Massive computational complexity and memory requirement of artificial intelligence models impede their deploy-ability on edge computing devices of the Internet of Things(IoT).While Power-of-Two(PoT)quantization is pro...Massive computational complexity and memory requirement of artificial intelligence models impede their deploy-ability on edge computing devices of the Internet of Things(IoT).While Power-of-Two(PoT)quantization is pro-posed to improve the efficiency for edge inference of Deep Neural Networks(DNNs),existing PoT schemes require a huge amount of bit-wise manipulation and have large memory overhead,and their efficiency is bounded by the bottleneck of computation latency and memory footprint.To tackle this challenge,we present an efficient inference approach on the basis of PoT quantization and model compression.An integer-only scalar PoT quantization(IOS-PoT)is designed jointly with a distribution loss regularizer,wherein the regularizer minimizes quantization errors and training disturbances.Additionally,two-stage model compression is developed to effectively reduce memory requirement,and alleviate bandwidth usage in communications of networked heterogenous learning systems.The product look-up table(P-LUT)inference scheme is leveraged to replace bit-shifting with only indexing and addition operations for achieving low-latency computation and implementing efficient edge accelerators.Finally,comprehensive experiments on Residual Networks(ResNets)and efficient architectures with Canadian Institute for Advanced Research(CIFAR),ImageNet,and Real-world Affective Faces Database(RAF-DB)datasets,indicate that our approach achieves 2×∼10×improvement in the reduction of both weight size and computation cost in comparison to state-of-the-art methods.A P-LUT accelerator prototype is implemented on the Xilinx KV260 Field Programmable Gate Array(FPGA)platform for accelerating convolution operations,with performance results showing that P-LUT reduces memory footprint by 1.45×,achieves more than 3×power efficiency and 2×resource efficiency,compared to the conventional bit-shifting scheme.展开更多
Molecular Dynamics(MD)simulation for computing Interatomic Potential(IAP)is a very important High-Performance Computing(HPC)application.MD simulation on particles of experimental relevance takes huge computation time,...Molecular Dynamics(MD)simulation for computing Interatomic Potential(IAP)is a very important High-Performance Computing(HPC)application.MD simulation on particles of experimental relevance takes huge computation time,despite using an expensive high-end server.Heterogeneous computing,a combination of the Field Programmable Gate Array(FPGA)and a computer,is proposed as a solution to compute MD simulation efficiently.In such heterogeneous computation,communication between FPGA and Computer is necessary.One such MD simulation,explained in the paper,is the(Artificial Neural Network)ANN-based IAP computation of gold(Au_(147)&Au_(309))nanoparticles.MD simulation calculates the forces between atoms and the total energy of the chemical system.This work proposes the novel design and implementation of an ANN IAP-based MD simulation for Au_(147)&Au_(309) using communication protocols,such as Universal Asynchronous Receiver-Transmitter(UART)and Ethernet,for communication between the FPGA and the host computer.To improve the latency of MD simulation through heterogeneous computing,Universal Asynchronous Receiver-Transmitter(UART)and Ethernet communication protocols were explored to conduct MD simulation of 50,000 cycles.In this study,computation times of 17.54 and 18.70 h were achieved with UART and Ethernet,respectively,compared to the conventional server time of 29 h for Au_(147) nanoparticles.The results pave the way for the development of a Lab-on-a-chip application.展开更多
The flexibility in radiotherapy can be improved if patients can be moved between any one of the department’s medical linear accelerators (LINACs) without the need to change anything in the patient’s treatment plan. ...The flexibility in radiotherapy can be improved if patients can be moved between any one of the department’s medical linear accelerators (LINACs) without the need to change anything in the patient’s treatment plan. For this to be possible, the dosimetric characteristics of the various accelerators must be the same, or nearly the same. The purpose of this work is to describe further and compare measurements and parameters after the initial vendor-recommended beam matching of the five LINACs. Deviations related to dose calculations and to beam matched accelerators may compromise treatment accuracy. The safest and most practical way to ensure that all accelerators are within clinical acceptable accuracy is to include TPS calculations in the LINACs matching evaluation. Treatment planning system (TPS) was used to create three photons plans with different field sizes 3 × 3 cm, 10 × 10 cm and 25 × 25 cm at a depth of 4.5 cm in Perspex. Calculated TPS plans were sent to Mosaiq to be delivered by five LINACs. TPS plans were compared with five LINACs measurements data using Gamma analyses of 2% and 2 mm. The results suggest that for four out of the five LINACs, there was generally good agreement, less than a 2% deviation between the planned dose distribution and the measured dose distribution. However, one specific LINAC named “Asterix” exhibited a deviation of 2.121% from the planned dose. The results show that all of the LINACs’ performance were within the acceptable deviation and delivering radiation dose consistently and accurately.展开更多
Deep neural networks(DNN)are widely used in image recognition,image classification,and other fields.However,as the model size increases,the DNN hardware accelerators face the challenge of higher area overhead and ener...Deep neural networks(DNN)are widely used in image recognition,image classification,and other fields.However,as the model size increases,the DNN hardware accelerators face the challenge of higher area overhead and energy consumption.In recent years,stochastic computing(SC)has been considered a way to realize deep neural networks and reduce hardware consumption.A probabilistic compensation algorithm is proposed to solve the accuracy problem of stochastic calculation,and a fully parallel neural network accelerator based on a deterministic method is designed.The software simulation results show that the accuracy of the probability compensation algorithm on the CIFAR-10 data set is 95.32%,which is 14.98%higher than that of the traditional SC algorithm.The accuracy of the deterministic algorithm on the CIFAR-10 dataset is 95.06%,which is 14.72%higher than that of the traditional SC algorithm.The results of Very Large Scale Integration Circuit(VLSI)hardware tests show that the normalized energy efficiency of the fully parallel neural network accelerator based on the deterministic method is improved by 31%compared with the circuit based on binary computing.展开更多
A novel TU derivative, N-phenyl-N＇-（γ-triethoxysilane）-propyl thiourea （STU）, is prepared and its binary accelerator system is investigated in detail. Compared to the control references, the optimum curing time ...A novel TU derivative, N-phenyl-N＇-（γ-triethoxysilane）-propyl thiourea （STU）, is prepared and its binary accelerator system is investigated in detail. Compared to the control references, the optimum curing time of NR compounds with STU is the shortest, indicating a more nucleophilic reaction occurs. The Py-GC/MS results present that the phenyl iso- thiocyanate fragment still remains in the NPUSTU compounds with or without extracting treatment, but no silane segment can be found in the vulcanizate with extracting treatment. Vibrations of C=S, NH and aromatic ring in FTIR experiments and a new methyne carbon peak, as well as the peaks of phenyl group of STU, in the solid state 13C-NMR experiments are found in the NR/STU vulcanizate with extracting treatment. Moreover, the crosslinking density of vulcanizates with STU evolves to lower level, indicating the sulfur atom of STU does not contribute to the sulfur crosslinking. Therefore, a new vulcanization kinetic mechanism of STU is propounded that the thiourea groups can graft to the rubber main chains as pendant groups by chemical bonds during the vulcanization process, which is in accordance with the experimental observations quite well.展开更多
Superconducting linear accelerators(SCL)have a high acceleration gradient and are capable of operating in a high-duty factor mode.For high-power and high-intensity SCL,the design of beam dynamics generally follows the...Superconducting linear accelerators(SCL)have a high acceleration gradient and are capable of operating in a high-duty factor mode.For high-power and high-intensity SCL,the design of beam dynamics generally follows the principle that the zero-current periodic phase advance(σ0)of each degree of freedom is less than 90°to avoid envelope instability caused by space charge.However,this principle is obtained under the condition of a completely periodic focusing channel,and it is ambiguous for pseudoperiodic structures,such as linear accelerators.Although transverse beam dynamics without acceleration have been studied by other researchers,it appears that there are some connections between pure 2D and 3D beam dynamics.Based on these two points,five focusing schemes for the solenoid and quadrupole doublet channels were designed to simulate the beam behavior with non-constantσ0.Among them,the four schemes follow the characteristics of variation in the zero-current longitudinal phase advance(σ0l)under a constant acceleration gradient and synchronous phase.The zero-current transverse phase advance(σ0t)is consistent withσ0l,based on the equipartition requirement.The initialσ0twas set to 120°,110°,100°,and 90°,and was then gradually decreased to approximately 40°at the end of the channel.The last scheme maintains the maximumσ0tof88°by reducing the acceleration gradient of the corresponding cavities,until the point at whichσ0tequals88°with a normal gradient.Using the stopbands obtained from the linearized envelope equations and multiparticle particle-in-cell(PIC)simulations,the transport properties of both continuous and 3D-bunched beams with the acceleration of the five focusing schemes were studied.It was found that for a CW beam,when tune depression>0.7,σ0tcan break through 90°when the beams were transported in both solenoid and quadrupole doublet periodic focusing channels.When tune depression<0.7,the conclusions were different.For the solenoid focusing system,σ0tcan partially break through 90°,and the beam quality is not significantly affected.For the quadrupole doublet focusing system,a partial breakthrough of 90°has a greater impact on the beam quality.The same conclusions were obtained for a bunched beam with acceleration.展开更多
The study revealed that ordinary acceler- ators for portlandcement also improves early strength of Belite cement remarkably, butone serious defect exists Which lead to a high strength reduction atthe age of 28 days. T...The study revealed that ordinary acceler- ators for portlandcement also improves early strength of Belite cement remarkably, butone serious defect exists Which lead to a high strength reduction atthe age of 28 days. The paper analyzes the action mechanism ofordinary accelerators on Belite cement. It shows that theseaccelerators can not accelerate the hydration of β-C_2S directly.展开更多
In this paper, the origin and type of radiation hazards as well as the main aspects of radiation protection for low-energy accelerators are discussed in general, and the problems of radiation protection and the experi...In this paper, the origin and type of radiation hazards as well as the main aspects of radiation protection for low-energy accelerators are discussed in general, and the problems of radiation protection and the experimental results of the operational monitoring of the five accelerators in the institute of Nuclear Science and Technology,Sichuan University, namely, one 1.2 M cyclofron, two Cockroft-waltons and two Van de Graafts, as well as a powerful electron accelerator for industrial irradiation are described. The discussion and evaluation are made according to the requirments of the National standards GB 5172-85.展开更多
In 2014 some research works have been carried out by the Electron Accelerator Group at IMP, including thehigh energy electron radiography (HEER), the electron accelerator design with short pulse for HEER, and a newtyp...In 2014 some research works have been carried out by the Electron Accelerator Group at IMP, including thehigh energy electron radiography (HEER), the electron accelerator design with short pulse for HEER, and a newtype low energy electrostatic electron accelerator for industrial applications.The verification experiments of HEER have been accomplished on THU electron accelerator in 2013, so moreexperimental works were proposed to continue with complicated parameters. This year, a specialized beam linefor HEER experiment was designed and built, which consists of two dipole magnets, ten quadrapole magnets andsome vacuum chambers with diagnostic probes.展开更多
To understand the nature of"matter",scientists have to answer some very basic questions.For example,why there are so many"elementary particles"?Are they really indivisible building blocks of our co...To understand the nature of"matter",scientists have to answer some very basic questions.For example,why there are so many"elementary particles"?Are they really indivisible building blocks of our cosmos?Is the cosmos following some unknown physical laws,which potentially subvert our current understanding of the whole world?How should we understand the"dark matter"in the cosmos?Could the four fundamental forces comply with any common physical law in the same framework?展开更多
This paper provides a review of the compact intense electron-beam accelerators (IEBAs) based on liquid pulse forming lines (PFLs) that havebeen developed at the National University of Defense Technology (NUDT) in Chin...This paper provides a review of the compact intense electron-beam accelerators (IEBAs) based on liquid pulse forming lines (PFLs) that havebeen developed at the National University of Defense Technology (NUDT) in China. The history and roadmap of the compact IEBAs used todrive high-power microwave (HPM) devices at NUDT are reviewed. The properties of both de-ionized water and glycerin as energy storagemedia are presented. Research into the breakdown properties of liquid dielectrics and the desire to maximize energy storage have resulted in theinvention of several coaxial PFLs with different electromagnetic structures, which are detailed in this paper. These high energy density liquidPFLs have been used to increase the performance of IEBA subsystems, based on which the SPARK (Single Pulse Accelerator with spark gaps)and HEART (High Energy-density Accelerator with Repetitive Transformer) series of IEBAs were constructed. This paper also discusses howthese compact IEBAs have been used to drive typical HPM devices and concludes by summarizing the associated achievements and theconclusions that can be drawn from the results.展开更多
In order to lower the imidization temperature of polyamic acids（PAA）, the catalytic activities of the curing agents p-hydroxybenzoic acid（PHA）, quinoline（QL）, benzimidazole（BI）, benzotriazole（BTA）, triethyla...In order to lower the imidization temperature of polyamic acids（PAA）, the catalytic activities of the curing agents p-hydroxybenzoic acid（PHA）, quinoline（QL）, benzimidazole（BI）, benzotriazole（BTA）, triethylamine（Et_3N） and 1, 8-diazabicyclo [5.4.0]undec-7-ene（DBU） were investigated in the process of thermal imidization of PAA. In addition, the effect of these various curing agents on the thermal stabilities and mechanical properties of the resultant polyimide（PI） films was determined. Quinoline was found to be an effective curing accelerator in the use of two-step method for synthesizing PI. Due to its moderate base strength, low steric crowding effect and moderate boiling point, quinoline could not only accelerate PAA to achieve imidization completely at 180 ℃, but also maintain the mechanical properties and thermal stability of the ordinary PI film. Any residual quinoline could be removed from PI films by heating at 250 ℃ for 4 h.展开更多
Laser wakefield accelerators（LWFAs） are compact accelerators which can produce femtosecond high-energy electron beams on a much smaller scale than the conventional radiofrequency accelerators. It is attributed to th...Laser wakefield accelerators（LWFAs） are compact accelerators which can produce femtosecond high-energy electron beams on a much smaller scale than the conventional radiofrequency accelerators. It is attributed to their high acceleration gradient which is about 3 orders of magnitude larger than the traditional ones. The past decade has witnessed the major breakthroughs and progress in developing the laser wakfield accelerators. To achieve the LWFAs suitable for applications,more and more attention has been paid to optimize the LWFAs for high-quality electron beams. A single-staged LWFA does not favor generating controllable electron beams beyond 1 Ge V since electron injection and acceleration are coupled and cannot be independently controlled. Staged LWFAs provide a promising route to overcome this disadvantage by decoupling injection from acceleration and thus the electron-beam quality as well as the stability can be greatly improved.This paper provides an overview of the physical conceptions of the LWFA, as well as the major breakthroughs and progress in developing LWFAs from single-stage to two-stage LWFAs.展开更多
Objective:The aim of this study was to measure the leakage by two methods with ion chamber and ready packs film,and to investigate the feasibility and the advantages of using two dosimetry methods for assessing leakag...Objective:The aim of this study was to measure the leakage by two methods with ion chamber and ready packs film,and to investigate the feasibility and the advantages of using two dosimetry methods for assessing leakage radiation around the head of the linear accelerators.Methods:Measurements were performed using a 30 cm3ion chamber;the gantry at 0°,the X-ray head at 0°,the field size at between the central axis and a plane surface at a FSD of 100 as a reference,a series of concentric circles having radii of 50,75,and 100 cm with their common centre at the reference point.The absorbed dose was measured at the reference point,and this would be used as the reference dose.With the diaphragm closed,the measurements were taken along the circumference of the three circles and at 45°intervals.Results:Leakage radiations while the treatment head was in the vertical position varied between 0.016%–0.04%.With the head lying horizontally,leakage radiation was the same order magnitude and varied between 0.02%–0.07%.In the second method,the verification was accomplished by closing the collimator jaws and covering the head of the treatment unit with the ready pack films.The films were marked to permit the determination of their positions on the machine after exposed and processed.With the diaphragm closed,and the ready packs films around the linear accelerator the beam turned on for 2500 cGy(2500 MU).The optical density of these films was measured and compared with this of the reference dose.Leakage radiation varied according to the film positions and the magnitude of leakage was between 0.005%–0.075%.Conclusion:The differences between the values of the leakage radiation levels observed at different measurement points do not only reflect differences in the effective shielding thickness of the head wall,but are also related to differences in the distances between the target and the measurement points.The experimental errors involved in dosimetric measurement also contribute to such differences.展开更多
Deep learning accelerators(DLAs)have been proved to be efficient computational devices for processing deep learning algorithms.Various DLA architectures are proposed and applied to different applications and tasks.How...Deep learning accelerators(DLAs)have been proved to be efficient computational devices for processing deep learning algorithms.Various DLA architectures are proposed and applied to different applications and tasks.However,for most DLAs,their programming interfaces are either difficult to use or not efficient enough.Most DLAs require programmers to directly write instructions,which is time-consuming and error-prone.Another prevailing programming interface for DLAs is high-performance libraries and deep learning frameworks,which are easy to be used and very friendly to users,but their high abstraction level limits their control capacity over the hardware resources thus compromises the efficiency of the accelerator.A design of the programming interface is for DLAs.First various existing DLAs and their programming methods are analyzed and a methodology for designing programming interface for DLAs is proposed,which is a high-level assembly language(called DLA-AL),assembler and runtime for DLAs.DLA-AL is composed of a low-level assembly language and a set of high-level blocks.It allows experienced experts to fully exploit the potential of DLAs and achieve near-optimal performance.Meanwhile,by using DLA-AL,end-users who have little knowledge of the hardware are able to develop deep learning algorithms on DLAs spending minimal programming efforts.展开更多
We present an analysis of electromagnetic oscillations in a spherical conducting cavity filled concentrically with either dielectric or vacuum layers. The fields are given analytically, and the resonant frequency is d...We present an analysis of electromagnetic oscillations in a spherical conducting cavity filled concentrically with either dielectric or vacuum layers. The fields are given analytically, and the resonant frequency is determined numerically. An important special case of a spherical conducting cavity with a smaller dielectric sphere at its center is treated in more detail. By numerically integrating the equations of motion we demonstrate that the transverse electric oscillations in such cavity can be used to accelerate strongly relativistic electrons. The electron’s trajectory is assumed to be nearly tangential to the dielectric sphere. We demonstrate that the interaction of such electrons with the oscillating magnetic field deflects their trajectory from a straight line only slightly. The Q factor of such a resonator only depends on losses in the dielectric. For existing ultra low loss dielectrics, Q can be three orders of magnitude better than obtained in existing cylindrical cavities.展开更多
This study contains comparative research of sulfate resistance of ordinary Portland cement pastes with addition of basic aluminum sulfate (hydroxosulfate, Al(OH)1.78(SO4)0.61) and amorphous Al(OH)3. Over 3 months of s...This study contains comparative research of sulfate resistance of ordinary Portland cement pastes with addition of basic aluminum sulfate (hydroxosulfate, Al(OH)1.78(SO4)0.61) and amorphous Al(OH)3. Over 3 months of storing in sodium sulfate solution, the most significant expansion and deterioration occur in case of samples with aluminum hydroxide. During sulfate resistance test, the ratio between aluminum nuclei in AFt (ettringite) and AFm phases was studied by 27Al-MAS NMR, and the impact of aluminum-containing admixtures on this ratio was specified. In accordance with NMR data, in samples with Al(OH)3, the rate of secondary ettringite formation becomes noticeable after one month of storing in sulfate solution. For samples with Al(OH)1.78(SO4)0.61, the rate of elongation and temper of changes in molar ratio between sulfoaluminate phases were comparable with reference samples without admixtures.展开更多
Dielectric laser accelerators(DLAs)are considered promising candidates for on-chip particle accelerators that can achieve high acceleration gradients.This study explores various combinations of dielectric materials an...Dielectric laser accelerators(DLAs)are considered promising candidates for on-chip particle accelerators that can achieve high acceleration gradients.This study explores various combinations of dielectric materials and accelerated structures based on the inverse Cherenkov effect.The designs utilize conventional processing methods and laser parameters currently in use.We optimize the structural model to enhance the gradient of acceleration and the electron energy gain.To achieve higher acceleration gradients and energy gains,the selection of materials and structures should be based on the initial electron energy.Furthermore,we observed that the variation of the acceleration gradient of the material is different at different initial electron energies.These findings suggest that on-chip accelerators are feasible with the help of these structures and materials.展开更多
Laser wakefield accelerators （LWFAs） are considered to be one of the most compeuuve next- generation accelerator candidates. In this paper, we will study the potential high-flux electron beam production of an LWFA d...Laser wakefield accelerators （LWFAs） are considered to be one of the most compeuuve next- generation accelerator candidates. In this paper, we will study the potential high-flux electron beam production of an LWFA driven by petawatt-level laser pulses. In our three-dimensional particle-in-cell simulations, an optimal set of parameters gives -40 nC of charge with 2 PW laser power, thus -400 kA of instantaneous current if we assume the electron beam duration is 100 fs. This high flux and its secondary radiation are widely applicable in nuclear and QED physics, industrial imaging, medical and biological studies.展开更多
Deep learning has now been widely used in intelligent apps of mobile devices.In pursuit of ultra-low power and latency,integrating neural network accelerators(NNA)to mobile phones has become a trend.However,convention...Deep learning has now been widely used in intelligent apps of mobile devices.In pursuit of ultra-low power and latency,integrating neural network accelerators(NNA)to mobile phones has become a trend.However,conventional deep learning programming frameworks are not well-developed to support such devices,leading to low computing efficiency and high memory-occupation.To address this problem,a 2-stage pipeline is proposed for optimizing deep learning model inference on mobile devices with NNAs in terms of both speed and memory-footprint.The 1 st stage reduces computation workload via graph optimization,including splitting and merging nodes.The 2 nd stage goes further by optimizing at compilation level,including kernel fusion and in-advance compilation.The proposed optimizations on a commercial mobile phone with an NNA is evaluated.The experimental results show that the proposed approaches achieve 2.8×to 26×speed up,and reduce the memory-footprint by up to 75%.展开更多
基金This work was supported by Open Fund Project of State Key Laboratory of Intelligent Vehicle Safety Technology by Grant with No.IVSTSKL-202311Key Projects of Science and Technology Research Programme of Chongqing Municipal Education Commission by Grant with No.KJZD-K202301505+1 种基金Cooperation Project between Chongqing Municipal Undergraduate Universities and Institutes Affiliated to the Chinese Academy of Sciences in 2021 by Grant with No.HZ2021015Chongqing Graduate Student Research Innovation Program by Grant with No.CYS240801.
文摘Massive computational complexity and memory requirement of artificial intelligence models impede their deploy-ability on edge computing devices of the Internet of Things(IoT).While Power-of-Two(PoT)quantization is pro-posed to improve the efficiency for edge inference of Deep Neural Networks(DNNs),existing PoT schemes require a huge amount of bit-wise manipulation and have large memory overhead,and their efficiency is bounded by the bottleneck of computation latency and memory footprint.To tackle this challenge,we present an efficient inference approach on the basis of PoT quantization and model compression.An integer-only scalar PoT quantization(IOS-PoT)is designed jointly with a distribution loss regularizer,wherein the regularizer minimizes quantization errors and training disturbances.Additionally,two-stage model compression is developed to effectively reduce memory requirement,and alleviate bandwidth usage in communications of networked heterogenous learning systems.The product look-up table(P-LUT)inference scheme is leveraged to replace bit-shifting with only indexing and addition operations for achieving low-latency computation and implementing efficient edge accelerators.Finally,comprehensive experiments on Residual Networks(ResNets)and efficient architectures with Canadian Institute for Advanced Research(CIFAR),ImageNet,and Real-world Affective Faces Database(RAF-DB)datasets,indicate that our approach achieves 2×∼10×improvement in the reduction of both weight size and computation cost in comparison to state-of-the-art methods.A P-LUT accelerator prototype is implemented on the Xilinx KV260 Field Programmable Gate Array(FPGA)platform for accelerating convolution operations,with performance results showing that P-LUT reduces memory footprint by 1.45×,achieves more than 3×power efficiency and 2×resource efficiency,compared to the conventional bit-shifting scheme.
文摘Molecular Dynamics(MD)simulation for computing Interatomic Potential(IAP)is a very important High-Performance Computing(HPC)application.MD simulation on particles of experimental relevance takes huge computation time,despite using an expensive high-end server.Heterogeneous computing,a combination of the Field Programmable Gate Array(FPGA)and a computer,is proposed as a solution to compute MD simulation efficiently.In such heterogeneous computation,communication between FPGA and Computer is necessary.One such MD simulation,explained in the paper,is the(Artificial Neural Network)ANN-based IAP computation of gold(Au_(147)&Au_(309))nanoparticles.MD simulation calculates the forces between atoms and the total energy of the chemical system.This work proposes the novel design and implementation of an ANN IAP-based MD simulation for Au_(147)&Au_(309) using communication protocols,such as Universal Asynchronous Receiver-Transmitter(UART)and Ethernet,for communication between the FPGA and the host computer.To improve the latency of MD simulation through heterogeneous computing,Universal Asynchronous Receiver-Transmitter(UART)and Ethernet communication protocols were explored to conduct MD simulation of 50,000 cycles.In this study,computation times of 17.54 and 18.70 h were achieved with UART and Ethernet,respectively,compared to the conventional server time of 29 h for Au_(147) nanoparticles.The results pave the way for the development of a Lab-on-a-chip application.
文摘The flexibility in radiotherapy can be improved if patients can be moved between any one of the department’s medical linear accelerators (LINACs) without the need to change anything in the patient’s treatment plan. For this to be possible, the dosimetric characteristics of the various accelerators must be the same, or nearly the same. The purpose of this work is to describe further and compare measurements and parameters after the initial vendor-recommended beam matching of the five LINACs. Deviations related to dose calculations and to beam matched accelerators may compromise treatment accuracy. The safest and most practical way to ensure that all accelerators are within clinical acceptable accuracy is to include TPS calculations in the LINACs matching evaluation. Treatment planning system (TPS) was used to create three photons plans with different field sizes 3 × 3 cm, 10 × 10 cm and 25 × 25 cm at a depth of 4.5 cm in Perspex. Calculated TPS plans were sent to Mosaiq to be delivered by five LINACs. TPS plans were compared with five LINACs measurements data using Gamma analyses of 2% and 2 mm. The results suggest that for four out of the five LINACs, there was generally good agreement, less than a 2% deviation between the planned dose distribution and the measured dose distribution. However, one specific LINAC named “Asterix” exhibited a deviation of 2.121% from the planned dose. The results show that all of the LINACs’ performance were within the acceptable deviation and delivering radiation dose consistently and accurately.
文摘Deep neural networks(DNN)are widely used in image recognition,image classification,and other fields.However,as the model size increases,the DNN hardware accelerators face the challenge of higher area overhead and energy consumption.In recent years,stochastic computing(SC)has been considered a way to realize deep neural networks and reduce hardware consumption.A probabilistic compensation algorithm is proposed to solve the accuracy problem of stochastic calculation,and a fully parallel neural network accelerator based on a deterministic method is designed.The software simulation results show that the accuracy of the probability compensation algorithm on the CIFAR-10 data set is 95.32%,which is 14.98%higher than that of the traditional SC algorithm.The accuracy of the deterministic algorithm on the CIFAR-10 dataset is 95.06%,which is 14.72%higher than that of the traditional SC algorithm.The results of Very Large Scale Integration Circuit(VLSI)hardware tests show that the normalized energy efficiency of the fully parallel neural network accelerator based on the deterministic method is improved by 31%compared with the circuit based on binary computing.
基金financially supported by the National Natural Science Foundation of China(Nos.51003031 and 51303026)Science Foundation for Universities and Institutions of Dongguan City(No.2012108102008)the Research Fund for the Doctoral Program of Dongguan University of Technology(No.ZJ121002)
文摘A novel TU derivative, N-phenyl-N＇-（γ-triethoxysilane）-propyl thiourea （STU）, is prepared and its binary accelerator system is investigated in detail. Compared to the control references, the optimum curing time of NR compounds with STU is the shortest, indicating a more nucleophilic reaction occurs. The Py-GC/MS results present that the phenyl iso- thiocyanate fragment still remains in the NPUSTU compounds with or without extracting treatment, but no silane segment can be found in the vulcanizate with extracting treatment. Vibrations of C=S, NH and aromatic ring in FTIR experiments and a new methyne carbon peak, as well as the peaks of phenyl group of STU, in the solid state 13C-NMR experiments are found in the NR/STU vulcanizate with extracting treatment. Moreover, the crosslinking density of vulcanizates with STU evolves to lower level, indicating the sulfur atom of STU does not contribute to the sulfur crosslinking. Therefore, a new vulcanization kinetic mechanism of STU is propounded that the thiourea groups can graft to the rubber main chains as pendant groups by chemical bonds during the vulcanization process, which is in accordance with the experimental observations quite well.
基金the National Natural Science Foundation of China(Nos.11375122 and 11875197).
文摘Superconducting linear accelerators(SCL)have a high acceleration gradient and are capable of operating in a high-duty factor mode.For high-power and high-intensity SCL,the design of beam dynamics generally follows the principle that the zero-current periodic phase advance(σ0)of each degree of freedom is less than 90°to avoid envelope instability caused by space charge.However,this principle is obtained under the condition of a completely periodic focusing channel,and it is ambiguous for pseudoperiodic structures,such as linear accelerators.Although transverse beam dynamics without acceleration have been studied by other researchers,it appears that there are some connections between pure 2D and 3D beam dynamics.Based on these two points,five focusing schemes for the solenoid and quadrupole doublet channels were designed to simulate the beam behavior with non-constantσ0.Among them,the four schemes follow the characteristics of variation in the zero-current longitudinal phase advance(σ0l)under a constant acceleration gradient and synchronous phase.The zero-current transverse phase advance(σ0t)is consistent withσ0l,based on the equipartition requirement.The initialσ0twas set to 120°,110°,100°,and 90°,and was then gradually decreased to approximately 40°at the end of the channel.The last scheme maintains the maximumσ0tof88°by reducing the acceleration gradient of the corresponding cavities,until the point at whichσ0tequals88°with a normal gradient.Using the stopbands obtained from the linearized envelope equations and multiparticle particle-in-cell(PIC)simulations,the transport properties of both continuous and 3D-bunched beams with the acceleration of the five focusing schemes were studied.It was found that for a CW beam,when tune depression>0.7,σ0tcan break through 90°when the beams were transported in both solenoid and quadrupole doublet periodic focusing channels.When tune depression<0.7,the conclusions were different.For the solenoid focusing system,σ0tcan partially break through 90°,and the beam quality is not significantly affected.For the quadrupole doublet focusing system,a partial breakthrough of 90°has a greater impact on the beam quality.The same conclusions were obtained for a bunched beam with acceleration.
文摘The study revealed that ordinary acceler- ators for portlandcement also improves early strength of Belite cement remarkably, butone serious defect exists Which lead to a high strength reduction atthe age of 28 days. The paper analyzes the action mechanism ofordinary accelerators on Belite cement. It shows that theseaccelerators can not accelerate the hydration of β-C_2S directly.
文摘In this paper, the origin and type of radiation hazards as well as the main aspects of radiation protection for low-energy accelerators are discussed in general, and the problems of radiation protection and the experimental results of the operational monitoring of the five accelerators in the institute of Nuclear Science and Technology,Sichuan University, namely, one 1.2 M cyclofron, two Cockroft-waltons and two Van de Graafts, as well as a powerful electron accelerator for industrial irradiation are described. The discussion and evaluation are made according to the requirments of the National standards GB 5172-85.
文摘In 2014 some research works have been carried out by the Electron Accelerator Group at IMP, including thehigh energy electron radiography (HEER), the electron accelerator design with short pulse for HEER, and a newtype low energy electrostatic electron accelerator for industrial applications.The verification experiments of HEER have been accomplished on THU electron accelerator in 2013, so moreexperimental works were proposed to continue with complicated parameters. This year, a specialized beam linefor HEER experiment was designed and built, which consists of two dipole magnets, ten quadrapole magnets andsome vacuum chambers with diagnostic probes.
文摘To understand the nature of"matter",scientists have to answer some very basic questions.For example,why there are so many"elementary particles"?Are they really indivisible building blocks of our cosmos?Is the cosmos following some unknown physical laws,which potentially subvert our current understanding of the whole world?How should we understand the"dark matter"in the cosmos?Could the four fundamental forces comply with any common physical law in the same framework?
基金This work was supported by the National Natural Science Foundation of China under Grant No.51677190the Hunan Provincial Natural Science Foundation of China under Grant No.2017JJ1005.
文摘This paper provides a review of the compact intense electron-beam accelerators (IEBAs) based on liquid pulse forming lines (PFLs) that havebeen developed at the National University of Defense Technology (NUDT) in China. The history and roadmap of the compact IEBAs used todrive high-power microwave (HPM) devices at NUDT are reviewed. The properties of both de-ionized water and glycerin as energy storagemedia are presented. Research into the breakdown properties of liquid dielectrics and the desire to maximize energy storage have resulted in theinvention of several coaxial PFLs with different electromagnetic structures, which are detailed in this paper. These high energy density liquidPFLs have been used to increase the performance of IEBA subsystems, based on which the SPARK (Single Pulse Accelerator with spark gaps)and HEART (High Energy-density Accelerator with Repetitive Transformer) series of IEBAs were constructed. This paper also discusses howthese compact IEBAs have been used to drive typical HPM devices and concludes by summarizing the associated achievements and theconclusions that can be drawn from the results.
文摘In order to lower the imidization temperature of polyamic acids（PAA）, the catalytic activities of the curing agents p-hydroxybenzoic acid（PHA）, quinoline（QL）, benzimidazole（BI）, benzotriazole（BTA）, triethylamine（Et_3N） and 1, 8-diazabicyclo [5.4.0]undec-7-ene（DBU） were investigated in the process of thermal imidization of PAA. In addition, the effect of these various curing agents on the thermal stabilities and mechanical properties of the resultant polyimide（PI） films was determined. Quinoline was found to be an effective curing accelerator in the use of two-step method for synthesizing PI. Due to its moderate base strength, low steric crowding effect and moderate boiling point, quinoline could not only accelerate PAA to achieve imidization completely at 180 ℃, but also maintain the mechanical properties and thermal stability of the ordinary PI film. Any residual quinoline could be removed from PI films by heating at 250 ℃ for 4 h.
基金Project supported by the National Natural Science Foundation of China(Grant Nos.11127901,11425418,and 61221064)the National Basic Research Program of China(Grant No.2011CB808100)the Science and Technology Talent Project of Shanghai City,China(Grant Nos.12XD1405200 and 12ZR1451700)
文摘Laser wakefield accelerators（LWFAs） are compact accelerators which can produce femtosecond high-energy electron beams on a much smaller scale than the conventional radiofrequency accelerators. It is attributed to their high acceleration gradient which is about 3 orders of magnitude larger than the traditional ones. The past decade has witnessed the major breakthroughs and progress in developing the laser wakfield accelerators. To achieve the LWFAs suitable for applications,more and more attention has been paid to optimize the LWFAs for high-quality electron beams. A single-staged LWFA does not favor generating controllable electron beams beyond 1 Ge V since electron injection and acceleration are coupled and cannot be independently controlled. Staged LWFAs provide a promising route to overcome this disadvantage by decoupling injection from acceleration and thus the electron-beam quality as well as the stability can be greatly improved.This paper provides an overview of the physical conceptions of the LWFA, as well as the major breakthroughs and progress in developing LWFAs from single-stage to two-stage LWFAs.
文摘Objective:The aim of this study was to measure the leakage by two methods with ion chamber and ready packs film,and to investigate the feasibility and the advantages of using two dosimetry methods for assessing leakage radiation around the head of the linear accelerators.Methods:Measurements were performed using a 30 cm3ion chamber;the gantry at 0°,the X-ray head at 0°,the field size at between the central axis and a plane surface at a FSD of 100 as a reference,a series of concentric circles having radii of 50,75,and 100 cm with their common centre at the reference point.The absorbed dose was measured at the reference point,and this would be used as the reference dose.With the diaphragm closed,the measurements were taken along the circumference of the three circles and at 45°intervals.Results:Leakage radiations while the treatment head was in the vertical position varied between 0.016%–0.04%.With the head lying horizontally,leakage radiation was the same order magnitude and varied between 0.02%–0.07%.In the second method,the verification was accomplished by closing the collimator jaws and covering the head of the treatment unit with the ready pack films.The films were marked to permit the determination of their positions on the machine after exposed and processed.With the diaphragm closed,and the ready packs films around the linear accelerator the beam turned on for 2500 cGy(2500 MU).The optical density of these films was measured and compared with this of the reference dose.Leakage radiation varied according to the film positions and the magnitude of leakage was between 0.005%–0.075%.Conclusion:The differences between the values of the leakage radiation levels observed at different measurement points do not only reflect differences in the effective shielding thickness of the head wall,but are also related to differences in the distances between the target and the measurement points.The experimental errors involved in dosimetric measurement also contribute to such differences.
基金Supported by the National Key Research and Development Program of China(No.2017YFA0700902,2017YFB1003101)the 973 Program of China(No.2015CB358800)National Science and Technology Major Project(No.2018ZX01031102)
文摘Deep learning accelerators(DLAs)have been proved to be efficient computational devices for processing deep learning algorithms.Various DLA architectures are proposed and applied to different applications and tasks.However,for most DLAs,their programming interfaces are either difficult to use or not efficient enough.Most DLAs require programmers to directly write instructions,which is time-consuming and error-prone.Another prevailing programming interface for DLAs is high-performance libraries and deep learning frameworks,which are easy to be used and very friendly to users,but their high abstraction level limits their control capacity over the hardware resources thus compromises the efficiency of the accelerator.A design of the programming interface is for DLAs.First various existing DLAs and their programming methods are analyzed and a methodology for designing programming interface for DLAs is proposed,which is a high-level assembly language(called DLA-AL),assembler and runtime for DLAs.DLA-AL is composed of a low-level assembly language and a set of high-level blocks.It allows experienced experts to fully exploit the potential of DLAs and achieve near-optimal performance.Meanwhile,by using DLA-AL,end-users who have little knowledge of the hardware are able to develop deep learning algorithms on DLAs spending minimal programming efforts.
文摘We present an analysis of electromagnetic oscillations in a spherical conducting cavity filled concentrically with either dielectric or vacuum layers. The fields are given analytically, and the resonant frequency is determined numerically. An important special case of a spherical conducting cavity with a smaller dielectric sphere at its center is treated in more detail. By numerically integrating the equations of motion we demonstrate that the transverse electric oscillations in such cavity can be used to accelerate strongly relativistic electrons. The electron’s trajectory is assumed to be nearly tangential to the dielectric sphere. We demonstrate that the interaction of such electrons with the oscillating magnetic field deflects their trajectory from a straight line only slightly. The Q factor of such a resonator only depends on losses in the dielectric. For existing ultra low loss dielectrics, Q can be three orders of magnitude better than obtained in existing cylindrical cavities.
文摘This study contains comparative research of sulfate resistance of ordinary Portland cement pastes with addition of basic aluminum sulfate (hydroxosulfate, Al(OH)1.78(SO4)0.61) and amorphous Al(OH)3. Over 3 months of storing in sodium sulfate solution, the most significant expansion and deterioration occur in case of samples with aluminum hydroxide. During sulfate resistance test, the ratio between aluminum nuclei in AFt (ettringite) and AFm phases was studied by 27Al-MAS NMR, and the impact of aluminum-containing admixtures on this ratio was specified. In accordance with NMR data, in samples with Al(OH)3, the rate of secondary ettringite formation becomes noticeable after one month of storing in sulfate solution. For samples with Al(OH)1.78(SO4)0.61, the rate of elongation and temper of changes in molar ratio between sulfoaluminate phases were comparable with reference samples without admixtures.
基金the National Natural Science Foundation of China(Grant No.11975214)。
文摘Dielectric laser accelerators(DLAs)are considered promising candidates for on-chip particle accelerators that can achieve high acceleration gradients.This study explores various combinations of dielectric materials and accelerated structures based on the inverse Cherenkov effect.The designs utilize conventional processing methods and laser parameters currently in use.We optimize the structural model to enhance the gradient of acceleration and the electron energy gain.To achieve higher acceleration gradients and energy gains,the selection of materials and structures should be based on the initial electron energy.Furthermore,we observed that the variation of the acceleration gradient of the material is different at different initial electron energies.These findings suggest that on-chip accelerators are feasible with the help of these structures and materials.
基金supported by Extreme Light Infrastructure- Nuclear Physics (ELI-NP) Phase Ⅱa project co-financed by the Romanian Government and European Union through the European Regional Development FundThe EPOCH code project was funded by the UK EPSRC grants EP/G054950/1, EP/ G056803/1, EP/G055165/1 and EP/ M022463/1
文摘Laser wakefield accelerators （LWFAs） are considered to be one of the most compeuuve next- generation accelerator candidates. In this paper, we will study the potential high-flux electron beam production of an LWFA driven by petawatt-level laser pulses. In our three-dimensional particle-in-cell simulations, an optimal set of parameters gives -40 nC of charge with 2 PW laser power, thus -400 kA of instantaneous current if we assume the electron beam duration is 100 fs. This high flux and its secondary radiation are widely applicable in nuclear and QED physics, industrial imaging, medical and biological studies.
基金Supported by the National Key Research and Development Program of China(No.2017YFB1003101,2018AAA0103300,2017YFA0700900)the National Natural Science Foundation of China(No.61702478,61732007,61906179)+2 种基金the Beijing Natural Science Foundation(No.JQ18013)the National Science and Technology Major Project(No.2018ZX01031102)the Beijing Academy of Artificial Intelligence
文摘Deep learning has now been widely used in intelligent apps of mobile devices.In pursuit of ultra-low power and latency,integrating neural network accelerators(NNA)to mobile phones has become a trend.However,conventional deep learning programming frameworks are not well-developed to support such devices,leading to low computing efficiency and high memory-occupation.To address this problem,a 2-stage pipeline is proposed for optimizing deep learning model inference on mobile devices with NNAs in terms of both speed and memory-footprint.The 1 st stage reduces computation workload via graph optimization,including splitting and merging nodes.The 2 nd stage goes further by optimizing at compilation level,including kernel fusion and in-advance compilation.The proposed optimizations on a commercial mobile phone with an NNA is evaluated.The experimental results show that the proposed approaches achieve 2.8×to 26×speed up,and reduce the memory-footprint by up to 75%.
