Artificial vascular graft(AVG)fistula is widely used for hemodialysis treatment in patients with renal failure.However,it has poor elasticity and compliance,leading to stenosis and thrombosis.The ideal artificial bloo...Artificial vascular graft(AVG)fistula is widely used for hemodialysis treatment in patients with renal failure.However,it has poor elasticity and compliance,leading to stenosis and thrombosis.The ideal artificial blood vessel for dialysis should replicate the structure and components of a real artery,which is primarily maintained by collagen in the extracellular matrix(ECM)of arterial cells.Studies have revealed that in hepatitis B virus(HBV)-induced liver fibrosis,hepatic stellate cells(HSCs)become hyperactive and produce excessive ECM fibers.Furthermore,mechanical stimulation can encourage ECM secretion and remodeling of a fiber structure.Based on the above factors,we transfected HSCs with the hepatitis B viral X(HBX)gene for simulating the process of HBV infection.Subsequently,these HBX-HSCs were implanted into a polycaprolactonepolyurethane(PCL-PU)bilayer scaffold in which the inner layer is dense and the outer layer consists of pores,which was mechanically stimulated to promote the secretion of collagen nanofiber from the HBX-HSCs and to facilitate crosslinking with the scaffold.We obtained an ECM-PCL-PU composite bionic blood vessel that could act as access for dialysis after decellularization.Then,the vessel scaffold was implanted into a rabbit’s neck arteriovenous fistula model.It exhibited strong tensile strength and smooth blood flow and formed autologous blood vessels in the rabbit’s body.Our study demonstrates the use of human cells to create biomimetic dialysis blood vessels,providing a novel approach for creating clinical vascular access for dialysis.展开更多
当前的目标分割模型难以兼顾分割性能与推断效率,为此提出一种基于尺度注意知识迁移的自蒸馏目标分割方法。首先,构建了一个仅利用主干特征的目标分割网络作为推断网络,实现高效的前向推断过程。其次,提出了一种基于尺度注意知识的自蒸...当前的目标分割模型难以兼顾分割性能与推断效率,为此提出一种基于尺度注意知识迁移的自蒸馏目标分割方法。首先,构建了一个仅利用主干特征的目标分割网络作为推断网络,实现高效的前向推断过程。其次,提出了一种基于尺度注意知识的自蒸馏学习模型:一方面,设计了具有尺度注意机制的金字塔特征模块,利用尺度注意机制自适应地捕获不同语义水平的上下文信息,提取更具区分性的自蒸馏知识;另一方面,融合交叉熵、KL(Kullback-Leibler)散度和L2距离构造蒸馏损失,高效驱动蒸馏知识向分割网络迁移,提升泛化性能。该方法在COD(Camouflaged Object Detection)、DUT-O(Dalian University of Technology-OMRON)、SOC(Salient Objects in Clutter)等五个目标分割数据集上进行了验证:将所提推断网络作为基准网络,所提自蒸馏模型分割性能在Fβ指标上平均提升3.01%,比免教师(TF)自蒸馏模型增加了1.00%;所提网络与近期的残差分割网络(R2Net)相比,参数量减少了2.33×10^(6),推断帧率提升了2.53%,浮点运算量减少了40.50%,分割性能提升了0.51%。实验结果表明:所提方法能有效兼顾性能与效率,适用于计算和存储资源受限的应用场景。展开更多
An in-depth understanding of the fracture behavior and mechanism of metallic shells under internal explosive loading can help develop material designs for warheads and regulate the quantity and mass distribution of th...An in-depth understanding of the fracture behavior and mechanism of metallic shells under internal explosive loading can help develop material designs for warheads and regulate the quantity and mass distribution of the fragments formed.This study investigated the fragmentation performance of a new high-carbon silicon-manganese(HCSiMn)steel cylindrical shell through fragment recovery experiments.Compared with the conventional 45Cr steel shell,the number of small mass fragments produced by the HCSi Mn steel shell was significantly increased with a scale parameter of 0.57 g fitted by the Weibull distribution model.The fragmentation process of the HCSi Mn shell exhibited more brittle tensile fracture characteristics,with the microcrack damage zone on the outer surface being the direct cause of its high fragmentation.On the one hand,the doping of alloy elements resulted in grain refinement by forming metallographic structure of tempered sorbite,so that microscopic intergranular fracture reduces the characteristic mass of the fragments;on the other hand,the distribution of alloy carbides can exert a"pinning"effect on the substrate grains,causing more initial cracks to form and propagate along the brittle carbides,further improving the shell fragmentation.Although the killing power radius for light armored vehicles was slightly reduced by about 6%,the dense killing radius of HCSiMn steel projectile against personnel can be significantly increased by about 26%based on theoretical assessment.These results provided an experimental basis for high fragmentation warhead design,and to some extent,revealed the correlation mechanism between metallographic structure and shell fragmentation.展开更多
Under low signal-to-noise ratio(SNR)and burst noise conditions,the speech enhancement effect of existing deep learning network models is not satisfactory.In contrast,humans can exploit the long-term correlation of spe...Under low signal-to-noise ratio(SNR)and burst noise conditions,the speech enhancement effect of existing deep learning network models is not satisfactory.In contrast,humans can exploit the long-term correlation of speech to form an integrated perception of different speech signals.Thus,describing the long-term dependencies of speech can help improve the enhancement performance under low SNR and burst noise conditions.Inspired by this feature,a time domain end-to-end monaural speech enhancement model TU-net that fuses the multi-head self-attention mechanism and U-net deep network is proposed.The TU-net model adopts the codec layer structure of U-net to achieve multi-scale feature fusion.It introduces the dual-path Transformer module using the multi-head self-attention mechanism to calculate the speech mask and better model long-term correlation.The TU-net model is trained with a weighted sum loss function in the time,time-frequency,and perceptual domains.Simulation experiments are carried out and the results show that with maintaining relatively fewer network model parameters,TU-net outperforms other similar monaural enhancement network models in several evaluation metrics such as perceptual evaluation of speech quality(PESQ),short-time objective intelligibility(STOI)and SNR gain under low SNR and burst noise conditions.展开更多
Because of potential high energy densities,microfluidic fuel cells can serve as micro-scale power sources.Because microfluidic fuel cells typically operate in the co-laminar flow regime to enable a membrane-less desig...Because of potential high energy densities,microfluidic fuel cells can serve as micro-scale power sources.Because microfluidic fuel cells typically operate in the co-laminar flow regime to enable a membrane-less design,they generally suffer from severe mass transfer limitations with respect to diffusion transport.To address this issue,a novel channel design that integrates slanted groove micro-mixers on the side walls of the channel is proposed.Numerical modeling on the design of groove micro-mixers and grooveless design demonstrates a mass transfer enhancement that has a 115%higher limiting current density and well-controlled convective mixing between the oxidant and the fuel streams with the use of slanted groove micro-mixers.Moreover,the growth of the thickness of the depletion boundary layer is found to be terminated within approximately 2 mm from the channel entrance,which is distinct from the constantly growing pattern in the grooveless design.In addition,a simplified mass transfer model capable of modeling the mass transfer prFocess with the presence of the transverse secondary flow is developed.Further,a dimensionless correlation is derived to analyze the effects of the design parameters on the limiting current density.The present theoretical study paves the way towards an optimal design of a microfluidic fuel cell integrating groove micro-mixers.展开更多
We report an efficient mid-infrared extracavity optical parametric oscillator(OPO)based on the nonlinear crystal BaGa4Se7 pumped by a diode-side-pumped Q-switched Nd:Y3 Al5O12(Nd:YAG)laser.The maximum pulse energy of ...We report an efficient mid-infrared extracavity optical parametric oscillator(OPO)based on the nonlinear crystal BaGa4Se7 pumped by a diode-side-pumped Q-switched Nd:Y3 Al5O12(Nd:YAG)laser.The maximum pulse energy of 1.03mJ at 4.25μm is obtained with the repetition rate of 10 Hz and pulse width of 12.6 ns when the pump energy was 13.5mJ,corresponding to an optical-to-optical conversion efficiency of 7.6% from 1.064μm to 4.25μm.The idler wave slope conversion efficiency was 12%.To the best of our knowledge,it is the highest reported conversion efficiency for the compact BaGa4Se7 OPO driven by the Nd:YAG laser.展开更多
基金supported by the National Natural Science Foundation of China(No.81770294)the Natural Science Foundation of Fujian Province(No.2023J05261),China.
文摘Artificial vascular graft(AVG)fistula is widely used for hemodialysis treatment in patients with renal failure.However,it has poor elasticity and compliance,leading to stenosis and thrombosis.The ideal artificial blood vessel for dialysis should replicate the structure and components of a real artery,which is primarily maintained by collagen in the extracellular matrix(ECM)of arterial cells.Studies have revealed that in hepatitis B virus(HBV)-induced liver fibrosis,hepatic stellate cells(HSCs)become hyperactive and produce excessive ECM fibers.Furthermore,mechanical stimulation can encourage ECM secretion and remodeling of a fiber structure.Based on the above factors,we transfected HSCs with the hepatitis B viral X(HBX)gene for simulating the process of HBV infection.Subsequently,these HBX-HSCs were implanted into a polycaprolactonepolyurethane(PCL-PU)bilayer scaffold in which the inner layer is dense and the outer layer consists of pores,which was mechanically stimulated to promote the secretion of collagen nanofiber from the HBX-HSCs and to facilitate crosslinking with the scaffold.We obtained an ECM-PCL-PU composite bionic blood vessel that could act as access for dialysis after decellularization.Then,the vessel scaffold was implanted into a rabbit’s neck arteriovenous fistula model.It exhibited strong tensile strength and smooth blood flow and formed autologous blood vessels in the rabbit’s body.Our study demonstrates the use of human cells to create biomimetic dialysis blood vessels,providing a novel approach for creating clinical vascular access for dialysis.
文摘当前的目标分割模型难以兼顾分割性能与推断效率,为此提出一种基于尺度注意知识迁移的自蒸馏目标分割方法。首先,构建了一个仅利用主干特征的目标分割网络作为推断网络,实现高效的前向推断过程。其次,提出了一种基于尺度注意知识的自蒸馏学习模型:一方面,设计了具有尺度注意机制的金字塔特征模块,利用尺度注意机制自适应地捕获不同语义水平的上下文信息,提取更具区分性的自蒸馏知识;另一方面,融合交叉熵、KL(Kullback-Leibler)散度和L2距离构造蒸馏损失,高效驱动蒸馏知识向分割网络迁移,提升泛化性能。该方法在COD(Camouflaged Object Detection)、DUT-O(Dalian University of Technology-OMRON)、SOC(Salient Objects in Clutter)等五个目标分割数据集上进行了验证:将所提推断网络作为基准网络,所提自蒸馏模型分割性能在Fβ指标上平均提升3.01%,比免教师(TF)自蒸馏模型增加了1.00%;所提网络与近期的残差分割网络(R2Net)相比,参数量减少了2.33×10^(6),推断帧率提升了2.53%,浮点运算量减少了40.50%,分割性能提升了0.51%。实验结果表明:所提方法能有效兼顾性能与效率,适用于计算和存储资源受限的应用场景。
基金funded by the National Natural Science Foundation of China (Grant Nos.12302444 and 12202349)。
文摘An in-depth understanding of the fracture behavior and mechanism of metallic shells under internal explosive loading can help develop material designs for warheads and regulate the quantity and mass distribution of the fragments formed.This study investigated the fragmentation performance of a new high-carbon silicon-manganese(HCSiMn)steel cylindrical shell through fragment recovery experiments.Compared with the conventional 45Cr steel shell,the number of small mass fragments produced by the HCSi Mn steel shell was significantly increased with a scale parameter of 0.57 g fitted by the Weibull distribution model.The fragmentation process of the HCSi Mn shell exhibited more brittle tensile fracture characteristics,with the microcrack damage zone on the outer surface being the direct cause of its high fragmentation.On the one hand,the doping of alloy elements resulted in grain refinement by forming metallographic structure of tempered sorbite,so that microscopic intergranular fracture reduces the characteristic mass of the fragments;on the other hand,the distribution of alloy carbides can exert a"pinning"effect on the substrate grains,causing more initial cracks to form and propagate along the brittle carbides,further improving the shell fragmentation.Although the killing power radius for light armored vehicles was slightly reduced by about 6%,the dense killing radius of HCSiMn steel projectile against personnel can be significantly increased by about 26%based on theoretical assessment.These results provided an experimental basis for high fragmentation warhead design,and to some extent,revealed the correlation mechanism between metallographic structure and shell fragmentation.
基金supported by the National Natural Science Foundation of China(62071484)。
文摘Under low signal-to-noise ratio(SNR)and burst noise conditions,the speech enhancement effect of existing deep learning network models is not satisfactory.In contrast,humans can exploit the long-term correlation of speech to form an integrated perception of different speech signals.Thus,describing the long-term dependencies of speech can help improve the enhancement performance under low SNR and burst noise conditions.Inspired by this feature,a time domain end-to-end monaural speech enhancement model TU-net that fuses the multi-head self-attention mechanism and U-net deep network is proposed.The TU-net model adopts the codec layer structure of U-net to achieve multi-scale feature fusion.It introduces the dual-path Transformer module using the multi-head self-attention mechanism to calculate the speech mask and better model long-term correlation.The TU-net model is trained with a weighted sum loss function in the time,time-frequency,and perceptual domains.Simulation experiments are carried out and the results show that with maintaining relatively fewer network model parameters,TU-net outperforms other similar monaural enhancement network models in several evaluation metrics such as perceptual evaluation of speech quality(PESQ),short-time objective intelligibility(STOI)and SNR gain under low SNR and burst noise conditions.
基金supported by the National Natural Science Foundation of China(No.51606164).
文摘Because of potential high energy densities,microfluidic fuel cells can serve as micro-scale power sources.Because microfluidic fuel cells typically operate in the co-laminar flow regime to enable a membrane-less design,they generally suffer from severe mass transfer limitations with respect to diffusion transport.To address this issue,a novel channel design that integrates slanted groove micro-mixers on the side walls of the channel is proposed.Numerical modeling on the design of groove micro-mixers and grooveless design demonstrates a mass transfer enhancement that has a 115%higher limiting current density and well-controlled convective mixing between the oxidant and the fuel streams with the use of slanted groove micro-mixers.Moreover,the growth of the thickness of the depletion boundary layer is found to be terminated within approximately 2 mm from the channel entrance,which is distinct from the constantly growing pattern in the grooveless design.In addition,a simplified mass transfer model capable of modeling the mass transfer prFocess with the presence of the transverse secondary flow is developed.Further,a dimensionless correlation is derived to analyze the effects of the design parameters on the limiting current density.The present theoretical study paves the way towards an optimal design of a microfluidic fuel cell integrating groove micro-mixers.
基金supported by the Science and Technology on Plasma Physics Laboratory(No.6142A04030403)the Innovation Fund of the Chinese Academy of Sciences(No.CXJJ-17-M164)
文摘We report an efficient mid-infrared extracavity optical parametric oscillator(OPO)based on the nonlinear crystal BaGa4Se7 pumped by a diode-side-pumped Q-switched Nd:Y3 Al5O12(Nd:YAG)laser.The maximum pulse energy of 1.03mJ at 4.25μm is obtained with the repetition rate of 10 Hz and pulse width of 12.6 ns when the pump energy was 13.5mJ,corresponding to an optical-to-optical conversion efficiency of 7.6% from 1.064μm to 4.25μm.The idler wave slope conversion efficiency was 12%.To the best of our knowledge,it is the highest reported conversion efficiency for the compact BaGa4Se7 OPO driven by the Nd:YAG laser.
