Isotopic temperature in spallation reactions

The possibility of applying the double-isotope-ratio method for calculating temperature to spallation products formed by high-energy protons is discussed. It is concluded that, at a specific incident-particle energy, spallation is characterized by a single temperature of the formation of products for targets in the mass range 56 ≤ A ≤ 208. No energy dependence of the temperature that was calculated by the double-isotope-ratio method was observed in the energy range 0.3 ≤ E ≤ 300 GeV.     

A Bayesian-neural-network prediction for fragment production in proton induced spallation reaction

Fragment production in spallation reactions yields key infrastructure data for various applications. Based on the empirical SPACS parameterizations, a Bayesian-neural-network (BNN) approach is established to predict the fragment cross sections in proton-induced spallation reactions. A systematic investigation has been performed for the measured proton-induced spallation reactions of systems ranging from intermediate to heavy nuclei systems and incident energies ranging from 168 MeV/u to 1500 MeV/u. By learning the residuals between the experimental measurements and SPACS predictions, it is found that the BNN-predicted results are in good agreement with the measured results. The established method is suggested to benefit the related research on nuclear astrophysics, nuclear radioactive beam sources, accelerator driven systems, proton therapy, etc.     

Cross sections of spallation residues produced in 1A GeV 208Pb on proton reactions

Spallation residues produced in 1 GeV per nucleon 208Pb on proton reactions have been studied using the Fragment Separator facility at GSI. Isotopic production cross sections of elements from 61Pm to 82Pb have been measured down to 0.1 mb with a high accuracy. The recoil kinetic energies of the produced fragments were also determined. The obtained cross sections agree with most of the few existing gamma-spectroscopic data. The data are compared with different intranuclear-cascade and evaporation-fission models. Drastic deviations were found for a standard code used in technical applications.     

基于神经网络的图像识别方法研究

图像识别涉及大量的信息运算,要求处理速度快、识别精度高,神经网络的实时性和容错性要符合图像识别的要求。利用改进的BP神经网络算法对旋转畸变图像进行了定位和识别,改进算法将附加动量项与自适应学习速率相结合,有效地抑制了网络陷入局部极小点,提高了网络的训练速度。结果表明,基于神经网络的图像识别方法是有效的、可行的。     

基于增强GA-BP神经网络的软件错误定位方法

在软件开发和后期维护的过程中,进行软件调试来定位错误并修正错误是其中最复杂且成本最大的一部分;文章针对现有基于神经网络的软件错误定位方法中的权值和阈值设定不方便、鲁棒性差等问题,结合正交实验设计思想和遗传算法(Genetic Algorithm),提出了一种基于增强遗传BP神经网络的软件错误定位方法;并将其同基于GA-BP神经网络的和基于BP神经网络的定位方法都在MATLAB上进行了实验,实验数据来源西门子测试集,从结果上看,基于增强GA-BP神经网络的软件错误定位方法在定位错误的效率和精确度上都有一些进步。     

贝叶斯优化卷积神经网络公共场所异常声识别

针对公共场所异常声的感知和识别问题,提出一种基于贝叶斯优化卷积神经网络的识别方法。提取声信号的Gammatone倒谱系数、倍频程功率谱、短时能量和谱质心,组合成声信号的特征图。构建卷积神经网络作为分类器,利用递增的卷积核设置和池化操作处理不同尺度的特征。基于贝叶斯优化算法优化卷积神经网络的模型参数,对包括火苗噼啪声、婴儿啼哭声、烟花燃放声、玻璃破碎声和警报声的5种公共场所异常声进行识别。该方法的识别结果与基于不同的特征提取和分类器方案得到的识别结果进行比较,结果表明该方法的识别效果优于其他特征提取和分类器方案的识别效果。最后分析了该方法在不同信噪比噪声干扰下的识别结果,验证了该方法的有效性。     

Micro-punching process based on spallation delamination induced by laser driven-flyer

In this article, we proposed a micro-punching process for microstructure on films based on laser driven-flyer induced spallation delamination phenomenon at the interface between a film and its substrate. To validate such a micro-punching process, a series of experiments were carried out for fabrication of microstructures on Au films coated on K9 glass substrates and polyimide substrate. Results show that through such a punching process, the microstructure on Au films can be fabricated efficiently and the spatial resolution is able to reach micron level. Moreover, we found that this method was more suitable for films coated on soft substrates rather than that coated on brittle substrates due to the additional destruction of the brittle substrate. This micro-punching process has a wide range of potential application in microfluidic devices, biodevices and other MEMS devices.     

利用几何特性及神经网络进行人脸探测技术的研究

在人脸识别过程中 ,首先也是最重要的一个环节是人脸探测 ,因为一旦从图像中定位并提取到了人脸 ,那么下一步的人脸识别工作就变得非常容易。眼睛是人脸图像中最容易探测的部位 ,而且通过探测双眼来发现人脸最符合人的视觉习惯。提出了一种基于几何特征分析和人工神经网络的由粗到细的两级人脸探测方法。在第一级中 ,眼睛和脸是通过测量眼睛的尺寸和眼睛与脸的位置关系探测到的 ,第一级的输出是一个尺寸归一化的人脸 ,但偶尔也伴随着一个或多个因对复杂背景中与眼睛类似的物体的误判而得到的非人脸图像 ;第二级神经网络正是用来过滤掉第一级中被误判的人脸。实验表明 ,这种由粗到细的两级人脸探测系统具有很高的稳定性和探测正确率     

基于神经网络的电子器件分段建模方法

针对电路仿真领域中对具有复杂非线性特性的电子器件不能采用单-神经网络建模问题,提出了一种分段建模方法,降低了模型的复杂度,并通过对稳压二极管的建模验证了该方法的有效性.     

基于模糊神经网络的短时交通流预测方法研究

为满足交通控制和诱导系统的实时性需求,减少交通拥挤状况,降低交通事故突发频率,需要对短时交通流进行预测。当前的短时交通流预测方法是采用K-近邻的非参数回归对其进行预测,预测过程中没有将预测模型中关键因素对交通流的影响进行详细的说明,导致预测结果不准确,存在短时交通流预测误差较大的问题。为此,提出一种基于模糊神经网络的短时交通流预测方法。该方法首先以历史短时交通流数据样本序列为基础,将提取的关联维数作为短时交通流的混沌特征量,然后以该特征量为依据,对短时交通流数据进行聚类,使相同的短时交通流聚合类样本比不同的交通流聚合类样本更为贴近,采用高斯过程回归对短时交通流预测模型进行建设,建设过程中利用差分方法对短时交通流预测序列进行平稳化操作之后,对短时交通流预测模型进行训练,将GPR模型引入至短时交通流预测过程中,得到交通流预测方差估计值,并确定交通流预测值置信区间,由此实现短时交通流的预测。由此实现短时交通流的预测。实验结果证明,所提方法可以准确地预测交通运输系统的实时状况,为车辆行驶的最佳路线进行了有效引导,减少了自然影响方面和人为因素对短时交通流预测结果的干扰,为交通部门对交通路况的控制管理提供了依据。     

基于轻量级卷积门控循环神经网络的语音增强方法*

针对在基于深度学习语音增强的方法中因采用因果式的网络输入导致语音增强性能下降的问题,提出了一种基于轻量级卷积门控循环神经网络(LCGRU)的语音增强方法。门控循环神经网络能够建模语音信号的时间相关性,但是其全连接结构破坏了语音信号的时频结构特征,并且参数数量庞大,不利于网络的训练。对此,本文采用卷积核替代门控循环神经网络中的全连接结构,在对语音信号时间相关性建模的同时保留了语音信号的时频结构特征,同时降低了网络的参数数量。为充分利用先前帧的特征信息,该网络单元当前时刻的输入融合了上一时刻的输入与输出。针对网络训练过程中容易产生过拟合的问题,本文采用了线性门控机制来控制信息的传输,这缓解了网络训练过程中的过拟合问题,提高了网络的语音增强性能。实验结果表明,本文所提出的网络结构在增强后的语音感知质量(PESQ),语音短时客观可懂度(STOI),分段信噪比(SSNR)等指标上均优于传统的网络结构。     

基于BP神经网络的闪光照相图像复原

针对闪光照相系统模糊较大、成像信噪比较低的问题,提出了一种基于BP神经网络的闪光照相图像复原方法。该方法利用BP神经网络的泛化能力,用样本图像对网络进行训练,建立退化图像与真实图像之间的非线性映射关系,然后将待复原图像分区,利用训练好的BP神经网络对待复原图像的边界区域进行复原处理。数值试验表明,在系统点扩展函数未知的情况下,该算法能较好再现图像边缘信息,复原出的图像在信噪比和视觉方面都有较大提高。     

基于BP神经网络的闪光照相图像复原

针对闪光照相系统模糊较大、成像信噪比较低的问题,提出了一种基于BP神经网络的闪光照相图像复原方法。该方法利用BP神经网络的泛化能力,用样本图像对网络进行训练,建立退化图像与真实图像之间的非线性映射关系,然后将待复原图像分区,利用训练好的BP神经网络对待复原图像的边界区域进行复原处理。数值试验表明,在系统点扩展函数未知的情况下,该算法能较好再现图像边缘信息,复原出的图像在信噪比和视觉方面都有较大提高。     

基于卷积神经网络与特征融合的天气识别方法

在太阳能热水器及太阳能电池等太阳能发电领域,下雨、下雪、阴天等气候因素将严重影响发电效果,而太阳能随动系统工作也必须消耗能量,所以迅速判断当前的天气状况,并设计自适应的开关随动系统极其重要。当天气状况为阴雨或者雪天时,系统应当关闭从而减少能耗。鉴于传统的天气识别方法效率低、准确度差、计算量大的问题,在公开的天气图像基础上创建了一个具有多种类别的天气分类集,并提供了一种基于卷积神经网络与特征融合的天气图像识别技术。通过采用传统方式获取图像的颜色、纹理、形状3种特征作为整个模型的底层特征,在原本的VGG16(visual geometry group-16)模型基础上进行了改进,从而提取图像的深层特征,最后将底层特征与深层特征融合起来在Softmax上进行输出,总识别率达到94%。     

基于远场信息和卷积神经网络的波前重构方法

探测波前相位信息是实现自适应光学波前补偿的关键,使用卷积神经网络（CNN）代替波前传感器进行波前重构,系统简单易于实现,同时重构过程不依赖迭代运算,快速实时。为准确提取远场中的波前特征,CNN需要事先使用大量样本进行训练。研究中根据4～30阶大气湍流泽尼克像差系数与其远场强度的对应关系,仿真制作样本数据集,训练CNN从输入的一帧远场图像中预测出畸变波前的泽尼克像差系数,重构原始波前。验证结果表明,该方法能快速实时地还原出波前相位信息,重构波前较原始波前具有极高的波面吻合度和较小的残差剩余量,有望实现实际自适应光学系统中的闭环校正。     

Thermonuclear reaction rates for proton induced reactions on 41K and neutron induced reactions on 41Ca

The yield of γ-rays from the reaction ⁴¹K(p, γ)⁴²Ca has been measured as a function of bombarding energy over the range 0.68–2.48 MeV and from the reaction ⁴¹K(p, αγ)³⁸Ar over the range 1.20–2.48 MeV, and the yield of neutrons from the reaction ⁴¹K(p, n)⁴¹Ca has been measured from threshold to a bombarding energy of 2.48 MeV. The energy dependence of the cross sections is compared with statistical-model calculations with global optical-model parameters in all particle channels. The calculations seriously overestimate the cross section for the neutron channel and underestimate those for the other channels. A reduction in the imaginary well depth in the neutron channel leads to good agreement with all the data. Statistical-model calculations with this modified set of parameters are then carried out to provide cross sections for the astrophysically interesting reactions ⁴¹Ca(n, p)⁴¹K, ⁴¹Ca(n, α)³⁸Ar, and ⁴¹Ca(n, γ)⁴²Ca. Thermonuclear reaction rates are calculated for all six reactions over the temperature range 5 × 10⁸–10¹⁰K which includes the range of temperatures of interest in nucleosynthesis calculations.     

Prediction of the cross-sections of isotopes produced in deuteron-induced spallation of long-lived fission products

The spallation cross-section data for the long-lived fission products(LLFPs) are scarce but required for the design of accelerator driven systems. In this paper, the isospin dependent quantum molecular dynamics model and the statistical code GEMINI are applied to simulate deuteron-induced spallation in the energy region of GeV/nucleon. By comparing the calculations with the experimental data, the applicability of the model is verified. The model is then applied to simulate the spallation of 90 Sr, 93 Zr, 107 Pd, and 137 Cs induced by deuterons at 200, 500 and 1000 MeV/nucleon. The cross-sections of isotopes, the cross-sections of long-lived nuclei, and the reaction energy are presented.Using the above observables, the feasibility of LLFP transmutation by spallation is discussed.     

A one-dimension convolutional neural network based interference classification method

Interference is a common problem in wireless communication, navigation and radar systems. A wide variety of interferences are used to degrade the communication quality especially in electronic warfare environment. In modern military communication systems, interference classification is an important module for its ability to obtain prior interference information before adopting related anti-interference method. This paper proposes a deep learning based interference classification method, which applies one-dimension convolutional neural networks to automatically extract interference features for classification. Computer simulations show better classification performance and lower computational complexity. Meanwhile, this proposed method is implied on software defined radios (SDR) hardware, more than 99% correct classification probability can be achieved with limited samples of the received signal, which verifies the robustness of this proposed method.     

基于人工神经网络的混响时间预测方法

在分析混响时间的各种影响因素的基础上,运用人工神经网络的BP算法,建立了厅堂混响时间的预测模型,该方法具有速度快、使用方便、精度高、可考虑的因素多等特点,具有一定的工程实用价值,有待实践的考验并逐步提高完善。