The transient critical boundary of dynamic security region (DSR) can be approximated by a few hyper planes correlated with instability separation modes. A method to fast predict instability separation modes is propose...The transient critical boundary of dynamic security region (DSR) can be approximated by a few hyper planes correlated with instability separation modes. A method to fast predict instability separation modes is proposed for DSR calculation in power injection space. The method identifies coherent generation groups by the developed K-medoids algorithm, taking a similarity matrix derived from the reachability Grammian as the index. As an experimental result, reachability Grammian matrices under local injections are approximately invariant. It indicates that the generator coherency identifications are nearly consistent for different injections. Then instability separation modes can be predicted at the normal operating point, while average initial acceleration is considered as the measure of the critical generator group to amend the error. Moreover, based on these predicted instability separation modes, a critical point search strategy for DSR calculation is illustrated in the reduced injection space of the critical generators. The proposed method was evaluated using New England Test System, and the computation accuracy and speed in determining the practical DSR were improved.展开更多
It is well known that -nitride semiconductors can generate the magnitude of MV/cm polarization electric field which is comparable with their ionization electric fields. To take full advantage of the polarization elect...It is well known that -nitride semiconductors can generate the magnitude of MV/cm polarization electric field which is comparable with their ionization electric fields. To take full advantage of the polarization electric field, we design an N-face AlGaN solar-blind avalanche photodiode (APD) with an Al<sub>0.45</sub>Ga<sub>0.55</sub>N/Al<sub>0.3</sub>Ga<sub>0.7</sub>N heterostructure as separate absorption and multiplication (SAM) regions. The simulation results show that the N-face APDs are more beneficial to improving the avalanche gain and reducing the avalanche breakdown voltage compared with the Ga-face APDs due to the effect of the polarization electric field. Furthermore, the Al<sub>0.45</sub>Ga<sub>0.55</sub>N/Al<sub>0.3</sub>Ga<sub>0.7</sub>N heterostructure SAM regions used in APDs instead of homogeneous Al<sub>0.45</sub>Ga<sub>0.55</sub>N SAM structure can increase significantly avalanche gain because of the increased hole ionization coefficient by using the relatively low Al-content AlGaN in the multiplication region. Meanwhile, a quarter-wave AlGaN/AlN distributed Bragg reflector structure at the bottom of the device is designed to remain a solar-blind characteristic of the heterostructure SAM-APDs.展开更多
Regional turbofan aircraft,which are used for medium-short distances,have a heightened risk of high-altitude Wake Vortices(VV)because of their tail-mounted engines and high horizontal tail configurations.For some regi...Regional turbofan aircraft,which are used for medium-short distances,have a heightened risk of high-altitude Wake Vortices(VV)because of their tail-mounted engines and high horizontal tail configurations.For some regional medium-short-range turbofan aircraft,this threat is higher than that for conventionally designed aircraft.To analyze the flight safety of turbofan aircraft during cruise,this study developed a model to assess wake vortex encounters based on evolutionary high-altitude wake flow patterns.First,the high-altitude wake vortex aircraft dissipation patterns were analyzed by combining Quick Access Recorder(QAR)flight data with the wake vortex evolution model.Then,to consider the uniqueness of the medium-short-range turbofan aircraft,the severity of the wake vortex encounters was simulated using an induced roll moment coefficient.The proposed high-altitude wake vortex encounter model was able to identify and assess the highaltitude wake vortex changes,the bearing moments at different altitudes,and the atmospheric pressure conditions.Using the latest wake separation standards from the International Civil Aviation Organization(ICAO),acceptable safety wake intervals for follower aircraft in different scenarios were determined for the safety assessment.The results indicate that compared to mid and low altitudes,the high-altitude aircraft wake vortex dissipation rate is faster,the ultimate bearing moment is weaker,and the roll moment coefficient is higher,which confirm that there is elevated wake vortex encounter severity for regional turbofan aircraft.As safety is found to deteriorate when encountering wake vortices at altitudes higher than 8 km,new medium-short-range turbofan regional aircraft require higher safety margins than the latest wake separation standards.展开更多
Laminar mixing in the stirred tank is widely encountered in chemical and biological industries.Isolated mixing regions(IMRs)usually exist when the fluid medium has high viscosity,which are not conducive to mixing.In t...Laminar mixing in the stirred tank is widely encountered in chemical and biological industries.Isolated mixing regions(IMRs)usually exist when the fluid medium has high viscosity,which are not conducive to mixing.In this work,the researches on IMRs,enhancement of laminar mixing and the phenomenon of particle clustering within IMRs are reviewed.For most studies,the aim is to destroy IMRs and improve the chaotic mixing.To this end,the mechanism of chaotic mixing and the structure of IMRs were well investigated.The methods developed to destroy IMRs include off-centered agitation,dynamic mixing protocol,special designs of impellers,baffles,etc.In addition,the methods to characterize the shape and size of IMRs as well as mixing effect by experiments and simulations are summarized.However,IMRs are not always nuisance,and it may be necessary in some situations.Finally,the present engineering applications are summarized,and the prospect of the future application is predicted.For example,particle clustering will form in the co-existing system of chaotic mixing and IMRs,which can be used for solid–liquid separation and recovery of particles from high viscosity fluid.展开更多
A separated flow model with stochastical trajectories has been developed to describe the fluid flow in a bubble stirred ladle.The bubble dispersion,turbulent characteristics and gas-liquid interactions can be predicte...A separated flow model with stochastical trajectories has been developed to describe the fluid flow in a bubble stirred ladle.The bubble dispersion,turbulent characteristics and gas-liquid interactions can be predicted by this mathematical model.The bubble flow as a dispersed phase is treated in a Lagrangian frame of reference and the analysis of the turbulent flow for liquid phase is conducted in a Eulerian field.The interactions between bubbles and liquid phases are considered as a bubble source term in the control equation for a continuous phase. The Monte Carlo sampling method is used to determine the bubble trajectories.The homoge- neous flow model is also taken into consideration so that it can be compared with the sepa- rated flow model.Numerical predictions using a water model of a ladle show that the pre- dicted results of the separated flow model agree satisfactorily with the experimental results, but the prediction of the homogeneous flow model are not in good agreement with the experi- mental results.展开更多
四元数神经网络将实值神经网络推广到了四元数代数中,其在偏振合成孔径雷达奇异点补偿、口语理解、机器人控制等任务中取得了比实值神经网络更高的精度或更快的收敛速度.四元数神经网络的性能在实验中已得到广泛验证,但四元数神经网络...四元数神经网络将实值神经网络推广到了四元数代数中,其在偏振合成孔径雷达奇异点补偿、口语理解、机器人控制等任务中取得了比实值神经网络更高的精度或更快的收敛速度.四元数神经网络的性能在实验中已得到广泛验证,但四元数神经网络的理论性质及其相较于实值神经网络的优势研究较少.从表示能力的角度出发,研究四元数神经网络的理论性质及其相较于实值神经网络的优势.首先,证明了四元数神经网络使用一个非分开激活的修正线性单元(rectified linear unit,ReLU)型激活函数时的通用近似定理.其次,研究了四元数神经网络相较于实值神经网络的逼近优势.针对分开激活的ReLU型激活函数,证明了单隐层实值神经网络需要约4倍参数量才能生成与单隐层四元数神经网络相同的最大凸线性区域数.针对非分开激活的ReLU型激活函数,证明了单隐层四元数神经网络与单隐层实值神经网络间的逼近分离:四元数神经网络可用相同的隐层神经元数量与权重模长表示实值神经网络,而实值神经网络需要指数多个隐层神经元或指数大的参数才可能近似四元数神经网络.最后,模拟实验验证了理论.展开更多
基金Supported by National Natural Science Foundation of China (No.50595413)Special Fund of the National Fundamental Research of China(No.2004CB217904)+4 种基金US EPRI under Agreement EP-P29464/C9966Foundation for the Author of National Excellent Doctoral Disserta-tion (No.200439)Key Project of Ministry of Education of China(No.105047)Program for New Century Excellent Talents in University,Fok Ying Tung Education Foundation (No.104019)Innovation Fund of Tianjin Municipal (2006-09)
文摘The transient critical boundary of dynamic security region (DSR) can be approximated by a few hyper planes correlated with instability separation modes. A method to fast predict instability separation modes is proposed for DSR calculation in power injection space. The method identifies coherent generation groups by the developed K-medoids algorithm, taking a similarity matrix derived from the reachability Grammian as the index. As an experimental result, reachability Grammian matrices under local injections are approximately invariant. It indicates that the generator coherency identifications are nearly consistent for different injections. Then instability separation modes can be predicted at the normal operating point, while average initial acceleration is considered as the measure of the critical generator group to amend the error. Moreover, based on these predicted instability separation modes, a critical point search strategy for DSR calculation is illustrated in the reduced injection space of the critical generators. The proposed method was evaluated using New England Test System, and the computation accuracy and speed in determining the practical DSR were improved.
基金Supported by the State Key Project of Research and Development Plan of China under Grant No 2016YFB0400903the National Natural Science Foundation of China under Grant Nos 61634002,61274075 and 61474060+2 种基金the Key Project of Jiangsu Province under Grant No BE2016174the Anhui University Natural Science Research Project under Grant No KJ2015A153the Open Fund of State KeyLab of Optical Technologies on Nano-fabrication and Micro-engineering
文摘It is well known that -nitride semiconductors can generate the magnitude of MV/cm polarization electric field which is comparable with their ionization electric fields. To take full advantage of the polarization electric field, we design an N-face AlGaN solar-blind avalanche photodiode (APD) with an Al<sub>0.45</sub>Ga<sub>0.55</sub>N/Al<sub>0.3</sub>Ga<sub>0.7</sub>N heterostructure as separate absorption and multiplication (SAM) regions. The simulation results show that the N-face APDs are more beneficial to improving the avalanche gain and reducing the avalanche breakdown voltage compared with the Ga-face APDs due to the effect of the polarization electric field. Furthermore, the Al<sub>0.45</sub>Ga<sub>0.55</sub>N/Al<sub>0.3</sub>Ga<sub>0.7</sub>N heterostructure SAM regions used in APDs instead of homogeneous Al<sub>0.45</sub>Ga<sub>0.55</sub>N SAM structure can increase significantly avalanche gain because of the increased hole ionization coefficient by using the relatively low Al-content AlGaN in the multiplication region. Meanwhile, a quarter-wave AlGaN/AlN distributed Bragg reflector structure at the bottom of the device is designed to remain a solar-blind characteristic of the heterostructure SAM-APDs.
基金supported by the National Natural Science Foundation of China(Nos.U2333209,U1733203)the National Key R&D Program of China(No.2021YFF0603904)the Civil Aviation Administration of China(No.AQ20200019)。
文摘Regional turbofan aircraft,which are used for medium-short distances,have a heightened risk of high-altitude Wake Vortices(VV)because of their tail-mounted engines and high horizontal tail configurations.For some regional medium-short-range turbofan aircraft,this threat is higher than that for conventionally designed aircraft.To analyze the flight safety of turbofan aircraft during cruise,this study developed a model to assess wake vortex encounters based on evolutionary high-altitude wake flow patterns.First,the high-altitude wake vortex aircraft dissipation patterns were analyzed by combining Quick Access Recorder(QAR)flight data with the wake vortex evolution model.Then,to consider the uniqueness of the medium-short-range turbofan aircraft,the severity of the wake vortex encounters was simulated using an induced roll moment coefficient.The proposed high-altitude wake vortex encounter model was able to identify and assess the highaltitude wake vortex changes,the bearing moments at different altitudes,and the atmospheric pressure conditions.Using the latest wake separation standards from the International Civil Aviation Organization(ICAO),acceptable safety wake intervals for follower aircraft in different scenarios were determined for the safety assessment.The results indicate that compared to mid and low altitudes,the high-altitude aircraft wake vortex dissipation rate is faster,the ultimate bearing moment is weaker,and the roll moment coefficient is higher,which confirm that there is elevated wake vortex encounter severity for regional turbofan aircraft.As safety is found to deteriorate when encountering wake vortices at altitudes higher than 8 km,new medium-short-range turbofan regional aircraft require higher safety margins than the latest wake separation standards.
基金supports from National Key Research and Develop-ment Program(2020YFA0906804)the National Natural Science Foundation of China(21776282,21978296 and 22078229)+4 种基金the NSFC Key Program(21938009)major project(91934301)the National Key R&D Program of China(2019YFC1905805)Chemistry and Chemical Engineering Guangdong Laboratory Shantou(1922006)Innovation Academy for Green Manufacture,Chinese Academy of Sciences(IAGM2020C06)。
文摘Laminar mixing in the stirred tank is widely encountered in chemical and biological industries.Isolated mixing regions(IMRs)usually exist when the fluid medium has high viscosity,which are not conducive to mixing.In this work,the researches on IMRs,enhancement of laminar mixing and the phenomenon of particle clustering within IMRs are reviewed.For most studies,the aim is to destroy IMRs and improve the chaotic mixing.To this end,the mechanism of chaotic mixing and the structure of IMRs were well investigated.The methods developed to destroy IMRs include off-centered agitation,dynamic mixing protocol,special designs of impellers,baffles,etc.In addition,the methods to characterize the shape and size of IMRs as well as mixing effect by experiments and simulations are summarized.However,IMRs are not always nuisance,and it may be necessary in some situations.Finally,the present engineering applications are summarized,and the prospect of the future application is predicted.For example,particle clustering will form in the co-existing system of chaotic mixing and IMRs,which can be used for solid–liquid separation and recovery of particles from high viscosity fluid.
文摘A separated flow model with stochastical trajectories has been developed to describe the fluid flow in a bubble stirred ladle.The bubble dispersion,turbulent characteristics and gas-liquid interactions can be predicted by this mathematical model.The bubble flow as a dispersed phase is treated in a Lagrangian frame of reference and the analysis of the turbulent flow for liquid phase is conducted in a Eulerian field.The interactions between bubbles and liquid phases are considered as a bubble source term in the control equation for a continuous phase. The Monte Carlo sampling method is used to determine the bubble trajectories.The homoge- neous flow model is also taken into consideration so that it can be compared with the sepa- rated flow model.Numerical predictions using a water model of a ladle show that the pre- dicted results of the separated flow model agree satisfactorily with the experimental results, but the prediction of the homogeneous flow model are not in good agreement with the experi- mental results.
文摘四元数神经网络将实值神经网络推广到了四元数代数中,其在偏振合成孔径雷达奇异点补偿、口语理解、机器人控制等任务中取得了比实值神经网络更高的精度或更快的收敛速度.四元数神经网络的性能在实验中已得到广泛验证,但四元数神经网络的理论性质及其相较于实值神经网络的优势研究较少.从表示能力的角度出发,研究四元数神经网络的理论性质及其相较于实值神经网络的优势.首先,证明了四元数神经网络使用一个非分开激活的修正线性单元(rectified linear unit,ReLU)型激活函数时的通用近似定理.其次,研究了四元数神经网络相较于实值神经网络的逼近优势.针对分开激活的ReLU型激活函数,证明了单隐层实值神经网络需要约4倍参数量才能生成与单隐层四元数神经网络相同的最大凸线性区域数.针对非分开激活的ReLU型激活函数,证明了单隐层四元数神经网络与单隐层实值神经网络间的逼近分离:四元数神经网络可用相同的隐层神经元数量与权重模长表示实值神经网络,而实值神经网络需要指数多个隐层神经元或指数大的参数才可能近似四元数神经网络.最后,模拟实验验证了理论.
