Search NASA⌕ Search

SEARCH · Search NASA

Results for “parameter selection”

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 19 records

A Reinforcement Learning Approach to Parameter Selection for Distributed Optimal Power Flow

With the increasing penetration of distributed energy resources, distributed optimization algorithms have attracted significant attention for power systems applications due to their potential for superior scalability, privacy, and robustness to a single point-of-failure. The Alternating Direction Method of Multipliers (ADMM) is a popular distributed optimization algorithm; however, its convergence performance is highly dependent on the selection of penalty parameters, which are usually chosen heuristically. In this work, we use reinforcement learning (RL) to develop an adaptive penalty parameter selection policy for alternating current optimal power flow (ACOPF) problem solved via ADMM with the goal of minimizing the number of iterations until convergence. We train our RL policy using deep Q-learning and show that this policy can result in significantly accelerated convergence (up to a 59% reduction in the number of iterations compared to existing, curvatureinformed penalty parameter selection methods). Furthermore, we show that our RL policy demonstrates promise for generalizability, performing well under unseen loading schemes as well as under unseen losses of lines and generators (up to a 50% reduction in iterations). This work thus provides a proof-of-concept for using RL for parameter selection in ADMM for power systems applications.

alternating current optimal power flow↗

Robust and Simple ADMM Penalty Parameter Selection

We present a new method for online selection of the penalty parameter for the alternating direction method of multipliers (ADMM) algorithm. ADMM is a widely used method for solving a range of optimization problems, including those that arise in signal and image processing. In its standard form, ADMM includes a scalar hyperparameter, known as the penalty parameter, which usually has to be tuned to achieve satisfactory empirical convergence. In this work, we develop a framework for analyzing the ADMM algorithm applied to a quadratic problem as an affine fixed point iteration. Using this framework, we develop a new method for automatically tuning the penalty parameter by detecting when it has become too large or small. We analyze this and several other methods with respect to their theoretical properties, i.e., robustness to problem transformations, and empirical performance on several optimization problems. Our proposed algorithm is based on a theoretical framework with clear, explicit assumptions and approximations, is theoretically covariant/invariant to problem transformations, is simple to implement, and exhibits competitive empirical performance.

42 ENGINEERING↗

An Analytic Framework for Optimal Milling Parameter Selection

The selection of optimal parameters is an important step in machining process definition. This paper presents a decision analytic framework for completing this task. The approach is demonstrated for titanium milling. An influence diagram showing the decision situation, the corresponding uncertainties, and the value is presented. The optimization objective is the selection of machining parameters that minimize cost while considering uncertainty in tool life and stability. The uncertainties are characterized using a probability distribution taking into account all available information. The cost associated with tool failure and unstable cutting conditions is incorporated in the cost formulation. A process probability tree showing the uncertainties and the corresponding costs is constructed. The optimization results show a 90% reduction in machining cost as compared to tool manufacturer-recommended parameters. The proposed framework is normative and robust and can be applied for optimizing process parameters in conventional (such as milling and turning) and non-conventional (such as electric discharge machining, electro-chemical machining) processes. A discussion section regarding inference, experimental design, and risk aversion is included.

Karandikar, Jaydeep↗

High Strain-Rate Strength Experiments on Molybdenum for Constitutive Model Parameter Selection

The constitutive model parameters for molybdenum obtained from the Plansee supplier have been constrained via Strength-Stabilized Rayleigh-Taylor experiments performed at the Los Alamos Neutron Science Center. Previous Hopkinson-bar measurements have produced a degenerate family of constitutive model parameters proven in the low strain rate regime. The experiments presented here provide a constraint at strain rates from 10 4 –10 6 s -1 which is used to select a preferred set of parameters for high strain-rate applications. Further analysis is warranted including determining a global fit in parameter space using all available high and low strain-rate data available.

36 MATERIALS SCIENCE↗

Effect of pulse sequence parameter selection on signal strength in positive-contrast MRI markers for MRI-based prostate postimplant assessment

Purpose: For postimplant dosimetric assessment, computed tomography (CT) is commonly used to identify prostate brachytherapy seeds, at the expense of accurate anatomical contouring. Magnetic resonance imaging (MRI) is superior to CT for anatomical delineation, but identification of the negative-contrast seeds is challenging. Positive-contrast MRI markers were proposed to replace spacers to assist seed localization on MRI images. Visualization of these markers under varying scan parameters was investigated. Methods: To simulate a clinical scenario, a prostate phantom was implanted with 66 markers and 86 seeds, and imaged on a 3.0T MRI scanner using a 3D fast radiofrequency-spoiled gradient recalled echo acquisition with various combinations of scan parameters. Scan parameters, including flip angle, number of excitations, bandwidth, field-of-view, slice thickness, and encoding steps were systematically varied to study their effects on signal, noise, scan time, image resolution, and artifacts. Results: The effects of pulse sequence parameter selection on the marker signal strength and image noise were characterized. The authors also examined the tradeoff between signal-to-noise ratio, scan time, and image artifacts, such as the wraparound artifact, susceptibility artifact, chemical shift artifact, and partial volume averaging artifact. Given reasonable scan time and managable artifacts, the authors recommended scan parameter combinations that can provide robust visualization of the MRI markers. Conclusions: The recommended MRI pulse sequence protocol allows for consistent visualization of the markers to assist seed localization, potentially enabling MRI-only prostate postimplant dosimetry.

60 APPLIED LIFE SCIENCES↗

Parameter Selection and Longitudinal Phase Space Simulation for a Single Stage X-Band FEL Driver at 250 MeV

Hard x-ray Free electron lasers (FEL) are being built or proposed at many accelerator laboratories as it supports wide range of applications in many aspects. Most of the hard x-ray FEL design is similar with the SLAC Linac Coherent Light Source (LCLS), which features a two (or multiple) stage bunch compression. For the first stage of the bunch compression, usually the beam is accelerated in a lower-frequency RF section (such as S-band for LCLS), and then the longitudinal phase space is linearized by a higher-frequency RF section (harmonic RF, such as X-band for LCLS). In this paper, a compact hard x-ray FEL design is proposed, which is based on X-band RF acceleration and eliminating the need of a harmonic RF. The parameter selection and relation is discussed, and the longitudinal phase space simulation is presented. The FEL coherence condition of the electron beam in the undulators requires a large charge density, a small emittance and small energy spread. The RMS electron bunch length from the injector is in the ps scale, with a bunch charge in the range of hundreds pC to several nC, which means that the current is roughly 0.1 kA. According to the requirement from soft x-ray lasing and hard x-ray lasing, a peak current of 1 kA and 3 kA is needed respectively. Thus the bunch has to be compressed. Usually a two stage bunch compression or multipole stage bunch compression is adopted. The z-correlated energy chirp is normally established by letting the beam pass through a section of RF cavities, with a RF phase off crest. As stated above, S-band RF (3 GHz) acceleration could be applied in this section. Due to the nature of RF acceleration wave, the chirp on the bunch is not linear, but has the RF curvature on it. In order to linearize the energy chirp, a harmonic RF section with higher frequency is needed. For LCLS a short X-band RF section (12 GHz) is used which is a fourth order harmonic. The linearized bunch is then passing by a dispersive region, in which the particles with different energy have different path length. A four dipole chicane is the natural choice for the dispersive region. As the example illustrated in Figure 1, the head of the bunch has smaller energy, and gets a stronger bending kick from the dipole magnet, then has a longer path length in the dispersive region. Similarly, the tail of the bunch has larger energy and shorter path length in the dispersive region. At the exit of the dispersive region, the relative longitudinal position of the head and tail of the bunch both move to the center of the bunch, so the bunch length will be shorter.

43 PARTICLE ACCELERATORS↗

Automatic parameter selection for electron ptychography via Bayesian optimization

Abstract Electron ptychography provides new opportunities to resolve atomic structures with deep sub-angstrom spatial resolution and to study electron-beam sensitive materials with high dose efficiency. In practice, obtaining accurate ptychography images requires simultaneously optimizing multiple parameters that are often selected based on trial-and-error, resulting in low-throughput experiments and preventing wider adoption. Here, we develop an automatic parameter selection framework to circumvent this problem using Bayesian optimization with Gaussian processes. With minimal prior knowledge, the workflow efficiently produces ptychographic reconstructions that are superior to those processed by experienced experts. The method also facilitates better experimental designs by exploring optimized experimental parameters from simulated data.

42 ENGINEERING↗

Data Visualization and Analytics for Optimal Process Parameter Selection in Turning

The objective of this project is to research physics-guided machine learning methods to recommend optimal tools and machining process parameters for turning applications using the MSC test database. For a given turning application, the MSC metalworking specialist needs to make decisions on tools and the associated process parameters for the MSC customer. For a given material, there are many alternatives for tools and a wide range of process parameters to consider. MSC has built a database of tools and parameters for different applications from the historical turning tests completed at various customer sites. The research project aims to use machine learning methods to predict optimal tools and process parameters for the MSC metalworking specialists using the MSC test database. This enables continuous learning of optimal tool and process parameters for different applications as new information is collected from testing. Through MSC, this information can be shared with machining shops across the US leading to improved productivity and efficiency.

42 ENGINEERING↗

Spatially dependent regularization parameter selection for total generalized variation-based image denoising

We propose a novel image denoising model based on the total generalized variation (TGV) regularization. In the model, a spatially dependent regularization parameter is utilized to adaptively fit the local image features, resulting in further exploitation of the denoising potential of the TGV regularization. The proposed model is formulated under a joint optimization framework, by which the estimations of the restored image and the regularization parameter are achieved simultaneously. Furthermore, the model is general purpose that can handle various types of noise occurring in image processing. An alternating minimization-based numerical scheme is especially developed, which leads to an efficient algorithmic solution to the nonconvex optimization problem. Numerical experiments are reported to illustrate the effectiveness of our model in terms of both peak signal-to-noise ratio and visual perception.

97 MATHEMATICS AND COMPUTING↗

Prediction of Silicon Content in a Blast Furnace via Machine Learning: A Comprehensive Processing and Modeling Pipeline

Silicon content plays an important role in determining the operational efficiency of blast furnaces (BFs) and their downstream processes in integrated steelmaking; however, existing sampling methods and first-principles models are somewhat limited in their capability and flexibility. Current data-based prediction models primarily rely on a limited set of manually selected furnace parameters. Additionally, different BFs present a diverse set of operating parameters and state variables that are known to directly influence the hot metal’s silicon content, such as fuel injection, blast temperature, and raw material charge composition, among other process variables that have their own impacts. The expansiveness of the parameter set adds complexity to parameter selection and processing. This highlights the need for a comprehensive methodology to integrate and select from all relevant parameters for accurate silicon content prediction. Providing accurate silicon content predictions would enable operators to adjust furnace conditions dynamically, improving safety and reducing economic risk. To address these issues, a two-stage approach is proposed. First, a generalized data processing scheme is proposed to accommodate diverse furnace parameters. Second, a robust modeling pipeline is used to establish a machine learning (ML) model capable of predicting hot metal silicon content with reasonable accuracy. The method employed herein predicted the average Si content of the upcoming furnace cast with an accuracy of 91% among 200 target predictions for a specific furnace provisioned by the XGBoost model. This prediction is achieved using only the past shift’s operating conditions, which should be available in real time. This performance provides a strong baseline for the modeling approach with potential for further improvement through provision of real-time features.

Chemistry↗

SU-D-12A-06: A Comprehensive Parameter Analysis for Low Dose Cone-Beam CT Reconstruction

Purpose: There is always a parameter in compressive sensing based iterative reconstruction (IR) methods low dose cone-beam CT (CBCT), which controls the weight of regularization relative to data fidelity. A clear understanding of the relationship between image quality and parameter values is important. The purpose of this study is to investigate this subject based on experimental data and a representative advanced IR algorithm using Tight-frame (TF) regularization. Methods: Three data sets of a Catphan phantom acquired at low, regular and high dose levels are used. For each tests, 90 projections covering a 200-degree scan range are used for reconstruction. Three different regions-of-interest (ROIs) of different contrasts are used to calculate contrast-to-noise ratios (CNR) for contrast evaluation. A single point structure is used to measure modulation transfer function (MTF) for spatial-resolution evaluation. Finally, we analyze CNRs and MTFs to study the relationship between image quality and parameter selections. Results: It was found that: 1) there is no universal optimal parameter. The optimal parameter value depends on specific task and dose level. 2) There is a clear trade-off between CNR and resolution. The parameter for the best CNR is always smaller than that for the best resolution. 3) Optimal parameters are also dose-specific. Data acquired under a high dose protocol require less regularization, yielding smaller optimal parameter values. 4) Comparing with conventional FDK images, TF-based CBCT images are better under a certain optimally selected parameters. The advantages are more obvious for low dose data. Conclusion: We have investigated the relationship between image quality and parameter values in the TF-based IR algorithm. Preliminary results indicate optimal parameters are specific to both the task types and dose levels, providing guidance for selecting parameters in advanced IR algorithms. This work is supported in part by NIH (1R01CA154747-01)

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗