Search NASA⌕ Search

SEARCH · Search NASA

Results for “demonstration box”

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 415 records · Page 23
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Target Selection and Validation of DESI Emission Line Galaxies

The Dark Energy Spectroscopic Instrument (DESI) will precisely constrain cosmic expansion and the growth of structure by collecting ~40 million extragalactic redshifts across ~80% of cosmic history and one-third of the sky. The Emission Line galaxy (ELG) sample, which will comprise about one-third of all DESI tracers, will be used to probe the universe over the 0.6 < z < 1.6 range, including the 1.1 < z < 1.6 range, which is expected to provide the tightest constraints. We present the target selection for the DESI Survey Validation (SV) and Main Survey ELG samples, which relies on the imaging of the Legacy Surveys. The Main ELG selection consists of a g-band magnitude cut and a (g – r) versus (r – z) color box, while the SV selection explores extensions of the Main selection boundaries. The Main ELG sample is composed of two disjoint subsamples, which have target densities of about 1940 deg –2 and 460 deg –2 , respectively. We first characterize their photometric properties and density variations across the footprint. We then analyze the DESI spectroscopic data that have been obtained from 2020 December to 2021 December in the SV and Main Survey. We establish a preliminary criterion for selecting reliable redshifts, based on the [O ιι] flux measurement, and assess its performance. Using this criterion, we are able to present the spectroscopic efficiency of the Main ELG selection, along with its redshift distribution. We thus demonstrate that the Main selection 1940 deg –2 subsample alone should provide 400 deg –2 and 460 deg –2 reliable redshifts in the 0.6 < z < 1.1 and the 1.1 < z < 1.6 ranges, respectively.

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRS1991

Synthesizing a protocol converter from executable protocol traces

Communicating finite state machines (CFSM's) with FIFO (first in, first out) queues are used to model a protocol converter. A protocol conversion algorithm is developed and presented for the CSFM model of the protocols A and B. A converter H for protocols A = (A0, A1) and B = (B0, B1) is viewed as a black box such that H is between sender A0 and receiver B1. This gives a resulting protocol X = (A0, H, B1). The conversion algorithm requires a specification of the message relationships between the messages of protocols A and B. It is assumed that protocols A and B have the required progress properties. The algorithm includes a search for related messages from the two protocols in an FIFO from a composite space formed by a Cartesian cross-product of state spaces A1 and B0. The search produces finite-length traces which are combined to form a state machine H, which is examined for freedom from unspecified receptions, deadlocks, and livelocks. A protocol conversion example demonstrates the applicability of the algorithm.

Rajagopal, Murali↗
NASA NTRS2010

Frequencies and Flutter Speed Estimation for Damaged Aircraft Wing Using Scaled Equivalent Plate Analysis

Equivalent plate analysis is often used to replace the computationally expensive finite element analysis in initial design stages or in conceptual design of aircraft wing structures. The equivalent plate model can also be used to design a wind tunnel model to match the stiffness characteristics of the wing box of a full-scale aircraft wing model while satisfying strength-based requirements An equivalent plate analysis technique is presented to predict the static and dynamic response of an aircraft wing with or without damage. First, a geometric scale factor and a dynamic pressure scale factor are defined to relate the stiffness, load and deformation of the equivalent plate to the aircraft wing. A procedure using an optimization technique is presented to create scaled equivalent plate models from the full scale aircraft wing using geometric and dynamic pressure scale factors. The scaled models are constructed by matching the stiffness of the scaled equivalent plate with the scaled aircraft wing stiffness. It is demonstrated that the scaled equivalent plate model can be used to predict the deformation of the aircraft wing accurately. Once the full equivalent plate geometry is obtained, any other scaled equivalent plate geometry can be obtained using the geometric scale factor. Next, an average frequency scale factor is defined as the average ratio of the frequencies of the aircraft wing to the frequencies of the full-scaled equivalent plate. The average frequency scale factor combined with the geometric scale factor is used to predict the frequency response of the aircraft wing from the scaled equivalent plate analysis. A procedure is outlined to estimate the frequency response and the flutter speed of an aircraft wing from the equivalent plate analysis using the frequency scale factor and geometric scale factor. The equivalent plate analysis is demonstrated using an aircraft wing without damage and another with damage. Both of the problems show that the scaled equivalent plate analysis can be successfully used to predict the frequencies and flutter speed of a typical aircraft wing.

Krishnamurthy, Thiagarajan↗
NASA NTRS1987

Effect of a localized minimum in equatorial field strength on resistive tearing instability in the geomagnetotail

A two-dimensional, resistive-MHD computer code is used to investigate the spontaneous reconnection of magnetotaillike configurations. The initial conditions adopted in the simulations are of two types: (1) in which the equatorial normal magnetic field component B(ze) declines monotonically down the tail, and (2) in which B(ze) exhibits a deep minimum in the near-earth plasma sheet. To represent the case where the earthward convection stops before the X line forms, zero-flow boundary conditions are imposed at the edges of the computational box. The initial configurations are in equilibrium and table within ideal MHD. The dynamic evolution of the system starts after the resistivity is turned on. The main results of these simulations basically support the neutral-line model of substorms and confirm Birn's (1980) computer studies. Spontaneous formation of an X-type neutral point and a single O-type plasmoid with strong tailward flow on the tailward side of the X point is found. The time interval from the turning on of the resistivity to the formation of a plasmoid is much shorter in the case where there is an initial deep minimum. A simple analytic calculation is also carried out to demonstrate why the configuration with a deep minimum is more susceptible to the development of the neutral point.

Hau, L.-N.↗
NASA NTRS1990

Thermal-distortion analysis of a spacecraft box truss in geostationary orbit

The Mission to Planet Earth enlists the use of a geostationary platform to support Earth science monitoring instruments. The strongback for a proposed geostationary platform is a deployable box truss that supports two large diameter passive microwave radiometer (PMR) and several other science instruments. A study was performed to estimate the north-south and east-west pointing errors at the mounting locations of the two PMRs due to on-orbit thermal distortions of the main truss. The baseline configuration indicated that the east-west pointing error greatly exceeded the required limits. Primary origins of the pointing errors were identified, and methods for their reduction were discussed. Thermal performance enhancements to the truss structure were modeled and analyzed, including state-of-the-art surface coatings and insulation techniques. Comparisons of the thermal enhancements to the baseline were performed. Results demonstrated that using a thermal enclosure insulating technique reduced external heat fluxes, and distributed those heat fluxes more evenly throughout the structure, sufficiently reducing the pointing error to satisfy pointing accuracy requirements for the PMR's.

Cosgrove, Patrick A.↗
NASA NTRS2011

Land and Atmosphere Near-Real-Time Capability for Earth Observing System

The past decade has seen a rapid increase in availability and usage of near-real-time data from satellite sensors. The EOSDIS (Earth Observing System Data and Information System) was not originally designed to provide data with sufficiently low latency to satisfy the requirements for near-real-time users. The EOS (Earth Observing System) instruments aboard the Terra, Aqua and Aura satellites make global measurements daily, which are processed into higher-level 'standard' products within 8-40 hours of observation and then made available to users, primarily earth science researchers. However, applications users, operational agencies, and even researchers desire EOS products in near-real-time to support research and applications, including numerical weather and climate prediction and forecasting, monitoring of natural hazards, ecological/invasive species, agriculture, air quality, disaster relief and homeland security. These users often need data much sooner than routine science processing allows, usually within 3 hours, and are willing to trade science product quality for timely access. While Direct Broadcast provides more timely access to data, it does not provide global coverage. In 2002, a joint initiative between NASA (National Aeronautics and Space Administration), NOAA (National Oceanic and Atmospheric Administration), and the DOD (Department of Defense) was undertaken to provide data from EOS instruments in near-real-time. The NRTPE (Near Real Time Processing Effort) provided products within 3 hours of observation on a best-effort basis. As the popularity of these near-real-time products and applications grew, multiple near-real-time systems began to spring up such as the Rapid Response System. In recognizing the dependence of customers on this data and the need for highly reliable and timely data access, NASA's Earth Science Division sponsored the Earth Science Data and Information System Project (ESDIS)-led development of a new near-real-time system called LANCE (Land, Atmosphere Near-Real-Time Capability for EOS) in 2009. LANCE consists of special processing elements, co-located with selected EOSDIS data centers and processing facilities. A primary goal of LANCE is to bring multiple near-real-time systems under one umbrella, offering commonality in data access, quality control, and latency. LANCE now processes and distributes data from the Moderate Resolution Imaging Spectroradiometer (MODIS), Atmospheric Infrared Sounder (AIRS), Advanced Microwave Scanning Radiometer Earth Observing System (AMSR-E), Microwave Limb Sounder (MLS) and Ozone Monitoring Instrument (OMI) instruments within 3 hours of satellite observation. The Rapid Response System and the Fire Information for Resource Management System (FIRMS) capabilities will be incorporated into LANCE in 2011. LANCE maintains a central website to facilitate easy access to data and user services. LANCE products are extensively tested and compared with science products before being made available to users. Each element also plans to implement redundant network, power and server infrastructure to ensure high availability of data and services. Through the user registration system, users are informed of any data outages and when new products or services will be available for access. Building on a significant investment by NASA in developing science algorithms and products, LANCE creates products that have a demonstrated utility for applications requiring near-real-time data. From lower level data products such as calibrated geolocated radiances to higher-level products such as sea ice extent, snow cover, and cloud cover, users have integrated LANCE data into forecast models and decision support systems. The table above shows the current near-real-time product categories by instrument. The ESDIS Project continues to improve the LANCE system and use the experience gained through practice to seek adjustments to improve the quality and performance of the system. For example, anGC-compliant Web Map Service (WMS) will be added shortly that will allow users to download geo-referenced MODIS images for arbitrary bounding boxes. Further, an OGC-compliant Web Coverage Service (WCS) will be added later this year that will expedite user access to arbitrary data subsets or re-formatted products. AIRS images are now served through WMS and available in multiple formats (PNG, GeoTIFF, KMZ). NASA has established a LANCE User Working Group to steer the development of the system and create a forum for sharing ideas and experiences that are expected to further improve the LANCE capabilities. The LANCE system has proved a success by satisfying the growing needs of the applications and operational communities for land and atmosphere data in near-real-time. NASA's Earth Sciences Division was able to leverage existing science research capabilities to provide the near-real-time community with products and imagery that support monitoring of disasters in a timely manner.

Murphy, Kevin J.↗
DOE OSTI2020

A Nonlocal-Gradient Descent Method for Inverse Design in Nanophotonics

Local-gradient-based optimization approaches lack nonlocal exploration abilityrequired for escaping from local minima when searching non-convex landscapes.A directional Gaussian smoothing (DGS) approach was recently proposed in [29]and used to define a truly nonlocal gradient, referred to as the DGS gradient, inorder to enable nonlocal exploration in high-dimensional black-box optimization.Promising results show that replacing the traditional local gradient with the nonlocalDGS gradient can significantly improve the performance of gradient-based methodsin optimizing highly multi-modal loss functions. However, the current DGS methodis designed for unbounded and uncontrained optimization problems, making itinapplicable to real-world engineering optimization problems where the tuningparameters are often bounded and the loss function is usually constrained byphysical processes. In this work, we propose to extend to the DGS approachto the constrained inverse design framework in order to find better optima ofmulti-modal loss functions. A series of adaptive strategies for smoothing radiusand learning rate updating are developed to improve the computational efficiencyand robustness. Our methodology is demonstrated by an example of designing ananoscale wavelength demultiplexer, and shows superior performance compared tothe state-of-the-art approaches. By incorporating volume constraints, the optimizeddesign achieves an equivalently high performance but significantly reduces theamount of material usage.

Bi, Sirui↗
DOE OSTI2024

Surrogate Model Guided Optimization of Expensive Black-Box Multi-Objective Problems: A Posteriori Methods

Many engineering applications require the simultaneous optimization of multiple conflicting objective functions. Often, these objective functions are evaluated using highly accurate computer simulations that are computationally too expensive to be evaluated hundreds or thousands of times during optimization. Thus, the goal is to find good approximations of the Pareto front using as few of these expensive simulations as possible. Here, we describe an optimization approach based on surrogate models and diverse sampling strategies to accelerate the search for the Pareto solutions. We use a separate surrogate model for approximating each objective function and then we use the surrogate models to inform where additional expensive simulations should be run. The surrogate models are updated in an active learning framework whenever new information from the expensive simulations becomes available. The sampling strategies aim at balancing local improvements of the approximate Pareto front and global exploration to identify the extrema and fill in large gaps of the approximate Pareto front. We demonstrate on a large set of benchmark problems the effectiveness of the method for finding good approximations of the Pareto front.

MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Mild hybridisation of turboprop engine with high-power-density integrated electric drives

Here this paper shares with the aerospace community a case study of turboprop mild hybridisation using a recently developed integrated drive system in the University of Nottingham, UK, within the ACHIEVE project under EU H2020 CleanSky 2 program (project No. 737814). The developed drive system enables green taxiing of a turboprop aircraft while on the ground with its engine off, and as an electrical generator when the turboprop is in the air. The entire system is designed to be able to integrate within the power auxiliary gear box (PAGB) of a turboprop aircraft. Some of the key features of the developed system include a high-speed permanent magnet machine (up to 14,200rpm) with dual three-phase design, SiC-based high power density (11.8kW/L for the power converter, 35.3kW/L and 7.2kW/kg for the machine active parts), integrated cooling design for high-temperature operation (>130°C ambient temperature), fault tolerance consideration with dual channel operation capabilities and sensorless control for entire operational conditions. This paper is giving an overview of the design process of the electrical machine, power converters, and its cooling of the entire drive. Numerical analysis (FEM and CFD) and some experimental results are presented to demonstrate the effectiveness and the desired performance of the developed integrated drive system.

42 ENGINEERING↗
DOE OSTI2022

Exploiting the Local Parabolic Landscapes of Adversarial Losses to Accelerate Black-Box Adversarial Attack

Existing black-box adversarial attacks on image classifiers update the perturbation at each iteration from only a small number of queries of the loss function. Since the queries contain very limited information about the loss, black-box methods usually require much more queries than white-box methods. We propose to improve the query efficiency of black-box methods by exploiting the smoothness of the local loss landscape. However, many adversarial losses are not locally smooth with respect to pixel perturbations. To resolve this issue, our first contribution is to theoretically and experimentally justify that the adversarial losses of many standard and robust image classifiers behave like parabolas with respect to perturbations in the Fourier domain. Our second contribution is to exploit the parabolic landscape to build a quadratic approximation of the loss around the current state, and use this approximation to interpolate the loss value as well as update the perturbation without additional queries. Since the local region is already informed by the quadratic fitting, we use large perturbation steps to explore far areas. We demonstrate the efficiency of our method on MNIST, CIFAR-10 and ImageNet datasets for various standard and robust models, as well as on Google Cloud Vision. The experimental results show that exploiting the loss landscape can help significantly reduce the number of queries and increase the success rate. Our codes are available at https://github.com/HoangATran/BABIES.

Tran, Hoang↗
DOE OSTI2021

Sequential Learning of Active Subspaces

This study shows that in recent years, active subspace methods (ASMs) have become a popular means of performing subspace sensitivity analysis on black-box functions. Naively applied, however, ASMs require gradient evaluations of the target function. In the event of noisy, expensive, or stochastic simulators, evaluating gradients via finite differencing may be infeasible. In such cases, often a surrogate model is employed, on which finite differencing is performed. When the surrogate model is a Gaussian process (GP), we show that the ASM estimator is available in closed form, rendering the finite-difference approximation unnecessary. We use our closed-form solution to develop acquisition functions focused on sequential learning tailored to sensitivity analysis on top of ASMs. We also show that the traditional ASM estimator may be viewed as a method of moments estimator for a certain class of GPs. We demonstrate how uncertainty on GP hyperparameters may be propagated to uncertainty on the sensitivity analysis, allowing model-based confidence intervals on the active subspace. Our methodological developments are illustrated on several examples.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2019

Design of an Automated Ultrasonic Scanning System for In-Situ Composite Cure Monitoring and Defect Detection

The preliminary design and development of an automated ultrasonic scanning system for in-situ composite cure monitoring and defect detection in the high temperature environment of an oven was completed. This preliminary design is a stepping stone to deployment in the high temperature and high pressure environment of an autoclave, the primary cure method of aerospace grade thermoset composites. Cure monitoring with real-time defect detection during the process could determine when defects form and how they move. In addition, real-time defect detection during cure could assist validating physics-based process models for predicting defects at all stages of the cure cycle. A physics-based process model for predicting porosity and fiber waviness originating during cure is currently under development by the NASA Advanced Composites Project (ACP). For the design, an ultrasonic contact scanner is enclosed in an insulating box that is placed inside an oven during cure. Throughout the cure cycle, the box is nitrogen-cooled to approximately room temperature to maintain a standard operating environment for the scanner. The composite part is mounted on the outside of the box in a vacuum bag on the build/tool plate. The build plate is attached to the bottom surface of the box. The scanner inspects the composite panel through the build plate, tracking the movement of defects introduced during layup and searching for new defects that may form during cure. The focus of this paper is the evaluation and selection of the build plate material and thickness. The selection was based on the required operating temperature of the scanner, the cure temperature of the composite material, thermal conductivity models of the candidate build plates, and a series of ultrasonic attenuation tests. This analysis led to the determination that a 63.5 mm thick build plate of borosilicate glass would be utilized for the system. The borosilicate glass plate was selected as the build plate material due to the low ultrasonic attenuation it demonstrated, its ability to efficiently insulate the scanner while supporting an elevated temperature on the part side of the plate, and the availability of a 63.5 mm thick plate without the need for lamination.

Hudson, Tyler B.↗
DOE OSTI2026

Systematic Construction of Time-Dependent Hamiltonians for Microwave-Driven Josephson Circuits

Time-dependent electromagnetic drives are fundamental for controlling complex quantum systems, including superconducting Josephson circuits. In these devices, accurate time-dependent Hamiltonian models are imperative for predicting their dynamics and designing high-fidelity quantum operations. Existing numerical methods, such as black-box quantization (BBQ) and energy-participation ratio (EPR), excel at modeling the static Hamiltonians of Josephson circuits. However, these techniques do not fully capture the behavior of driven circuits stimulated by external microwave drives, nor do they include a generalized approach to account for the inevitable noise and dissipation that enter through microwave ports. Here, we introduce numerical techniques that leverage classical microwave simulations, efficiently executable in finite-element solvers, to obtain the time-dependent Hamiltonian of microwave-driven superconducting circuits with arbitrary geometries under charge, flux, or mixed electromagnetic modulation. Importantly, our techniques do not rely on a lumped-element description of the superconducting circuit, in contrast to previous approaches to tackling this problem. We demonstrate the versatility of our approach by characterizing the driven properties of realistic circuit devices in complex electromagnetic environments, including coherent dynamics due to charge and flux modulation, as well as drive-induced relaxation and dephasing. Our techniques offer a powerful toolbox for optimizing circuit designs and advancing practical applications in superconducting quantum computing.

Lu, Yao [Yale U.; Yale U. (main); Fermilab] (ORCID↗
DOE OSTI2025

Solving sparse finite element problems on neuromorphic hardware

The finite element method (FEM) is one of the most important and ubiquitous numerical methods for solving partial differential equations (PDEs) on computers for scientific and engineering discovery. Applying the FEM to larger and more detailed scientific models has driven advances in high-performance computing for decades. Here we demonstrate that scalable spiking neuromorphic hardware can directly implement the FEM by constructing a spiking neural network that solves the large, sparse, linear systems of equations at the core of the FEM. We show that for the Poisson equation, a fundamental PDE in science and engineering, our neural circuit achieves meaningful levels of numerical accuracy and close to ideal scaling on modern, inherently parallel and energy-efficient neuromorphic hardware, specifically Intel’s Loihi 2 neuromorphic platform. We illustrate extensions to irregular mesh geometries in both two and three dimensions as well as other PDEs such as linear elasticity. Our spiking neural network is constructed from a recurrent network model of the brain’s motor cortex and, in contrast to black-box deep artificial neural network-based methods for PDEs, directly translates the well-understood and trusted mathematics of the FEM to a natively spiking neuromorphic algorithm.

Applied mathematics↗
DOE OSTI2025

Robustness of Deep Learning Classification to Adversarial Input on GPUs: Asynchronous Parallel Accumulation Is a Source of Vulnerability

The ability of machine learning (ML) classification models to resist small, targeted input perturbations—known as adversarial attacks—is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets used in robustness assessments. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.

Shanmugavelu, Sanjif [Maxeler Technologies, a Groq↗
NASA NTRS1990

Active control of a flexible structure using a modal positive position feedback controller

The feasibility of a new Modal Positive Position Feedback (MPPF) strategy in controlling the vibration of a complex flexible structure using a single piezo-electric active structural member is demonstrated. The control strategy generates its control forces by manipulating only the modal position signals of the structure to provide a damping action to undamped modes. This is in contrast to conventional modal controllers that rely in their operation on negative feedback of both the modal position and velocity. The proposed strategy is very simple to design and implement as it designs the controller at the uncoupled modal level and utilizes simple first order filters to achieve the Positive Position Feedback effect. The performance of the new strategy is enhanced by augmenting it with a time sharing strategy to share a small number of actuators between larger number of modes. The effectiveness of the new strategy is validated experimentally on a flexible box-type structure that has four bays and its first two bending modes are 2.015 and 6.535 Hz respectively. A single piezo-electric actuator is utilized as an active structural member to control several transverse bending modes of the structure. The performance of the active control system is determined in the time and the frequency domains. The results are compared with those obtained when using the Independent Modal Space Control (IMSC) of Meirovitch. The experimental results suggest the potential of the proposed strategy as a viable means for controlling the vibration of large flexible structures in real time.

Poh, S.↗
NASA NTRS1990

Active control of a flexible structure using a modal positive position feedback controller

The feasibility of a new Modal Positive Position Feedback (MPPF) strategy in controlling the vibration of a complex flexible structure using a single piezo-electric active structural member is demonstrated. The control strategy generates its control forces by manipulating only the modal position signals of the structure to provide a damping action to undamped modes. This is in contrast to conventional modal controllers that rely in their operation on negative feedback of both the modal position and velocity. The proposed strategy is very simple to design and implement as it designs the controller at the uncoupled modal level and utilizes simple first order filters to achieve the Positive Position Feedback effect. The performance of the new strategy is enhanced by augmenting it with a time sharing strategy to share a small number of actuators between larger number of modes. The effectiveness of the new strategy is validated experimentally on a flexible box-type structure that has four bays and its first two bending modes are 2.015 and 6.535 Hz, respectively. A single piezo-electric actuator is utilized as an active structural member to control several transverse bending modes of the structure. The performance of the active control system is determined in the time and the frequency domains. The results are compared with those obtained when using the Independent Modal Space Control (IMSC) of Meirovitch. The experimental results suggest the potential of the proposed strategy as a viable means for controlling the vibration of large flexible structures in real time.

Poh, S.↗
DOE OSTI2025

Black-box optimization of CT acquisition and reconstruction parameters: a reinforcement learning approach

Protocol optimization is critical in Computed Tomography (CT) for achieving desired diagnostic image quality while minimizing radiation dose. Due to the inter-effect of influencing CT parameters, traditional optimization methods rely on the testing of exhaustive combinations of these parameters. This poses a notable limitation due to the impracticality of exhaustive parameter testing. This study introduces a novel methodology leveraging Virtual Imaging Trials (VITs) and reinforcement learning to more efficiently optimize CT protocols. Computational phantoms with liver lesions were imaged using a validated CT simulator and reconstructed with a novel CT reconstruction Toolkit. The optimization parameter space included tube voltage, tube current, reconstruction kernel, slice thickness, and pixel size. The optimization process was done using a Proximal Policy Optimization (PPO) agent which was trained to maximize the Detectability Index (d’) of the liver lesion for each reconstructed image. Results showed that our reinforcement learning approach found the absolute maximum d’ across the test cases while requiring 79.7% fewer steps compared to an exhaustive search, demonstrating both accuracy and computational efficiency, offering a efficient and robust framework for CT protocol optimization. The flexibility of the proposed technique allows for use of varying image quality metrics as the objective metric to maximize for. Our findings highlight the advantages of combining VIT and reinforcement learning for CT protocol management.

Fenwick, David [Duke University Medical Center]↗