwhile, additional optimizations including loss scaling and autocast techniques can mitigate the accuracy loss but lead to inherent overhead and inadequate use of reduced precision. In this work, we leverage another reduced precision format, BFloat16, and introduce a throughput-oriented and accuracy-aware approach to maximize the performance potential of DNN training. Since the high throughput provided by BFloat16 format is accompanied by low precision of the floating-point representation, this approach achieves high throughput by using BFloat16 on all DNN operations and avoids the accuracy loss through a customized accuracy-aware normalization. Results show that our approach outperforms the state-of-the-art mixed-precision training by 1.21x on an NVIDIA A100 GPU.

Xie, Zhen↗

Efficiency Optimization of Dual Active Bridge Converter Based on dV/dt Snubber Capacitors

This paper presents a lossless regenerative dV/dt snubber circuit for PWM converters to achieve high-efficiency high power-density without significant cost and reliability penalties. The dV/dt snubber employs lossless capacitors for each MOSFET device in a converter to provide an additional path to store the switching energy during turn-off transient and decrease the actual turn-off loss of the MOSFET channel. A novel mathematical turn-off loss model is built to separate the actual turn-off loss of the MOSFET channel and the recycling switching energy stored in the total equivalent output capacitance including Coss, the dV/dt snubber, and etc. Necessary snubber optimization strategy is proposed in order to balance the turn-off loss with ZVS operation range and deadtime conduction loss, and thereby achieving an optimal efficiency performance. To prove the validity and accuracy of the novel turn-off loss model and optimization strategy, the dV/dt snubber has been incorporated into a 1.3kV/200kW DC/DC and a 1kV/70kVA DC/AC PWM converters. Experimental results are given to demonstrate the validity of the proposed dV/dt snubber capacitors and optimization strategy.

Xu, Wei↗

ARENA: Adversary-Resistant Evolving Neural Architectures

Neural networks are becoming the cornerstone for national security prediction tasks. However, designing them requires significant research and trial/error, as they have many hyperparameters, including their computation graph (“architecture”). Neural architecture search (NAS) employs secondary optimizers to search for architectures maximizing objectives like accuracy. Evolutionary algorithms (EAs) are the most used class of optimizer for NAS. However, existing Python libraries for writing EAs limit the complexity of experiments a user can design. In this project, we built ARENA, a Python framework that encodes complex, hyper-realistic EAs. ARENA collects detailed information as it runs and is flexible enough to encode non-EA search algorithms. We tested ARENA on 4 toy optimization problems by encoding 3 search algorithms for each—random search, an EA, and simulated annealing. We also designed an EA that performs NAS on the MNIST dataset. Our experiments suggest the potential for immediate mission impact through solving lab-wide optimization problems.

97 MATHEMATICS AND COMPUTING↗

Development of Fuel Shuffling Module for PHISICS

PHISICS (Parallel and Highly Innovative Simulation for the INL Code System) [4] code toolkit has been in development at the Idaho National Laboratory. This package is intended to provide a modern analysis tool for reactor physics investigation. It is designed with the mindset to maximize accuracy for a given availability of computational resources and to give state of the art tools to the modern nuclear engineer. This is obtained by implementing several different algorithms and meshing approaches among which the user will be able to choose, in order to optimize his computational resources and accuracy needs. The software is completely modular in order to simplify the independent development of modules by different teams and future maintenance. The package is coupled with the thermo-hydraulic code RELAP5-3D [3]. In the following the structure of the different PHISICS modules is briefly recalled, focusing on the new shuffling module (SHUFFLE), object of this paper.

97 MATHEMATICS AND COMPUTING↗

HLS4ML Integration with QICK

The QICK (Quantum Instrumentation Control Kit) integration aims to enhance the readout of superconducting qubits by leveraging machine learning (ML) techniques. These techniques offer high accuracy, increased speed, and better state preservation for qubit readouts. The integration process employs neural network algorithms to optimize system performance and achieve high accuracy rates. The QICK board uses an FPGA, allowing for efficient real-time processing and high-performance execution of machine learning algorithms.

Ali, Mohamud↗