Vibration energy harvesting can solve the energy supplying problem of systems like wireless sensors networks.The Galfenol cantilever beam energy harvester is suitable for this application.According to the electromecha...Vibration energy harvesting can solve the energy supplying problem of systems like wireless sensors networks.The Galfenol cantilever beam energy harvester is suitable for this application.According to the electromechanical conversion principle,the constitutive relation of Galfenol is built.A magnetization model is also established,based on the hysteresis model of Galfenol.Combining the magneto-mechanical coupling model,the constitutive relation of Galfenol and the electromagnetic induction law,the mathematical model of Galfenol vibration energy harvester is established.A hyperbolic curve-shaped cantilever beam is designed and its performance is compared to three types of cantilever beams:rectangle,trapezoidal,and triangle.The stress distribution,modal analysis and frequency response of these four shapes of beams are compared.The hyperbolic beam is more suitable for low frequency vibration harvesting.The strain on the beams,output voltage and power output response to these energy harvesters,under different natural vibration frequencies,are determined by simulation.Finally,the simulation results are compared to experimental electric outputs of all four types of prototype beams.The comparative study showed consistency between the experimental and the simulation results,and also that the peak-to-peak induction voltage value of the hyperbolic beam is larger than other shapes of energy harvesters,whose average value is at 269.8 mV.The maximum power output of the hyperbolic beam is 403.8μW when connected with a 50Ω resistance.展开更多
To explore the potential of distributed training on deep neural networks,we implement several distributed algorithms with the basis of swFlow on the world-leading supercomputer,Sunway TaihuLight.Based on two naive des...To explore the potential of distributed training on deep neural networks,we implement several distributed algorithms with the basis of swFlow on the world-leading supercomputer,Sunway TaihuLight.Based on two naive designs of parameter server and ring all-reduce,we present the limitation of the communication model and discuss the optimizations for adapting the five-level interconnect architecture of Sunway system.To reduce the communication bottleneck on large scale system,multi-severs and hierarchical ring all-reduce models are introduced.With a benchmark from deep learning-based cancerous region detection algorithm,the average parallel efficiency obtains over 80%for at most 1024 processors.It reveals the great opportunity for joint combination of deep learning and HPC system.展开更多
基金supported by the Natural Science Foundation of Zhejiang Province(Grant LY17E050026)the National Natural Science Foundation of China(Grant 51805126).
文摘Vibration energy harvesting can solve the energy supplying problem of systems like wireless sensors networks.The Galfenol cantilever beam energy harvester is suitable for this application.According to the electromechanical conversion principle,the constitutive relation of Galfenol is built.A magnetization model is also established,based on the hysteresis model of Galfenol.Combining the magneto-mechanical coupling model,the constitutive relation of Galfenol and the electromagnetic induction law,the mathematical model of Galfenol vibration energy harvester is established.A hyperbolic curve-shaped cantilever beam is designed and its performance is compared to three types of cantilever beams:rectangle,trapezoidal,and triangle.The stress distribution,modal analysis and frequency response of these four shapes of beams are compared.The hyperbolic beam is more suitable for low frequency vibration harvesting.The strain on the beams,output voltage and power output response to these energy harvesters,under different natural vibration frequencies,are determined by simulation.Finally,the simulation results are compared to experimental electric outputs of all four types of prototype beams.The comparative study showed consistency between the experimental and the simulation results,and also that the peak-to-peak induction voltage value of the hyperbolic beam is larger than other shapes of energy harvesters,whose average value is at 269.8 mV.The maximum power output of the hyperbolic beam is 403.8μW when connected with a 50Ω resistance.
基金supported by the National Key Research and Development Program of China(Grants no.2016YF B1000403)the Fundamental Research Funds for the Central Universities of China.
文摘To explore the potential of distributed training on deep neural networks,we implement several distributed algorithms with the basis of swFlow on the world-leading supercomputer,Sunway TaihuLight.Based on two naive designs of parameter server and ring all-reduce,we present the limitation of the communication model and discuss the optimizations for adapting the five-level interconnect architecture of Sunway system.To reduce the communication bottleneck on large scale system,multi-severs and hierarchical ring all-reduce models are introduced.With a benchmark from deep learning-based cancerous region detection algorithm,the average parallel efficiency obtains over 80%for at most 1024 processors.It reveals the great opportunity for joint combination of deep learning and HPC system.
