Search NASA⌕ Search

SEARCH · Search NASA

Results for “filter optimization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

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

At least 1,423 records · Page 79

Investigation of Off-Stoichiometry in Ternary Nitrides by EDS and HRTEM

We demonstrate an inverted metamorphic multijunction (IMM) photovoltaic cell comprising lattice-mismatched 1.2 eV AlGaInAs and 1.0 eV GaInAs junctions optimized for high-temperature thermophotovoltaic (TPV) applications. This device differs from traditional IMM solar cells because the mismatched junctions are grown at a single lattice constant. This architecture enables removal of the compositionally graded buffer which otherwise filters light from the junctions below and absorbs sub-bandgap light via free-carrier absorption. Sub-bandgap absorption dramatically reduces the efficiency of TPV systems using high reflectivity cells to enable bandedge spectrum filtering. Three components required development to enable this device: 1) a lattice-mismatched 1.2 eV AlGaInAs junction, 2) a metamorphic contact layer grown after the graded buffer, and 3) a transparent tunnel junction that sits in front of the 1.0 eV GaInAs junction. Growth conditions that minimize oxygen defect incorporation maximize AlGaInAs cell quality, enabling a 0.41 V bandgap open circuit voltage offset at 22 mA/cm2 under AM1.5D. A mismatched GaInAs:Se layer is developed as a low resistance contact. Lastly we develop a GaAsSb:C/GaInP:Se tunnel junction suitable for high-power densities with more transparency than the GaAsSb:C/GaInP:Se structure used in past IMM cells. We characterize the tandem device under a high-intensity spectrum that approximates the emission from a 2150 °C blackbody radiator and deduce a projected ideal TPV efficiency of 39.9% at ~30% of the blackbody irradiance, and 36% ideal TPV efficiency under the full 118 W/cm2 irradiance. Improvements to the back-surface reflectivity and series resistance are expected to increase the ideal TPV efficiency well above 40%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Inverted Metamorphic AlGaInAs/GaInAs Tandem Thermophotovoltaic Cell Designed for Thermal Energy Grid Storage Application

We demonstrate an inverted metamorphic multijunction (IMM) photovoltaic cell comprising lattice-mismatched 1.2 eV AlGaInAs and 1.0 eV GaInAs junctions optimized for high-temperature thermophotovoltaic (TPV) applications. This device differs from traditional IMM solar cells because the mismatched junctions are grown at a single lattice constant. This architecture enables removal of the compositionally graded buffer that otherwise filters light from the junctions below and absorbs sub-bandgap light via free-carrier absorption. Sub-bandgap absorption dramatically reduces the efficiency of TPV systems using high reflectivity cells to enable band edge spectrum filtering. Three components required development to enable this device: (1) a lattice-mismatched 1.2 eV AlGaInAs junction, (2) a metamorphic contact layer grown after the graded buffer, and (3) a transparent tunnel junction that sits in front of the 1.0 eV GaInAs junction. Growth conditions that minimize oxygen defect incorporation maximize AlGaInAs cell quality, enabling a 0.41 V bandgap open circuit voltage offset at 22 mA/cm2 under AM1.5D. A mismatched GaInAs:Se layer is developed as a low resistance contact. Lastly, we develop a GaAsSb:C/GaInP:Se tunnel junction suitable for high-power densities with more transparency than the GaAsSb:C/GaInAs:Se structure used in past IMM cells. We characterize the tandem device under a high-intensity spectrum that approximates the emission from a 2150°C blackbody radiator and deduce a projected ideal TPV efficiency of 39.9% at ~30% of the blackbody irradiance and 36% ideal TPV efficiency under the full 118 W/cm2 irradiance. Improvements to the back-surface reflectivity and series resistance are expected to increase the ideal TPV efficiency well above 40%.

25 ENERGY STORAGE↗

Homotopy Solver

This software implements parallel versions of an interior-point solver, based on the publicly available ipopt solver. Here we have full control over the linear solver and our algorithm is fully parallel thus enabling scalability to large-scale optimization problems. This package also has a parallel implementation of a homotopy solver developed under the scalable methods for contact LDRD project 23-ERD-017. This solver is an mfem-based implementation of algorithm described in ``A filter trust-region Newton continuation method for nonlinear complementarity problems''. Cosmin G. Petra, Nai-Yuan Chiang, Jingyi Wang, Tucker Hartland, and Michael Puso (submitted), LLNL-JRNL-869761.

Hartland, Tucker [Lawrence Livermore National Labo↗

Visibility of wavelet quantization noise

The discrete wavelet transform (DWT) decomposes an image into bands that vary in spatial frequency and orientation. It is widely used for image compression. Measures of the visibility of DWT quantization errors are required to achieve optimal compression. Uniform quantization of a single band of coefficients results in an artifact that we call DWT uniform quantization noise; it is the sum of a lattice of random amplitude basis functions of the corresponding DWT synthesis filter. We measured visual detection thresholds for samples of DWT uniform quantization noise in Y, Cb, and Cr color channels. The spatial frequency of a wavelet is r 2-lambda, where r is display visual resolution in pixels/degree, and lambda is the wavelet level. Thresholds increase rapidly with wavelet spatial frequency. Thresholds also increase from Y to Cr to Cb, and with orientation from lowpass to horizontal/vertical to diagonal. We construct a mathematical model for DWT noise detection thresholds that is a function of level, orientation, and display visual resolution. This allows calculation of a "perceptually lossless" quantization matrix for which all errors are in theory below the visual threshold. The model may also be used as the basis for adaptive quantization schemes.

NASA Center ARC↗

Design of 94/238 GHz Dual-Frequency Feed Horn for Atmospheric Radar Applications

A compact dual frequency feed horn is designed and optimized to operate at W-band (94.92 GHz) and G-band (238 GHz). The feed horn consists of a single spline profile horn, a diplexer having W- and G-band ports and a band pass filter at the W-band port. The filter provides isolation between the W- and G-band ports. Simulations show a directivity of 21.9 dB and -12 dB beam width of 28 degrees at 94.92 GHz and 26.27 dB and 17 degrees respectively at 238 GHz. The dual frequency horn will be used as part of a W/G-band breadboard atmospheric radar.

Focardi, Paolo↗

Onboard Navigation Error Analysis for Aerocapture at Uranus

Capturing into an orbit around Uranus using aerocapture allows one to design a mission with faster interplanetary trajectories and less propellant requirements. Such an aerocapture mission would rely on the onboard Guidance, Navigation, and Control (GNC) subsystems to successfully capture into an orbit around Uranus. Uncertainty in the state information and the noise in the sensor measurements induce navigation errors in the guidance and control subsystems, which can affect the overall performance of the aerocapture mission at Uranus. Understanding the effect of these navigation errors on mission performance is essential. To this end, this work considers different sensors with varying quality to understand their impact on the overall mission performance. In addition, this paper studies the impact of the uncertainty in the initial states used to initialize the onboard navigation filter and understands their effect on mission performance. This paper also shows the onboard navigation errors obtained from the Linear Covariance (LinCov) analysis and uses them for verification and validation (V&V) of the results from Program to Optimize and Simulate Trajectories-II (POST2).

Pardha Sai Chadalavada↗

Onboard Navigation Error Analysis for Aerocapture at Uranus

Capturing into an orbit around Uranus using aerocapture allows one to design a mission with faster interplanetary trajectories and less propellant requirements. Such an aerocapture mission would rely on the onboard Guidance, Navigation, and Control (GNC) subsystems to successfully capture into an orbit around Uranus. Uncertainty in the state information and the noise in the sensor measurements induce navigation errors in the guidance and control subsystems, which can affect the overall performance of the aerocapture mission at Uranus. Understanding the effect of these navigation errors on mission performance is essential. To this end, this work considers different sensors with varying quality to understand their impact on the overall mission performance. In addition, this paper studies the impact of the uncertainty in the initial states used to initialize the onboard navigation filter and understands their effect on mission performance. This paper also shows the onboard navigation errors obtained from the Linear Covariance (LinCov) analysis and uses them for verification and validation (V\&V) of the results from Program to Optimize and Simulate Trajectories-II (POST2).

Aerocapture↗

Speeding up Particle Track Reconstruction in the CMS Detector using a Vectorized and Parallelized Kalman Filter Algorithm

Building particle tracks is the most computationally intense step of event reconstruction at the LHC. With the increased instantaneous luminosity and associated increase in pileup expected from the High-Luminosity LHC, the computational challenge of track finding and fitting requires novel solutions. The current track reconstruction algorithms used at the LHC are based on Kalman filter methods that achieve good physics performance. By adapting the Kalman filter techniques for use on many-core SIMD architectures such as the Intel Xeon and Intel Xeon Phi and (to a limited degree) NVIDIA GPUs, we are able to obtain significant speedups and comparable physics performance. New optimizations, including a dedicated post-processing step to remove duplicate tracks, have improved the algorithm's performance even further. Here we report on the current structure and performance of the code and future plans for the algorithm.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

TOF Electron Energy Analyzer for Spin and Angular Resolved Photoemission Spectroscopy

Current pulsed laser and synchrotron x-ray sources provide new opportunities for Time-Of- Flight (TOF) based photoemission spectroscopy to increase photoelectron energy resolution and efficiency compared to current standard techniques. The principals of photoelectron timing front formation, temporal aberration minimization, and optimization of electron beam transmission are presented. We have developed these concepts into a high resolution Electron Optical Scheme (EOS) of a TOF Electron Energy Analyzer (TOF-EEA) for photoemission spectroscopy. The EOS of the analyzer includes an electrostatic objective lens, three columns of transport lenses and a 90 degree energy band pass filter (BPF). The analyzer has two modes of operation: Spectrometer Mode (SM) with straight passage of electrons through the EOS undeflected by the BPF, allowing the entire spectrum to be measured, and Monochromator Mode (MM) in which the BPF defines a certain energy window inside the scope of the electron energy spectrum.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

A Generalized Framework for Auto-tuning Stencil Computations

This work introduces a generalized framework for automatically tuning stencil computations to achieve superior performance on a broad range of multicore architectures. Stencil (nearest-neighbor) based kernels constitute the core of many important scientific applications involving block-structured grids. Auto-tuning systems search over optimization strategies to find the combination of tunable parameters that maximizes computational efficiency for a given algorithmic kernel. Although the auto-tuning strategy has been successfully applied to libraries, generalized stencil kernels are not amenable to packaging as libraries. Studied kernels in this work include both memory-bound kernels as well as a computation-bound bilateral filtering kernel. We introduce a generalized stencil auto-tuning framework that takes a straightforward Fortran expression of a stencil kernel and automatically generates tuned implementations of the kernel in C or Fortran to achieve performance portability across diverse computer architectures.

97 MATHEMATICS AND COMPUTING↗

Exploring the determinants of organic matter bioavailability through substrate-explicit thermodynamic modeling

Microbial decomposition of organic matter (OM) in river corridors is a major driver of nutrient and energy cycles in natural ecosystems. Recent advances in omics technologies enabled high-throughput generation of molecular data that could be used to inform biogeochemical models. With ultrahigh-resolution OM data becoming more readily available, in particular, the substrate-explicit thermodynamic modeling (SXTM) has emerged as a promising approach due to its ability to predict OM degradation and respiration rates from chemical formulae of compounds. This model implicitly assumes that all detected organic compounds are bioavailable, and that aerobic respiration is driven solely by thermodynamics. Despite promising demonstrations in previous studies, these assumptions may not be universally valid because OM degradation is a complex process governed by multiple factors. To identify key drivers of OM respiration, we performed a comprehensive analysis of diverse river systems using Fourier-transform ion cyclotron resonance mass spectrometry OM data and associated respiration measurements collected by the Worldwide Hydrobiogeochemistry Observation Network for Dynamic River Systems (WHONDRS) consortium. In support of our argument, we found that the incorporation of all compounds detected in the samples into the SXTM resulted in a poor correlation between the predicted and measured respiration rates. The data-model consistency was significantly improved by the selective use of a small subset (i.e., only about 5%) of organic compounds identified using an optimization method. Through a subsequent comparative analysis of the subset of compounds (which we presume as bioavailable) against the full set of compounds, we identified three major traits that potentially determine OM bioavailability, including: (1) thermodynamic favorability of aerobic respiration, (2) the number of C atoms contained in compounds, and (2) carbon/nitrogen (C/N) ratio. We found that all three factors serve as “filters” in that the compounds with undesirable properties in any of these traits are strictly excluded from the bioavailable fraction. This work highlights the importance of accounting for the complex interplay among multiple key traits to increase the predictive power of biogeochemical and ecosystem models.

59 BASIC BIOLOGICAL SCIENCES↗

Visualization of Hierarchical Nanodomains in Polymer/Fullerene Bulk Heterojunction Solar Cells

Here, traditional electron microscopy techniques such as bright-field imaging provide poor contrast for organic films and identification of structures in amorphous material can be problematic, particularly in high-performance organic solar cells. By combining energy-filtered corrected transmission electron microscopy, together with electron energy loss and X-ray energy-dispersive hyperspectral imaging, we have imaged PTB7/ PC 61 BM blended polymer optical photovoltaic films, and were able to identify domains ranging in size from several hundred nanometers to several nanometers in extent. This work verifies that microstructural domains exist in bulk heterojunctions in PTB7/PC 61 BM polymeric solar cells at multiple length scales and expands our understanding of optimal device performance providing insight for the design of even higher performance cells.

14 SOLAR ENERGY↗

Produced Water and Waste Heat-aided Blowdown Water Treatment: Using Chemical and Energy Synergisms for Value Creation

The project objective was to develop a cooling blowdown water (BDW) treatment process utilizing produced water (PW) and low-grade heat to maximize water reuse and saleable by-product generation while reducing chemical and energy footprints of the treatment. The proposed treatment process consists of mixing, softening, organics and suspended solids removal, reverse osmosis (RO), thermal desalination, and brine electrolysis. BDW samples collected from a local coal-fired power plant and PW samples from two shale gas production wells were used in this study. Each treatment unit was first designed and tested to quantify its treatment efficiency, and its chemical and energy requirements. In addition, a process model was developed and model simulations were conducted based on the experimental results and literature data to optimize the treatment process. A techno-economic analysis was conducted to quantify chemical and energy savings as well as production of 10-lb brine as a saleable product. With the field-collected BDW and PW samples, mixing experiments determined a volumetric mixing ratio 10:1 (BDW:PW) resulted in the best performance of multivalent ions removal and largest chemical savings for softening. Softening of the BDW/PW mixtures using alkaline chemicals (Na 2 CO 3 and NaOH) achieved 95%-100% removal of scaling-forming cations (Ca, Mg, Fe, Ba, Sr) and 60% of silicon, and 10% of total organic carbon (TOC). The mixing and softening treatments yielded an effluent with total dissolved solids (TDS) concentration of 23 g/L. Activated carbon (AC) filtration removed TOC to a low level (< 3 mg/L) and further removed remaining scale-forming divalent metals and silica from the softened water. The AC filtration resulted in a slight reduction of TDS from 23 g/L to 20 g/L, leaving behind only mostly monovalent ions (i.e., sodium and chloride) in the filtered water. These pretreatments yielded a feed water that met the criteria of the downstream reverse osmosis (RO) to prevent membrane fouling. A cross-flow RO system was used to further concentrate the TDS of the AC effluent. Various factors including TDS, pH, and applied pressure were examined and optimal conditions were determined for the co-treatment process. An integrated process consisting of mixing, softening, AC filtration and RO was used to treat a continuous flow (0.25 – 1.2 L/min, or 0.07 – 0.32 gpm) and successfully generated RO permeate as product water (TDS < 0.5 g/L) for reuse in cooling operation, and a concentrate (TDS ~ 45 g/L) to be further treated in a thermal desalination unit. These flow rates meet the FOA’s criterion of 0.01 – 1 gpm. Overall, the co-treatment of BDW/PW allowed shorter ramp-up time compared to treatment of BDW alone. It resulted in 40% and 55% savings of Na 2 CO 3(s) and NaOH, respectively, compared to treating the BDW and PW individually for the same level of softening. The co-treatment also resulted in a 29% energy saving compared to treatment of BDW only for the level of TDS concentration. A thermal desalination system was designed using CFD simulations and manufactured in the WVU Innovation Hub for further treatment of the RO concentrate to generate 10-lb brine. The system has a design flow rate of 2 gpm and has been successfully tested. A bench-scale brine electrolysis system was developed for on-site generation of chlorine/hypochlorite (Cl 2 /OCl - ) and caustic soda (NaOH) as useful chemicals for the co-treatment process. Using salt solutions (0.5 M and 1 M), the system achieved faradaic efficiencies of 93%-97% and 70%-77% for caustic soda and chlorine/hypochlorite generation, respectively. An economic analysis showed that the electricity costs for on-site generation of these chemicals were significantly lower than the chemical prices offered by suppliers. An industrial-scale process model consisting of mixing, softening, AC filtration, RO, thermal desalination, and brine electrolysis was developed using the Aspen Plus V9 in conjunction with Aspen Custom Modeler V9. The model serves as a solvable Aspen Plus model and as basis to form the costing infrastructure. In addition, techno-economic analysis considering capital, operating, and transportation costs was conducted. An optimization solution showed that produced water for mixing is still advantageous in low quantities. The optimum solution approaches a leveled cost of water (LCW) of 2 $/m 3 which becomes cost competitive with nominal water treatment prices.

20 FOSSIL-FUELED POWER PLANTS↗

Innovative high-temperature ammonium bifluoride fusion and rapid analysis of elements with nuclear forensic value

High-temperature ammonium bifluoride (ABF) fusions were evaluated for potential use in rapid dissolution of post-detonation nuclear debris. The ABF fusion was carried out in a Pt crucible which allowed evaluation of higher fusion and evaporation temperatures. The high-temperature ABF fusion dissolution method was evaluated using geological reference materials: USGS QLO-1a Quartz Latite, USGS SDC-1 Mica Schist, and NIST 278 Obsidian Rock. The optimized dissolution method involved a 10 min fusion at 540 °C, a 5 min reflux in 8 M HNO 3 , an evaporation at 300 °C and final dilution into 45 mL of 2% (v/v) HNO3. The final solution was filtered after heating at 105 °C using a hotblock. This dissolution method was simple, requiring only a hotplate or hotblock, filtered samples were available for ICP-MS analysis or radiochemical separation within 150 min, and was found to have high (>90%) recovery for many isotopes of interest in nuclear forensics applications. U and Pu in the dissolved material was separated using TEVA and UTEVA extraction chromatography columns, a process which resulted in >90% recovery. An irradiated U tracer was spiked into the material prior to dissolution and analyzed for recovery of major fission products and 239Np. The monitored radionuclides had recoveries of greater than 90%, except for the volatile radioiodine isotopes.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

MO-A-BRD-02: Noise Suppression for Dual-Energy CT Through Entropy Minimization

Purpose: In dual energy CT (DECT), noise amplification during signal decomposition significantly limits the utility of basis material images. Since clinically relevant objects contain a limited number of materials, we propose to suppress noise in decomposed images through entropy minimization within a 2D transformation space. Distinct from other noise suppression techniques, the entropy minimization method does not estimate and suppress noise based on spatial variations of signals and thus maximally preserves image spatial resolution. Methods: From decomposed images, we first generate a 2D plot of scattered data points, using basis material densities as coordinates. Data points representing the same material generate a cluster with a highly asymmetric shape. We orient an axis by minimizing the entropy in a 1D histogram of these points projected onto the axis. To suppress noise, we replace the pixel values of decomposed images with center-of-mass values in the direction perpendicular to the optimized axis. The proposed method's performance is assessed using a Catphan 600 phantom and an anthropomorphic head phantom. Electron density calculations are used to quantify its accuracy. Our results are compared to those without noise suppression, with a filtering method, and with a recently developed iterative method. Results: On both phantoms, the proposed method reduces noise standard deviations of the decomposed images by at least on order of magnitude. In the Catphan study, this method retains the spatial resolution of the CT images and increases the accuracy of electron density calculations. In the head phantom study, the proposed method outperforms the others in retaining fine, intricate structures. Conclusion: This work shows that the proposed method of noise suppression through entropy minimization for DECT suppresses noise without loss of spatial resolution while increasing electron density calculation accuracy. Future investigations will analyze possible bias and incorporate spatial information into the technique for further noise reduction. The work is partially supported by the NIH under the grant number R21EB012700.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

SU-E-T-326: Dosimetric Impact of Beam Energies and Jaw Tracking On Intracranial Tumors Using RapidArc

Purpose: To determine the dosimetric impact of jaw tracking and beam energies on dose conformity and normal-brain-tissue doses for intracranial tumors using VMAT (RapidArc). Methods: Seven patients with 1–2 and three patients with 4–6 intracranial tumors were planned using RapidArc for Varian TrueBeam STx machine with beam energies 6MV-FFF (Flattening-Filter-Free), 8MV, 10MV, and 10MV-FFF. The prescription dose ranged from 14–23Gy. Between 2 and 8 arcs were used with the following geometries: 2 full coplanar arcs and the non-coplanar half arcs. Plans were optimized (jaw tracking ON) with a high priority to Normal-Tissue-Objective and normal-brain-tissue. Plans were calculated on 1mm grid size using AAA algorithm and then normalized so that 99% of each target volume received the prescription dose. Plans for the 6MV-FFF were also optimized without jaw tracking (No-JT) for comparison. Plan quality was assessed by target coverage using Paddick Conformity Index (PCI), sparing of normal-brain-tissue through analysis of V4Gy, V8Gy and V12Gy, and integral dose. Results: The average PCI ± standard deviation was 0.76±0.21 and 0.76±0.22 for 6MV-FFF and 10 MV-FFF, respectively. The average ratio in normal brain tissue volume (reported as follows V4,V8,V12) were (1.12±0.07,1.12±0.07,1.14±0.04), (1.12±0.08,1.12±0.09,1.13±0.06), (1.19±0.10,1.18±0.10,1.20±0.04), and (1.04±0.03,1.03±0.03,1.03±0.04) for 8MV/6MV-FFF, 10MV-FFF/6MV-FFF, 10MV/6MV-FFF, 6MV-FFF No-JT/6MV-FFF, respectively. Statistically significant differences in normal-brain-tissue for V4, V8, and V12 were observed in all cases for the different energies (p-values <0.05). V4 data shows significant differences in JT vs. No-JT (p=0.04), however no difference was found for V8 and V12. Brain tissue sparing from best to worst occurred in this order 6MV-FFF, 6MV-FFF no-JT, 10MV-FFF, 8MV, and 10MV. The average ratio of integral brain dose was 1.05±0.04 (p=0.21), 1.04±0.05 (p=0.33), 1.09±0.06 (p=0.04), and 1.02±0.06 (p=0.61) for 8MV/6MV-FFF, 10MV-FFF/6MV-FFF, 10MV/6MV-FFF, and 6MV-FFF No-JT/6MV-FFF, respectively. Conclusion: Normal brain tissue and integral dose improved using the lower energy and FFF beams, though plan conformity showed energy independence.

60 APPLIED LIFE SCIENCES↗

TH-AB-207A-05: A Fully-Automated Pipeline for Generating CT Images Across a Range of Doses and Reconstruction Methods

Purpose: To evaluate the robustness of CAD or Quantitative Imaging methods, they should be tested on a variety of cases and under a variety of image acquisition and reconstruction conditions that represent the heterogeneity encountered in clinical practice. The purpose of this work was to develop a fully-automated pipeline for generating CT images that represent a wide range of dose and reconstruction conditions. Methods: The pipeline consists of three main modules: reduced-dose simulation, image reconstruction, and quantitative analysis. The first two modules of the pipeline can be operated in a completely automated fashion, using configuration files and running the modules in a batch queue. The input to the pipeline is raw projection CT data; this data is used to simulate different levels of dose reduction using a previously-published algorithm. Filtered-backprojection reconstructions are then performed using FreeCT-wFBP, a freely-available reconstruction software for helical CT. We also added support for an in-house, model-based iterative reconstruction algorithm using iterative coordinate-descent optimization, which may be run in tandem with the more conventional recon methods. The reduced-dose simulations and image reconstructions are controlled automatically by a single script, and they can be run in parallel on our research cluster. The pipeline was tested on phantom and lung screening datasets from a clinical scanner (Definition AS, Siemens Healthcare). Results: The images generated from our test datasets appeared to represent a realistic range of acquisition and reconstruction conditions that we would expect to find clinically. The time to generate images was approximately 30 minutes per dose/reconstruction combination on a hybrid CPU/GPU architecture. Conclusion: The automated research pipeline promises to be a useful tool for either training or evaluating performance of quantitative imaging software such as classifiers and CAD algorithms across the range of acquisition and reconstruction parameters present in the clinical environment. Funding support: NIH U01 CA181156; Disclosures (McNitt-Gray): Institutional research agreement, Siemens Healthcare; Past recipient, research grant support, Siemens Healthcare; Consultant, Toshiba America Medical Systems; Consultant, Samsung Electronics.

60 APPLIED LIFE SCIENCES↗

Multiobjective and categorical global optimization of photonic structures based on ResNet generative neural networks

Abstract We show that deep generative neural networks, based on global optimization networks (GLOnets), can be configured to perform the multiobjective and categorical global optimization of photonic devices. A residual network scheme enables GLOnets to evolve from a deep architecture, which is required to properly search the full design space early in the optimization process, to a shallow network that generates a narrow distribution of globally optimal devices. As a proof-of-concept demonstration, we adapt our method to design thin-film stacks consisting of multiple material types. Benchmarks with known globally optimized antireflection structures indicate that GLOnets can find the global optimum with orders of magnitude faster speeds compared to conventional algorithms. We also demonstrate the utility of our method in complex design tasks with its application to incandescent light filters. These results indicate that advanced concepts in deep learning can push the capabilities of inverse design algorithms for photonics.

Jiang, Jiaqi↗