SEARCH · Engineering Papers
Results for “.NET”
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.
Charged-to-neutral correlation at forward rapidity in Au+Au collisions at s NN = 200 GeV
Event-by-event fluctuations of the ratio of inclusive charged to photon multiplicities at forward rapidity in Au+Au collision at √s NN =200 GeV have been studied. Dominant contribution to such fluctuations is expected to come from correlated production of charged and neutral pions. We search for evidences of dynamical fluctuations of different physical origins. Observables constructed out of moments of multiplicities are used as measures of fluctuations. Mixed events and model calculations are used as baselines. Results are compared to the dynamical net-charge fluctuations measured in the same acceptance. A non-zero statistically significant signal of dynamical fluctuations is observed in excess to the model prediction when charged particles and photons are measured in the same acceptance. Thus, we find that, unlike dynamical net-charge fluctuation, charge-neutral fluctuation is not dominated by correlation due to particle decay. Results are compared to the expectations based on the generic production mechanism of pions due to isospin symmetry, for which no significant (< 1%) deviation is observed.
Kinetic relaxation and nucleation of Bose stars in self-interacting wave dark matter
We revisit kinetic relaxation and soliton/boson star nucleation in fuzzy scalar dark matter featuring short-ranged self-interactions H int = − λ | ψ | 4 / 2 m 2 , alongside gravitational self-interactions. We map out the full curve of nucleation timescale for both repulsive ( λ < 0 ) and attractive ( λ > 0 ) short-ranged self-interaction strength and in doing so reveal two new points. Firstly, besides the two usual terms, ∝ G 2 and ∝ λ 2 , in the total relaxation rate Γ relax , there is an additional cross term ∝ G λ arising due to interference between gravitational and short-ranged self-interaction scattering amplitudes. This yields a critical repulsive interaction strength λ cr ≃ − 2 π G m 2 / v 0 2 , at which the relaxation rate is smallest and serves as the transition point between typical net attractive self-interaction ( λ ≳ λ cr ) and net repulsive self-interaction ( − λ ≳ − λ cr ). Secondly, while in the net attractive regime, nucleation timescale is similar to inverse relaxation timescale τ nuc ∼ Γ relax − 1 ; in the net repulsive regime, nucleation occurs at a delayed time τ nuc ∼ ( λ / λ cr ) Γ relax − 1 . We confirm our analytical understanding by performing 3D field simulations with varying average mass density ρ ¯ , box size L and grid size N . Published by the American Physical Society 2024
Controllable topological insulator phases in litharge-phase InBi monolayer
Despite recent advances of layered square-net topological material models that possess ideal semimetallic electronic structures and promising potential in material applications, the identification of experimentally accessible two-dimensional square-net materials with related topological properties has proven challenging. Due to the highly tunable physical and topological properties of III-V semiconductors, we revisit the class of III-V materials and observe that the litharge-phase InBi is a layered square-net material and can be exfoliated into the InBi monolayer. We present a comprehensive first-principles study of the energy landscape of the InBi monolayer. We identify a paraelastic phase and three ferroelastic phases and study their topological properties. Specifically, we show that the paraelastic InBi monolayer is a trivial insulator due to the orbital-ordering-induced band inversion occurring between states with the same parity. Substituting one Bi atom per cell with another V-group element (N, P, As) or applying an electric field that breaks the inversion symmetry and changes the orbital onsite energy, the paraelastic InBi monolayer can be driven into the topological insulator phase. Furthermore, one of the ferroelastic phases of pure InBi, which can be obtained by gently straining the paraelastic phase, also possesses such topological insulating properties. Furthermore, these results provide several experimentally accessible routes to tune the nontrivial topology in the InBi monolayer, including creating heterostructures with piezoelectric or ferroelectric substrates and applying mechanical strain, making the InBi monolayer an ideal platform to study the interplay of reduced dimensionality, square-net chemical bonding networks, and band topology.
Scalable Hybrid Classification-Regression Solution for High-Frequency Nonintrusive Load Monitoring
Residential buildings with the ability to monitor and control their net-load (sum of load and generation) can provide valuable flexibility to power grid operators. We present a novel multiclass nonintrusive load monitoring (NILM) approach that enables effective net-load monitoring capabilities at high-frequency with minimal additional equipment and cost. The proposed machine learning based solution provides accurate multiclass state predictions while operating at a faster timescale (able to provide a prediction for each 60- Hz ac cycle used in US power grid) without relying on event-detection techniques. We also introduce an innovative hybrid classification-regression method that allows for the prediction of not only load on/off states but also individual load operating power levels. A test bed with eight residential appliances is used for validating the NILM approach. Results show that the overall method has high accuracy, good scaling and generalization properties.
A Methodology to Analyze the Impact of 30-Minute Wind Scheduling on Load-Following Requirements
This paper proposes a new and systematic approach to investigating the impact of wind transfer between balancing authorities (BAs) with half-hour scheduling on load following requirements, which was traditionally scheduled on an hourly basis. The analysis is conducted for a few BAs in the Pacific Northwest of the U.S. that are described as source BAs (sending renewables) and sink BAs (receiving renewables). The main hypothesis is that if the source BA could schedule interchange for wind on a half-hourly basis, it would make the schedule follow its net load more closely. The scheduling change in the source BA is matched by adding the corresponding component to the net load in the sink BA. The load-following requirements are calculated as: (a) by the difference between the net load and modified schedule in the source BA, and (b) the difference between modified net load and unchanged hourly schedule in the sink BA. Results are presented as hourly upward and downward load following requirements in the source and sink BA, and compared with the results obtained with all generators scheduled on an hourly basis. Thus, the proposed method can effectively help utilities better understand the impact of 30-minute scheduling and make better business decisions.
Network Time Synchronization of the Readout Electronics for a New Radioactive Gas Detection System
In systems with multiple radiation detectors, time synchronization of the data collected from different detectors is essential to reconstruct multi-detector events such as scattering and coincidences. In cases where the number of detectors exceeds the readout channels in a single data acquisition electronics module, multiple modules have to be synchronized, which is traditionally accomplished by distributing clocks and triggers via dedicated connections. To eliminate this added cabling complexity in the case of a new radioactive gas detection system prototype under development at the French Atomic Energy Commission, we implemented time synchronization between multiple XIA Pixie-Net detector readout modules through the existing Ethernet network, based on the IEEE 1588 precision time protocol. The detector system is dedicated to the measurement of radioactive gases at low activity and consists of eight large silicon pixels and two NaI(Tl) detectors, instrumented with a total of three 4-channel Pixie-Net modules. Detecting NaI(Tl)/silicon coincidences will make it possible to identify each radioisotope present in the sample. To allow these identifications at low activities, the Pixie-Net modules must be synchronized to a precision well below the targeted coincidence window of 500-1000 ns. Being equipped with an Ethernet PHY compatible with IEEE 1588 and synchronous Ethernet that outputs a locally generated but system-wide synchronized clock, the Pixie-Net can operate its analog to digital converters and digital processing circuitry with that clock and match time stamps for captured data across the three modules. Depending on the network configuration and synchronization method, the implementation is capable to achieve timing precisions between 300 ns and 200 ps.
Terrestrial carbon balance in a drier world: the effects of water availability in southwestern North America
Abstract Global modeling efforts indicate semiarid regions dominate the increasing trend and interannual variation of net CO 2 exchange with the atmosphere, mainly driven by water availability. Many semiarid regions are expected to undergo climatic drying, but the impacts on net CO 2 exchange are poorly understood due to limited semiarid flux observations. Here we evaluated 121 site‐years of annual eddy covariance measurements of net and gross CO 2 exchange (photosynthesis and respiration), precipitation, and evapotranspiration ( ET ) in 21 semiarid North American ecosystems with an observed range of 100 – 1000 mm in annual precipitation and records of 4–9 years each. In addition to evaluating spatial relationships among CO 2 and water fluxes across sites, we separately quantified site‐level temporal relationships, representing sensitivity to interannual variation. Across the climatic and ecological gradient, photosynthesis showed a saturating spatial relationship to precipitation, whereas the photosynthesis– ET relationship was linear, suggesting ET was a better proxy for water available to drive CO 2 exchanges after hydrologic losses. Both photosynthesis and respiration showed similar site‐level sensitivity to interannual changes in ET among the 21 ecosystems. Furthermore, these temporal relationships were not different from the spatial relationships of long‐term mean CO 2 exchanges with climatic ET . Consequently, a hypothetical 100‐mm change in ET , whether short term or long term, was predicted to alter net ecosystem production ( NEP ) by 64 gC m −2 yr −1 . Most of the unexplained NEP variability was related to persistent, site‐specific function, suggesting prioritization of research on slow‐changing controls. Common temporal and spatial sensitivity to water availability increases our confidence that site‐level responses to interannual weather can be extrapolated for prediction of CO 2 exchanges over decadal and longer timescales relevant to societal response to climate change.
The greenhouse gas cost of agricultural intensification with groundwater irrigation in a Midwest U.S. row cropping system
Abstract Groundwater irrigation of cropland is expanding worldwide with poorly known implications for climate change. This study compares experimental measurements of the net global warming impact of a rainfed versus a groundwater‐irrigated corn (maize)–soybean–wheat, no‐till cropping system in the Midwest US, the region that produces the majority of U.S. corn and soybean. Irrigation significantly increased soil organic carbon (C) storage in the upper 25 cm, but not by enough to make up for the CO 2 ‐equivalent (CO 2 e) costs of fossil fuel power, soil emissions of nitrous oxide (N 2 O), and degassing of supersaturated CO 2 and N 2 O from the groundwater. A rainfed reference system had a net mitigating effect of −13.9 (±31) g CO 2 e m −2 year −1 , but with irrigation at an average rate for the region, the irrigated system contributed to global warming with net greenhouse gas (GHG) emissions of 27.1 (±32) g CO 2 e m −2 year −1 . Compared to the rainfed system, the irrigated system had 45% more GHG emissions and 7% more C sequestration. The irrigation‐associated increase in soil N 2 O and fossil fuel emissions contributed 18% and 9%, respectively, to the system's total emissions in an average irrigation year. Groundwater degassing of CO 2 and N 2 O are missing components of previous assessments of the GHG cost of groundwater irrigation; together they were 4% of the irrigated system's total emissions. The irrigated system's net impact normalized by crop yield (GHG intensity) was +0.04 (±0.006) kg CO 2 e kg −1 yield, close to that of the rainfed system, which was −0.03 (±0.002) kg CO 2 e kg −1 yield. Thus, the increased crop yield resulting from irrigation can ameliorate overall GHG emissions if intensification by irrigation prevents land conversion emissions elsewhere, although the expansion of irrigation risks depletion of local water resources.
A substantial role of soil erosion in the land carbon sink and its future changes
Realistic representation of land carbon sink in climate models is vital for predicting carbon climate feedbacks in a changing world. Although soil erosion that removes land organic carbon has increased substantially since the onset of agriculture, it is rarely included in the current generation of climate models. Using an Earth system model with soil erosion represented, we estimated that on average soil erosion displaces 5% of newly fixed land organic carbon downslope annually in the continental US. In the lower Mississippi river basin and the Cascades, the fraction can be as large as 40%. About 12% of the eroded organic carbon is eventually exported to inland waters, which is equal to 14% of the simulated net carbon gain by terrestrial ecosystems. By comparing the eroded organic carbon export to rivers with the particulate organic carbon export to oceans, we demonstrated that a large fraction of the carbon export to rivers could have been mineralized in inland waters. Importantly, with a direct comparison of eroded and exported soil organic carbon and land net carbon uptake, we found that Earth system models that ignore soil erosion likely offset the erosional carbon loss by increasing heterotrophic respiration implicitly. But as soil erosion and heterotrophic respiration respond differently to a warming climate, this unrealistic compensation would lead to biased predictions of future land carbon sink.
Coupling plant litter quantity to a novel metric for litter quality explains C storage changes in a thawing permafrost peatland
Abstract Permafrost thaw is a major potential feedback source to climate change as it can drive the increased release of greenhouse gases carbon dioxide (CO 2 ) and methane (CH 4 ). This carbon release from the decomposition of thawing soil organic material can be mitigated by increased net primary productivity (NPP) caused by warming, increasing atmospheric CO 2 , and plant community transition. However, the net effect on C storage also depends on how these plant community changes alter plant litter quantity, quality, and decomposition rates. Predicting decomposition rates based on litter quality remains challenging, but a promising new way forward is to incorporate measures of the energetic favorability to soil microbes of plant biomass decomposition. We asked how the variation in one such measure, the nominal oxidation state of carbon (NOSC), interacts with changing quantities of plant material inputs to influence the net C balance of a thawing permafrost peatland. We found: (1) Plant productivity (NPP) increased post‐thaw, but instead of contributing to increased standing biomass, it increased plant biomass turnover via increased litter inputs to soil; (2) Plant litter thermodynamic favorability (NOSC) and decomposition rate both increased post‐thaw, despite limited changes in bulk C:N ratios; (3) these increases caused the higher NPP to cycle more rapidly through both plants and soil, contributing to higher CO 2 and CH 4 fluxes from decomposition. Thus, the increased C‐storage expected from higher productivity was limited and the high global warming potential of CH 4 contributed a net positive warming effect. Although post‐thaw peatlands are currently C sinks due to high NPP offsetting high CO 2 release, this status is very sensitive to the plant community's litter input rate and quality. Integration of novel bioavailability metrics based on litter chemistry, including NOSC, into studies of ecosystem dynamics, is needed to improve the understanding of controls on arctic C stocks under continued ecosystem transition.
Optimal stomatal theory predicts CO 2 responses of stomatal conductance in both gymnosperm and angiosperm trees
Optimal stomatal theory predicts that stomata operate to maximise photosynthesis (A net ) and minimise transpirational water loss to achieve optimal intrinsic water-use efficiency (iWUE). We tested whether this theory can predict stomatal responses to elevated atmospheric CO 2 (eCO 2 ), and whether it can capture differences in responsiveness among woody plant functional types (PFTs). We conducted a meta-analysis of tree studies of the effect of eCO 2 on iWUE and its components A net and stomatal conductance (g s ). We compared three PFTs, using the unified stomatal optimisation (USO) model to account for confounding effects of leaf–air vapour pressure difference (D). We expected smaller g s , but greater A net , responses to eCO 2 in gymnosperms compared with angiosperm PFTs. We found that iWUE increased in proportion to increasing eCO 2 in all PFTs, and that increases in A net had stronger effects than reductions in g s . The USO model correctly captured stomatal behaviour with eCO 2 across most datasets. The chief difference among PFTs was a lower stomatal slope parameter (g 1 ) for the gymnosperm, compared with angiosperm, species. Land surface models can use the USO model to describe stomatal behaviour under changing atmospheric CO 2 conditions.
Characterizing the marker-dye correction for Gafchromic EBT2 film: A comparison of three analysis methods
Purpose: Gafchromic EBT2 film has a yellow marker dye incorporated into the active layer of the film that can be used to correct the film response for small variations in thickness. This work characterizes the effect of the marker-dye correction on the uniformity and uncertainty of dose measurements with EBT2 film. The effect of variations in time postexposure on the uniformity of EBT2 is also investigated. Methods: EBT2 films were used to measure the flatness of a {sup 60}Co field to provide a high-spatial resolution evaluation of the film uniformity. As a reference, the flatness of the {sup 60}Co field was also measured with Kodak EDR2 films. The EBT2 films were digitized with a flatbed document scanner 24, 48, and 72 h postexposure, and the images were analyzed using three methods: (1) the manufacturer-recommended marker-dye correction, (2) an in-house marker-dye correction, and (3) a net optical density (OD) measurement in the red color channel. The field flatness was calculated from orthogonal profiles through the center of the field using each analysis method, and the results were compared with the EDR2 measurements. Uncertainty was propagated through a dose calculation for each analysis method. The change in the measured field flatness for increasing times postexposure was also determined. Results: Both marker-dye correction methods improved the field flatness measured with EBT2 film relative to the net OD method, with a maximum improvement of 1% using the manufacturer-recommended correction. However, the manufacturer-recommended correction also resulted in a dose uncertainty an order of magnitude greater than the other two methods. The in-house marker-dye correction lowered the dose uncertainty relative to the net OD method. The measured field flatness did not exhibit any unidirectional change with increasing time postexposure and showed a maximum change of 0.3%. Conclusions: The marker dye in EBT2 can be used to improve the response uniformity of the film. Depending on the film analysis method used, however, application of a marker-dye correction can improve or degrade the dose uncertainty relative to the net OD method. The uniformity of EBT2 was found to be independent of the time postexposure.
Permanent-magnet energy spectrometer for electron beams from radiotherapy accelerators
Purpose: The purpose of this work was to adapt a lightweight, permanent magnet electron energy spectrometer for the measurement of energy spectra of therapeutic electron beams. Methods: An irradiation geometry and measurement technique were developed for an approximately 0.54-T, permanent dipole magnet spectrometer to produce suitable latent images on computed radiography (CR) phosphor strips. Dual-pinhole electron collimators created a 0.318-cm diameter, approximately parallel beam incident on the spectrometer and an appropriate dose rate at the image plane (CR strip location). X-ray background in the latent image, reduced by a 7.62-cm thick lead block between the pinhole collimators, was removed using a fitting technique. Theoretical energy-dependent detector response functions (DRFs) were used in an iterative technique to transform CR strip net mean dose profiles into energy spectra on central axis at the entrance to the spectrometer. These spectra were transformed to spectra at 95-cm source to collimator distance (SCD) by correcting for the energy dependence of electron scatter. The spectrometer was calibrated by comparing peak mean positions in the net mean dose profiles, initially to peak mean energies determined from the practical range of central-axis percent depth-dose (%DD) curves, and then to peak mean energies that accounted for how the collimation modified the energy spectra (recalibration). The utility of the spectrometer was demonstrated by measuring the energy spectra for the seven electron beams (7–20 MeV) of an Elekta Infinity radiotherapy accelerator. Results: Plots of DRF illustrated their dependence on energy and position in the imaging plane. Approximately 15 iterations solved for the energy spectra at the spectrometer entrance from the measured net mean dose profiles. Transforming those spectra into ones at 95-cm SCD increased the low energy tail of the spectra, while correspondingly decreasing the peaks and shifting them to slightly lower energies. Energy calibration plots of peak mean energy versus peak mean position of the net mean dose profiles for each of the seven electron beams followed the shape predicted by the Lorentz force law for a uniform z-component of the magnetic field, validating its being modeled as uniform (0.542 ± 0.027 T). Measured Elekta energy spectra and their peak mean energies correlated with the 0.5-cm (7–13 MeV) and the 1.0-cm (13–20 MeV) R{sub 90} spacings of the %DD curves. The full-width-half-maximum of the energy spectra decreased with decreasing peak mean energy with the exception of the 9-MeV beam, which was anomalously wide. Similarly, R{sub 80–20} decreased linearly with peak mean energy with the exception of the 9 MeV beam. Both were attributed to suboptimal tuning of the high power phase shifter for the recycled radiofrequency power reentering the traveling wave accelerator. Conclusions: The apparatus and analysis techniques of the authors demonstrated that an inexpensive, lightweight, permanent magnet electron energy spectrometer can be used for measuring the electron energy distributions of therapeutic electron beams (6–20 MeV). The primary goal of future work is to develop a real-time spectrometer by incorporating a real-time imager, which has potential applications such as beam matching, ongoing beam tune maintenance, and measuring spectra for input into Monte Carlo beam calculations.
Statistical-learning strategies generate only modestly performing predictive models for urinary symptoms following external beam radiotherapy of the prostate: A comparison of conventional and machine-learning methods
Purpose: Given the paucity of available data concerning radiotherapy-induced urinary toxicity, it is important to ensure derivation of the most robust models with superior predictive performance. This work explores multiple statistical-learning strategies for prediction of urinary symptoms following external beam radiotherapy of the prostate. Methods: The performance of logistic regression, elastic-net, support-vector machine, random forest, neural network, and multivariate adaptive regression splines (MARS) to predict urinary symptoms was analyzed using data from 754 participants accrued by TROG03.04-RADAR. Predictive features included dose-surface data, comorbidities, and medication-intake. Four symptoms were analyzed: dysuria, haematuria, incontinence, and frequency, each with three definitions (grade ≥ 1, grade ≥ 2 and longitudinal) with event rate between 2.3% and 76.1%. Repeated cross-validations producing matched models were implemented. A synthetic minority oversampling technique was utilized in endpoints with rare events. Parameter optimization was performed on the training data. Area under the receiver operating characteristic curve (AUROC) was used to compare performance using sample size to detect differences of ≥0.05 at the 95% confidence level. Results: Logistic regression, elastic-net, random forest, MARS, and support-vector machine were the highest-performing statistical-learning strategies in 3, 3, 3, 2, and 1 endpoints, respectively. Logistic regression, MARS, elastic-net, random forest, neural network, and support-vector machine were the best, or were not significantly worse than the best, in 7, 7, 5, 5, 3, and 1 endpoints. The best-performing statistical model was for dysuria grade ≥ 1 with AUROC ± standard deviation of 0.649 ± 0.074 using MARS. For longitudinal frequency and dysuria grade ≥ 1, all strategies produced AUROC>0.6 while all haematuria endpoints and longitudinal incontinence models produced AUROC<0.6. Conclusions: Logistic regression and MARS were most likely to be the best-performing strategy for the prediction of urinary symptoms with elastic-net and random forest producing competitive results. The predictive power of the models was modest and endpoint-dependent. New features, including spatial dose maps, may be necessary to achieve better models.
The weak land carbon sink hypothesis
Over the past three decades, assessments of the contemporary global carbon budget consistently report a strong net land carbon sink. Here, we review evidence supporting this paradigm and quantify the differences in global and Northern Hemisphere estimates of the net land sink derived from atmospheric inversion and satellite-derived vegetation biomass time series. Our analysis, combined with additional synthesis, supports a hypothesis that the net land sink is substantially weaker than commonly reported. At a global scale, our estimate of the net land carbon sink is 0.8 ± 0.7 petagrams of carbon per year from 2000 through 2019, nearly a factor of two lower than the Global Carbon Project estimate. With concurrent adjustments to ocean (+8%) and fossil fuel (−6%) fluxes, we develop a budget that partially reconciles key constraints provided by vegetation carbon, the north-south CO 2 gradient, and O 2 trends. We further outline potential modifications to models to improve agreement with a weaker land sink and describe several approaches for testing the hypothesis.
Citywide Impacts of E-Commerce: Does Parcel Delivery Travel Outweigh Household Shopping Travel Reductions?
E-commerce has facilitated online ordering of goods by households in recent years. This technological advancement has disrupted shopping related transportation. While the National Household Travel Survey (NHTS [1]) finds that household shopping frequency has declined in the last 10-20 years, deliveries by parcel delivery trucks and vans [2] have increased. However, the net effect of these phenomena on overall trip making, vehicle-miles traveled (VMT) and fuel consumption has not been quantified. From a regional planning perspective, understanding the net effect is important for informing city policies--for example, in regards to land use and transportation planning. The objective of this research is to address this gap. In this study, the net regional impact of e-commerce on transportation and fuel consumption is evaluated. The approach relies on a powerful, agent-based modeling framework (POLARIS [3]) that models decisions made by individual household and commercial agents. E-commerce demand is modeled for each household using a bilevel multinomial probit structure that evaluates e-commerce participation and ordering frequency. Last-mile delivery tours were constructed using GIS-based tools and information from a major parcel delivery company [4]. After integrating the resulting supply and demand models with all other passenger and commercial traffic within POLARIS, a traffic simulation was performed and subsequently VMT and energy consumption were analyzed. The study finds that while e-commerce has generated an increase in parcel truck delivery trips, the net effect of e-commerce is a reduction in VMT and fuel consumption due major reductions in these quantities via shopping trip reductions.
Toward Large-Scale Image Segmentation on Summit
Semantic segmentation of images is an important computer vision task that emerges in a variety of application domains such as medical imaging, robotic vision and autonomous vehicles to name a few. While these domain-specific image analysis tasks involve relatively small image sizes (~ 10 2 × 10 2 ), there are many applications that need to train machine learning models on image data with extents that are orders of magnitude larger (~10 4 × 10 4 ). Training deep neural network (DNN) models on large extent images is extremely memory-intensive and often exceeds the memory limitations of a single graphical processing unit, a hardware accelerator of choice for computer vision workloads. Here, an efficient, sample parallel approach to train U-Net models on large extent image data sets is presented. Its advantages and limitations are analyzed and near-linear strong-scaling speedup demonstrated on 256 nodes (1536 GPUs) of the Summit supercomputer. Using a single node of the Summit supercomputer, an early evaluation of a recently released model parallel framework called GPipe is demonstrated to deliver ~ 2X speedup in executing a U-Net model with an order of magnitude larger number of trainable parameters than reported before. Performance bottlenecks for pipelined training of U-Net models are identified and mitigation strategies to improve the speedups are discussed. Together, these results open up the possibility of combining both approaches into a unified scalable pipelined and data parallel algorithm to efficiently train U-Net models with very large receptive fields on data sets of ultra-large extent images.