Cryogenic superconducting devices such as a Superconducting Magnetic Energy Storage device can require current leads that must operate under conditions where the current is delivered in pulses. The ...focus of this paper was to observe the electrical and thermal behavior of conduction-cooled superconducting current leads in hopes to develop a better understanding of normal zone propagation when pulsed currents are applied for future design purposes. Leads made of copper tubing soldered in parallel to 2-G HTS tape were made to allow the normal zone propagation to be observed. The leads were evacuated and connected to a power source at one end and the cold finger of a cryocooler at the other end. The test fixture allowed for the temperature and voltage distribution measurements to be made. Pulsed dc current was supplied to allow the observation of lead response as a function of current level for different ramp cycles. A transient 1-D numerical model was also developed to calculate the current lead temperature in space and time. The lead dimensions used in the model match those used in the experiment. Results of the experiment validate the numerical model and are used to develop a better understanding of the thermal and electrical properties of superconducting current leads under pulsed conditions.
Quench Detection and Protection of an HTS Coil Sheehan, Evan; Pfotenhauer, John; Miller, Franklin ...
IOP conference series. Materials Science and Engineering,
12/2017, Letnik:
278, Številka:
1
Journal Article
Recenzirano
Odprti dostop
A pulsed, modular HTS magnet for energy storage applications was constructed and tested. Charge and discharge pulses were accomplished in about 1 second. A recuperative cryogenic cooling system ...supplies 42 to 80 Kelvin helium gas to the magnet. A practical solution to overvoltage and overcurrent protection has been implemented digitally using LabVIEW. Voltages as little as 46 μV greater than the expected value trigger the protection system, which stops the pulse profile and begins an immediate current ramp down to zero over 1 second. The protection system has displayed its effectiveness in HTS transition detection and damage prevention. Experimentation has demonstrated that current pulses on the order of seconds with amplitudes of up to 110 Amps can be achieved for extended periods. Higher currents produce joint heating in excess of the available cooling from the existing cryogenic system.
HTS coils designed to carry pulsed currents have been built and tested demonstrating fast charge and discharge times on the order of a second. Gaseous Helium from 40 to 80 K is used as a coolant. ...Losses due to ac, hysteresis, and transport currents associated with current entry into and exiting from HTS are small and do not produce thermal transients that induce current sharing or normal transitions for reasonable margins and operating temperatures. Short lengths of HTS and small coils show current sharing before normal transitions, whereas current sharing voltages are not identified in large coils since current sharing voltages are small compared to inductive and resistive voltages. Pulse shape, duration, and frequency are varied with no degradation in performance.
High-effectiveness heat exchangers are a ubiquitous component of cryogenic systems, but their performance typically falls short of model-based expectations. The following thesis details modeling ...efforts of a heat exchanger designed to achieve an effectiveness in excess of 99% within a prescribed volume, weight, pressure drop, and operating conditions for SHI Cryogenics of America. With such a high effectiveness requirement, simulation efforts focused on making minimal assumptions and detailed models. Axial conduction and parallel flow passage imbalance are major contributing factors to heat exchanger inefficiency. Consequently, a staggered stacked slotted plate geometry was chosen as the most promising design to achieve the desired effectiveness. The large number of design parameters initially spurred the development of model interfacing with the University of Wisconsin Center for High Throughput Computing, but licensing issues limited its usability. Instead, the number of parameters was thoughtfully reduced based on manufacturing limitations and scaling considerations. Manufacturing options for machining the slots and bonding the plates were investigated. A MATLAB model which included axial conduction, parasitic heat loads, and material property variation accounted for losses in order to avoid inflating the efficiency estimate. Fluent was used to develop accurate Nusselt number and Darcy friction factor correlations, which were used as inputs to the MATLAB model. The MATLAB model was validated by comparing its results to an analytic, constant property effectiveness-NTU solution as well as experimental data from a similar heat exchanger. The Fluent correlations were compared to other correlations typically used for slotted plate heat exchangers.
Progress on the UN Sustainable Development Goals (SDGs) is hampered by a persistent lack of data regarding key social, environmental, and economic indicators, particularly in developing countries. ...For example, data on poverty - the first of seventeen SDGs - is both spatially sparse and infrequently collected in Sub-Saharan Africa due to the high cost of surveys. Here we propose a novel method for estimating socioeconomic indicators using open-source, geolocated textual information from Wikipedia articles. We demonstrate that modern NLP techniques can be used to predict community-level asset wealth and education outcomes using nearby geolocated Wikipedia articles. When paired with nightlights satellite imagery, our method outperforms all previously published benchmarks for this prediction task, indicating the potential of Wikipedia to inform both research in the social sciences and future policy decisions.
Efficient Conditional Pre-training for Transfer Learning Chakraborty, Shuvam; Uzkent, Burak; Ayush, Kumar ...
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW),
2022-June
Conference Proceeding
Almost all the state-of-the-art neural networks for computer vision tasks are trained by (1) pre-training on a large-scale dataset and (2) finetuning on the target dataset. This strategy helps reduce ...dependence on the target dataset and improves convergence rate and generalization on the target task. Although pre-training on large-scale datasets is very useful for new methods or models, its foremost disadvantage is high training cost. To address this, we propose efficient filtering methods to select relevant subsets from the pre-training dataset. Additionally, we discover that lowering image resolutions in the pre-training step offers a great trade-off between cost and performance. We validate our techniques by pre-training on ImageNet in both the unsupervised and supervised settings and finetuning on a diverse collection of target datasets and tasks. Our proposed methods drastically reduce pre-training cost and provide strong performance boosts. Finally, we improve the current standard of ImageNet pre-training by 1-3% by tuning available models on our subsets and pre-training on a dataset filtered from a larger scale dataset.
Almost all the state-of-the-art neural networks for computer vision tasks are trained by (1) pre-training on a large-scale dataset and (2) finetuning on the target dataset. This strategy helps reduce ...dependence on the target dataset and improves convergence rate and generalization on the target task. Although pre-training on large-scale datasets is very useful, its foremost disadvantage is high training cost. To address this, we propose efficient filtering methods to select relevant subsets from the pre-training dataset. Additionally, we discover that lowering image resolutions in the pre-training step offers a great trade-off between cost and performance. We validate our techniques by pre-training on ImageNet in both the unsupervised and supervised settings and finetuning on a diverse collection of target datasets and tasks. Our proposed methods drastically reduce pre-training cost and provide strong performance boosts. Finally, we improve standard ImageNet pre-training by 1-3% by tuning available models on our subsets and pre-training on a dataset filtered from a larger scale dataset.
Despite recent progress in computer vision, finegrained interpretation of satellite images remains challenging because of a lack of labeled training data. To overcome this limitation, we construct a ...novel dataset called WikiSatNet by pairing georeferenced Wikipedia articles with satellite imagery of their corresponding locations. We then propose two strategies to learn representations of satellite images by predicting properties of the corresponding articles from the images. Leveraging this new multi-modal dataset, we can drastically reduce the quantity of human-annotated labels and time required for downstream tasks. On the recently released fMoW dataset, our pre-training strategies can boost the performance of a model pre-trained on ImageNet by up to 4:5% in F1 score.