人工神经网络法校正ICP┐AES中重叠光谱干扰

本文将反向传播人工神经网络（ＢＰ－ＡＮＮ）用于ＩＣＰ－ＡＥＳ中重叠光谱干扰的校正。利用模拟的Ｃｅ４１３．３８０ｎｍ和Ｐｒ４１３．３６１ｎｍ光谱对神经网络的训练方式、输入值范围、噪声影响等作了较详细的讨论。     

基于BP算法神经网络的物理实验曲线拟合

介绍了人工神经元网络的基本原理和BP算法神经网络。将BP算法神经网络应用于物理实验中的幅频特性曲线拟合，在MATLAB下通过训练和仿真验证了应用BP算法神经网络拟合实验曲线的优越性。     

粒子群优化算法在混合气体红外光谱定量分析中的应用

通过将粒子群优化技术及BP神经网络技术相结合,建立了三种烃烷混合气体的红外光谱定量分析模型。混合气体主要由甲烷、乙烷、丙烷三种组分气体组成,三种组分气体浓度范围分别为0.01%～0.1%。文章首先采用主成分分析技术从红外光谱1 866个数据中提取了5个特征变量作为神经网络的输入,将气体浓度作为网络输出。然后将粒子群优化算法与BP神经网络技术相结合,对网络的隐含层节点数进行了优化选择。再对结构优化后的网络进行训练,建立气体分析模型。分析模型的标准气体验证实验结果表明,采用此方法建立混合气体红外光谱定量分析模型所用时间(大约4 600 s)比单纯采用BP神经网络进行遍历优化建模所用时间(大约24 500 s)降低5倍以上,模型预测精度水平相当,网络结构大致相同,具有一定的实践意义和应用潜力。     

Identification of Rhubarb Samples by Using NIR Spectrometry and Takagi–Sugeno Fuzzy System

Abstract

The Takagi–Sugeno fuzzy system is implemented based on several back‐propagation neural networks (BP‐NNs) and has been applied to identification of official and unofficial rhubarb samples based on their near‐infrared spectra. Rhubarb is one of the most important Chinese medical herbs. It is of importance to identify official and unofficial rhubarb samples based on nondestructive near‐infrared spectrometry for quality control in Chinese herbal products. Near‐infrared diffuse reflectance spectrometry and the Takagi–Sugeno fuzzy system were used to classify 52 rhubarb samples, and the effects of the number of hidden neurons and of momentum parameters on prediction were investigated. The results obtained by using the Takagi–Sugeno fuzzy system were better than those by commonly used BP networks. With proper network training parameters, 100% correctness can be obtained by using the Takagi–Sugeno fuzzy system.     

Using Short Wave Visible–Near Infrared Reflectance Spectroscopy to Predict Soil Properties and Content

In this study, short wave visible–near infrared reflectance spectroscopy was evaluated for prediction of diverse soil properties related to four different soil series of several regions in Jiangxi, China. A total of 240 soil samples were collected for the calibration (n = 168) and prediction (n = 72) sets. The used wavelength range of short wave visible–near infrared reflectance spectroscopy is 325–1075 nm. Partial least squares regression and back propagation neural network were used to develop models for soil properties such as organic matter and extractable forms of calcium, magnesium, and potassium. Performance of these models was also compared and analyzed. The input of back propagation neural network was the first six principal components resulted from the principal component analysis and the optimal number of latent variables obtained from partial least squares regression. The overall results showed that the performance of partial least squares regression model was inferior to all back propagation neural network models. The best prediction was obtained with latent variables as input of back propagation neural network model for organic matter (determination coefficient = 0.84 and relative predictive determinant = 2.38), which was classified as very good model predictions. The prediction of calcium, magnesium, and potassium was classified as fair (determination coefficient = 0.56–0.68 and relative predictive determinant = 1.51–1.61), where quantitative predictions were considered possible. It is recommended to adopt latent variables as input for back propagation neural network model predicting soil properties with short wave visible–near infrared reflectance spectroscopy. In conclusion, short wave visible–near infrared reflectance spectroscopy was variably successful in estimating soil properties and showed potential for substituting laboratory analyses.     

人工神经网络分类鉴别苦丁茶红外光谱

为了分类鉴别苦丁茶,采用竞争神经网络(CNN)和反向传播人工神经网络(BP网络)两种模式的人工神经网络(ANN)分别分析了各种苦丁茶的红外谱图。作者采用25个样本作训练集,11个样本作检验集,用两种网络进行了训练。结果表明,CNN网络和BP网络均能够有效地实现苦丁茶产地的鉴别,但CNN网络能够进一步地区分苦丁茶的级别。实验表明,CNN速度快,预测结果准确,可望用竞争神经网络(CNN)和红外光谱法结合分类鉴别苦丁茶。     

应用可见-近红外光谱技术快速无损鉴别婴幼儿奶粉品种

为了快速无损鉴别婴幼儿奶粉品种,提出了结合偏最小二乘(PLS)法和人工神经网络(ANN)综合预测婴幼儿奶粉品种的新方法。获取婴幼儿奶粉样本在400～1 000 nm波段的漫反射光谱,采取平均平滑法和多元散射校正(MSC)进行预处理,用PLS建立校正模型进行模式特征分析及主成分的提取。经过交互验证法判别,提取7个主成分作为神经网络的输入变量,奶粉的品种值作为输出,建立了三层BP神经网络。9个典型品种的婴幼儿奶粉各取样本30个,共计270个作为训练集。随机抽取的各个品种的10个样本,共90个作为预测检验样本,结果表明,90个未知样本的品种预测准确率为100％。说明提出的方法具有很好的分类和鉴别作用,为婴幼儿奶粉的品种快速无损鉴别提供了一种新方法。     

Enhanced susceptibility in LNiO3 perovskites ( L = La,Pr,Nd,Nd0.5Sm0. 5)

The temperature dependence of the resistivity rho(T) and of the dc magnetic susceptibility chi(T) were measured on high-quality LNiO3 (L = La,Pr,Nd,Nd0.5Sm0.5) samples synthesized under high oxygen pressure. Subtraction of the rare-earth contribution to chi(T) allows the presentation of the evolution of the susceptibility of the NiO3 array from Pauli to Curie-Weiss paramagnetism with decreasing bandwidth. A metal-insulator transition occurring at a temperature T(t) = T(N) is first order for L = Pr and Nd; it becomes second order and produces no anomaly in chi(-1)(T) at a T(t)>T(N) for L = Nd0.5Sm0.5. In the antiferromagnetic domain T相似文献   

A novel calibration approach to structured light 3D vision inspection

Structured light 3D vision inspection is a commonly used method for various 3D surface profiling techniques. In this paper, a novel approach is proposed to generate the sufficient calibration points with high accuracy for structured light 3D vision. This approach is based on a flexible calibration target, composed of a photo-electrical aiming device and a 3D translation platform. An improved algorithm of back propagation (BP) neural network is also presented, and is successfully applied to the calibration of structured light 3D vision inspection. Finally, using the calibration points and the improved algorithm of BP neural network, the best network structure is established. The training accuracy for the best BP network structure is 0.083 mm, and its testing accuracy is 0.128 mm.     

神经网络模型在LED便携式近红外整粒小麦成分测量仪上的建立

采用三层反向传播人工神经网络,在LED便携式近红外整粒小麦成分测量仪上,建立了测定整粒小麦中粗蛋白含量的定量分析模型。给出了LED便携式近红外整粒小麦成分测量仪的结构框图、光路结构图及所采集的整粒小麦样品光谱图。简介了人工神经网络的原理并给出了建模的结果。用人工神经网络建立模型,得到校准集和预测集的相关系数分别为0.90和0.96,相对标准偏差分别为3.77%和4.46%。在所使用的有限个分立波长的测定仪上,由于仪器光路、电路及整粒样品状态的影响,存在着一定的非线性。采用人工神经网络建立的模型,其结果优于线性模型的结果。     

基于炉口辐射和改进神经网络的转炉终点预测模型

针对国内外转炉炼钢终点控制的现状,建立了一种用于终点预测的神经网络模型.以炉口辐射信息获取系统为实验平台,运用光纤谱分复用和颜色空间模型转换技术.分析发现了光谱与图像信息特征量在吹炼过程中晕现出中前期类似、末期相反的规律.从得到的特征规律曲线中选用一些关键特征量,在改进的修正系数算法基础上,进行了模型的训练和预测分析.实验结果表明:响应时间在2 s以内,满足快速判定的时间要求;改进算法的模型预测精度高于常规算法,该系统可以止常工作在转炉炼钢的恶劣环境下,达到了预期效果.     

硅基支撑系统的光刻工艺

为提高硅基支持系统制备中光刻过程的线宽精度,用正交试验分析了前烘、曝光、显影主要步骤中一些主要参数对制备结果的影响,得到了它们的影响程度的规律以及较优的参数组合。在此基础上设计和训练了合适的前向误差反向传播神经网络,对主要工艺参数进行了进一步的分析、预测和优选,并用实验加以验证。最终得到在胶厚约1.55 m时,前烘温度100 ℃,时间90 s; 曝光时间5 s; 显影温度15 ℃,时间90 s时,光刻后图形的线宽偏差最小,达到了0.3 m以下。     

Melting of the rare earth metals and f-electron delocalization

Melting curves for Pr, Nd, Sm, Gd, and Y were measured in a diamond-anvil-cell to nearly 100 GPa and 4000 K. f-electron volume collapses are observed as triple points for Pr (24 GPa and 1400 K) and Gd (65 GPa and 3100 K). These pressures coincide with the volume collapses observed at room temperature. For Nd and Sm, the f-electron volume collapse has not been observed at room temperature but appears at approximately 2000-2500 K as a broad minimum in the melting curve, similar to that of Ce, near 50 GPa (Nd) and 70 GPa (Sm). The melting curve of Y goes smoothly along the entire rare earth sequence.     

Magnetic properties of R3Co29Si4B10 (R=La, Ce, Pr, Nd, Sm, Gd and Dy) borides

The crystal structure and magnetic properties of quaternary rare-earth intermetallic borides R₃Co₂₉Si₄B₁₀ with R=La, Ce, Pr, Nd, Sm, Gd and Dy have been studied by X-ray powder diffraction and magnetization measurements. All compounds crystallize in a tetragonal crystal structure with the space group P4/nmm. Compounds with R=La, Ce, Pr, Nd and Sm are ferromagnets, while ferrimagnetic behavior is observed for R=Gd and Dy. The Curie temperatures vary between 149 K and 210 K. The Curie temperatures in R₃Co₂₉Si₄B₁₀ (R=Ce, Pr, Nd, Sm, Gd, Dy) compounds are roughly proportional to the de Gennes factors.     

采用BP神经网络研究C-F键核自旋偶合常数

通常理论研究核自旋偶合常数的方法是基于线性模型进行拟合和预测，该方法在拟合和预测中仍有较大误差. 本文在前面工作的基础上，提出了基于非线性模型对C-F键核自旋偶合常数进行研究的观点，采用BP神经网络方法对C-F键核自旋偶合常数的函数关系式进行拟合，并用拟合结果对4种化合物的偶合常数进行预测. 结果表明，采用非线性的BP神经网络方法其训练效果与预测效果均优于线性模型方法；其预测误差对文中的4种化合物不超过0.40%.     

BP神经网络结合ATLD与三维荧光光谱法测量水中多环芳烃

多环芳烃(PAHs)是煤,石油,木材,烟草等燃料和有机高分子化合物等有机物不完全燃烧时产生的一种持久性有机污染物。迄今已发现有200多种PAHs,其中有多种PAHs具有致癌性。PAHs广泛分布于我们生活的环境中,水中的PAHs主要来源于生活污水,工业排水和大气沉降。使用三维荧光光谱法,结合BP神经网络与交替三线性分解(ATLD)算法对水中的PAHs进行定性和定量分析。以苊（ANA）和芴（FLU）2种PAHs为目标分析物,用甲醇（光谱级）制备样本。使用FS920稳态荧光光谱仪对样本进行检测,设置激发波长为200～370 nm,间隔10 nm记录一个数据;发射波长为240～390 nm,间隔2 nm记录一个数据。设置初始发射波长总是滞后激发波长40 nm,以消除一级瑞利散射的干扰。随后使用BP神经网络法对待测样本数据进行预处理。利用BP神经网络基于误差反向传播算法(error back propagation training,BP)原理,对测得的三维荧光数据进行数据压缩处理,该方法具有柔性的网络结构与很强的非线性映射能力,网络的输入层、隐含层和输出层的神经元个数可根据实际情况设定,并且网络的结构不同时,性能也有所差异。随后,用ATLD算法分解预处理后的三维荧光光谱数据。采用核一致诊断法确定待测样本的组分数为2。结果表明,ATLD算法分解得到两种PAHs(ANA和FLU)的激发、发射光谱图与目标光谱非常相似,能实现光谱重叠严重的PAHs(ANA和FLU)的快速定性和定量分析,实现了以“数学分离”代替“化学分离”。将预测样本导入训练好的BP神经网络中,得到处理后待测样本数据的网络均方差（MSE）均小于0.003,网络的峰值信噪比（PSNR）均大于120dB（数据压缩中典型的峰值信噪比值在30～40 dB之间,越高越好）,可见BP神经网络对样本数据的压缩效果较好。BP神经网络训练后,得到输出值与目标值之间的拟合度高,拟合系数达0.998,具有较好的数据压缩效果。使用ATLD算法对待测样本进行分解后得到平均回收率为97.1％和98.9％,预测均方根误差为0.081 8和0.098 5 μg·L-1。三维荧光光谱结合BP神经网络和ATLD能够实现痕量PAHs的快速检测。     

Magnetic susceptibility and electrical resistivity of some Th2Zn17-type rare-earth zinc phases

We report measurements of the magnetic susceptibility and electrical resistivity of the iostructural compounds RE₂Zn₁₇ (RE=La, Ce, Pr, Nd, Sm, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu). The composition dependence of the lattice parameter and effective moment indicate that all the RE ions are trivalent except Yb which is divalent. Magnetic order is observed in compounds where RE=Ce, Pr, Nd, Sm, Gd, Tb, Dy, Ho and Er. A second transition is seen for RE=Pr, Ho, Sm and Tb. Superzone boundary effects are observed in the electrical resistivity of these four alloys as well as in Er₂Zn₁₇. Resistivity measurements reveal concentrated Kondo behavior (or 4f instability) of Ce in Ce₂Zn₁₇.     

稀土氨羧络合物的吸收光谱研究

本文系统地研究了EDTA,NTA,CyDTA,TTHA与镨,钕,钐,铕,锹,铒络合物的可见光吸收光谱及三阶导数光谱,并以Nd,Ho-TTHA体系测定了混合稀土氧化物中的Nd,Ho,首次以Sm-DTPA体系的三阶导数我谱成功地进行了纯氧化镧中杂质氧化钐的测定。     

ELECTRONIC STRUCTURE AND CRYSTALLINE ELECTRIC FIELD IN Rco5(R= Y,La, Ce,Pr, Nd,Sm, Gd,and Tb)

Self consistent charge and spin polarized local spin-density approximation functional theory calculations based on the discrete variational method have been performed for RCo₅(R=Y, La, Ce, Pr, Nd, Sm, Gd, and Tb) compounds. The partial density of states of the Pr atom in the PrCo₆Co₁₂ cluster is established to be strikingly similar to that of the Ce atom in the CeCo₆Co₁₂ cluster, supporting the suggestion that the Pr atom is valence fluctuating. The radii ＜r_4f＞ and ＜r₄f²＞ of the 4f electrons of the R atom from La to Tb, except Ce, show the lanthanide contraction. The crystalline electric field (CEF) parameter A⁰₂ at the R site is calculated using a real charge distribution ρ(R) in the cluster, except for Pr and Nd, and is in agreement with that evalu ated based on the single-ion model. This result shows that the CEF parameter A⁰₂ is mainly determined by the near electronic structure. There exists a hybridization in a certain degree between the light rare-earth R-4f and Co-3d orbitals in some single-electron-molecular-orbitals, which are n ear the Fermi energy level and occupied by electrons. For light rare-earths the R-4f electrons in R Co₆Co₁₂(R=Y, La, Ce, Pr, Nd, and Sm) clu sters are not localized entirely and a small amount of the R-4f electrons have itinerant properties.     

基于主成分分析和人工神经网络的激光诱导击穿光谱塑料分类识别方法研究

研究了人工神经网络在激光诱导击穿光谱(LIBS)塑料分类识别方面的应用。选用七种常见的塑料作为实验样品,获得每种样品的170组LIBS光谱数据,利用主成分分析获得前五个主成分的得分矩阵。用每种塑料样品的130组光谱数据的主成分得分矩阵作为训练集,建立反向传播(BP)人工神经网络模型。将其余40组主成分得分作为测试数据输入训练好的模型进行分类识别,其识别准确度达到97.5%。实验结果表明,通过采用主成分分析与BP人工神经网络相结合的方法,可以很好地进行塑料激光诱导击穿光谱的分类识别,对塑料的回收利用有重要价值。