基于光纤传感技术的油液污染度软测量研究

采用基于Mie理论的激光散射法测量油液污染度时,油液污染度与监测参数——入射光强、散射光强、出射光强以及颗粒参数之间存在着复杂的非线性关系,给油液颗粒污染度的准确测量带来困难。利用支持向量机优良的非线性映射和强大的泛化能力,建立了一个基于最小二乘支持向量机的油液污染度软测量模型,给出了相应的系统结构和算法;仿真和实际运行结果表明基于LS-SVM的油液污染度软测量模型具有较高的估算精度与泛化能力,为油液污染度的在线测量提供了一种简单、可靠的新方法。     

基于支持向量机的模拟电路软故障诊断

提出了基于支持向量机的模拟电路软故障诊断新方法.该方法提取电路的频域响应为故障特征,利用支持向量机对故障进行识别分类.支持向量机具有结构简单、泛化能力强的特点,对小样本分类具有良好的识别效果.以Sallen-Key滤波电路为诊断例,实验结果表明该方法故障诊断准确率大于99%.     

小样本数据的支持向量机回归模型参数及预测区间研究

支持向量机是由统计学习理论发展起来的机器学习算法,它从结构风险最小化的角度保证了模型的最大泛化能力.文中运用支持向量机进行小样本数据回归分析研究.首先利用推广性的界理论指导支持向量机回归模型参数的选取,以保证模型具有最大的推广能力;其次,运用基于正态分布和基于t分布的两种区间预测方法进行了预测值的区间估计;最后,利用模拟序列和真实的航空发动机油样光谱分析数据作为实验数据,建立了支持向量机回归分析模型,并与最小二乘法进行了比较.结果表明,所提出的支持向量机模型参数选取和区间估计方法适用于小样本数据的回归分析,具有较高的预测精度.     

装载机载重测量的支持向量机软测量建模方法

在阐述支持向量机(SVM)和最小二乘支持向量机(LS-SVM)的原理和算法并对两者的特点进行比较后,为装载机载重测量建立了基于LS-SVM的软测量模型,并从核函数选择以及核参数确定两方面阐述了LS-SVM软测量建模的过程,最后与RBF函数网络以及BP网络的软测量建模结果进行对比.仿真分析结果表明,LS-SVM同时兼顾了精度和泛化能力两方面的性能,其最大泛化误差仅为6.863 8×10-6,是适合装载机载重软测量的建模方法.     

基于支持向量机方法的压电智能结构系统辨识

支持向量机(SVM)是一种基于统计学习理论和结构风险最小化原则的新型机器学习方法,克服了传统机器学习方法在训练中的局部极小问题,过学习和欠学习等问题,具有很好的泛化能力.本文介绍了应用于回归分析的最小二乘支持向量机(LS-SVM)的基本理论,然后以随机激励下压电智能结构的响应数据作为训练样本集,采用LS-SVM方法辨识系统,仿真结果验证了这种方法的有效性.     

基于遗传算法和最小二乘支持向量机的织物剪切性能预测

提出了一种基于最小二乘支持向量机的织物剪切性能预测模型,并且采用遗传算法进行最小二乘支持向量机的参数优化,将获得的样本进行归一化处理后,将其输入预测模型以得到预测结果.仿真结果表明,基于最小二乘支持向量机的预测模型比BP神经网络和线性回归方法具有更高的精度和范化能力.     

基于LS-SVM的某型装备保障性能评估研究

针对军事油料装备保障性能指标参数多样性、变化性的特点,研究了新型支持向量机算法--最小二乘支持向量机,并设计了基于多元分类的最小二乘支持向量机;建立了装备保障性能评估的最小二乘支持向量机决策模型;以某型油料装备为例,运用模型确立的保障性能评估指标体系和支持矢量学习决策模式,对其保障性能进行了评估.该研究结果对于优化提高装备保障性能,增强装备保障能力具有重要现实意义.     

基于支持向量机的齿轮故障诊断方法研究

故障样本的不足从一定程度上制约了基于知识的方法在实际故障诊断中的应用，针对这一问题，利用支持向量机在小样本情况下具有较强分类能力的特点，提出了一种基于支持向量机的齿轮故障诊断方法。该方法采用小波变换对齿轮的振动信号进行处理来构造特征向量，并直接输入到支持向量机的多故障分类器中进行故障识别。试验结果表明该方法是有效、可行的，且在小样本情况下比BP神经网络具有更高的诊断精度。     

基于小波包和进化支持向量机的齿轮早期诊断研究

针对齿轮早期故障的特征不明显，提出了一种基于小波包和进化支持向量机的齿轮故障诊断方法，该方法既充分利用了小波包优良的时频局部化特性，又利用了支持向量机在小样本情况下出色的学习性能和良好的推广特性，以及遗传算法的全局优化能力。在齿轮试验台上的应用结果表明，经过特征提取和参数优化后，提高了支持向量机的分类能力。     

基于声成像模式识别的故障诊断方法研究

利用非接触式声学故障诊断技术的工作优点,发展了基于支持向量机的声成像模式识别技术,并引入到故障诊断领域。针对某种机械故障特点,在利用波束形成算法得到声像进行噪声源识别与定位的基础上,对声像进行图像处理,提取声像的纹理特征和奇异值特征,采用支持向量机进行训练分类,进而用于机械工作状态的诊断。通过仿真及实验,得到了较好的诊断效果,表明基于声像的图像特征提取技术在结合支持向量机后可应用于机械故障诊断,为声成像方法在故障诊断领域的应用提供参考。     

Coherent imaging with two-dimensional focal-plane arrays: design and applications

Scanned, single-channel optical heterodyne detection has been used in a variety of lidar applications from ranging and velocity measurements to differential absorption spectroscopy. We describe the design of a coherent camera system that is based on a two-dimensional staring array of heterodyne receivers for coherent imaging applications. Experimental results with a single HgCdTe detector translated in the image plane to form a synthetic two-dimensional array demonstrate the ability to obtain passive heterodyne images of chemical vapor plumes that are invisible to normal video infrared cameras. We describe active heterodyne imaging experiments with use of focal-plane arrays that yield hard-body Doppler lidar images and also demonstrate spatial averaging to reduce speckle effects in static coherent images.     

粗略不相似度量及其在层次聚类中的应用

局部结构特征在数据分析过程中具有重要的作用.为获得简单有效的数据集局部结构化特征检测方法,本文结合重采样误差分析和传统的近邻选择方法提出了一种检测局部结构特征的方向一致性度量—粗略不相似性度量.该度量是一种优化的近邻选择方法,不仅考虑了传统的欧氏距离排序,而且考虑了局部方向结构特征.因其计算和存储复杂度小以及具有优越的结构检测性能,可应用于无监督学习形成一种层次化的子图聚类算法—RDClust,与经典聚类算法相比,其优势在于:一是计算复杂度较小,是近似线性算法;二是无需对类的形状和分布形式做任何的假设,可自动体现数据集的局部结构;三是有一个近邻参数,且该参数对结果较鲁棒.在人工和真实数据集上的实验显示了新的度量方式应用于新算法的优越性能.     

基于啁啾脉冲的反射层析激光雷达成像

反射层析激光雷达成像具有系统简单及对湍流不敏感的优势,但普通脉冲发射信号难以兼顾远距离和高精度探测的要求.基于上述情况,本文提出了一种基于啁啾脉冲信号的反射层析激光雷达成像处理方法.该方法首先对啁啾脉冲回波进行相干压缩处理,通过对处理后信号的包络提取得到目标在各方向的反射层析投影数据,最后利用卷积反投影算法实现高分辨力的图像重构.仿真结果表明,在同一投影角度,利用该方法得到的包络与目标反射率投影相一致;在投影角度范围大于60°时能够得到目标的轮廓信息,角度范围越大,成像越精确.研究结果验证了该方法的有效性.     

A PU-based meshless Shepard interpolation method satisfying delta property

A novel meshless Shepard interpolation method (MSIM) based on the partition of unity (PU) approach is proposed to solve the elasticity problems. In the proposed MSIM interpolation, the Shepard shape functions are used for the partition of unity and the local cover functions are separately constructed for the nodes on the boundary and inside the domain. Three distinguished features of MSIM are: the interpolation property, the arbitrarily high order consistency, and the low computational expense of the shape functions. These properties are desirable in the realization and implementation of the meshless methods. Numerical examples demonstrate the effectiveness and robustness of the present method.     

Calibration method for multiangle lidar measurements

A new method based on a two-angle approach is developed to determine the lidar solution constant from scanning elastic lidar data, hence providing a relative calibration for each lidar scan. Once the solution constant is determined, the vertical profiles of atmospheric extinction can be calculated. With this calibration method a minimization technique is used that replaces the linear regression used in a known two-angle approach that requires only local atmospheric homogeneity over a restricted altitude calibration range rather than overall horizontal homogeneity. Lidar signals from at least one pair of elevation angles are used, averaged in time when the system is operated in a permanent two-angle mode, or an arbitrary number of signal pairs is used, when a two-dimensional lidar scan is being processed. The method is tested extensively with synthetic data. The calibration method is a robust tool for determining the solution constant to the lidar equation and for obtaining vertical profiles of atmospheric extinction.     

Analytical differentiation of the differential-absorption-lidar data distorted by noise

A method of analytical differentiation is developed for processing differential absorption lidar (DIAL) data. The method is based on simple analytical transformation of the DIAL on and off signal ratio. The derivatives consequently are found for either individual data points or local zones of the measurement range. The method makes possible the separation of local zones of interest and the separate investigation of these. The smoothing level is established by the selected value of the exponent in a transformation formula rather than by the selection of the resolution range. The method does not require the calculation of local signal increments. This reduces significantly the high-frequency noise in the measured concentration. The method is general and can be used for different experimental data, including inelastic (Raman) lidar data. The processing technique is practical and does not require a determination of the solution for a large set of algebraic equations. It is based on the simple repetition of the same type of calculations with different constants. The method can easily be implemented for practical computations.     

Parry arc: a polarization lidar, ray-tracing, and aircraft case study

Using simple ray-tracing simulations, the cause of the rare Parry arc has been linked historically to horizontally oriented columns that display the peculiar ability to fall with a pair of prism faces closely parallel to the ground. Although we understand the aerodynamic forces that orient the long-column axis in the horizontal plane, which gives rise to the relatively common tangent arcs of the 22 degrees halo, the mechanism leading to the Parry crystal orientation has never been resolved adequately. On 16 November 1998, at the University of Utah Facility for Atmospheric Remote Sensing, we studied a cirrus cloud producing a classic upper Parry arc using polarization lidar and an aircraft with a new high-resolution ice crystal imaging probe. Scanning lidar data, which reveal extremely high linear depolarization ratios delta a few degrees off the zenith direction, are simulated with ray-tracing theory to determine the ice crystal properties that reproduce this previously unknown behavior. It is found that a limited range of thick-plate crystal axis (length-to-diameter) ratios from ~0.75 to 0.93 generates a maximum delta approximately 2.0-5.0 for vertically polarized 0.532-mum light when the lidar is tilted 1 degrees -2 degrees off the zenith. Halo simulations based on these crystal properties also generate a Parry arc. However, although such particles are abundant in the in situ data in the height interval indicated by the lidar, one still has to invoke an aerodynamic stabilization force to produce properly oriented particles. Although we speculate on a possible mechanism, further research is needed into this new explanation for the Parry arc.     

Determination of overlap in lidar systems

The overlap profile, also known as crossover function or geometric form factor, is often a source of uncertainty for lidar measurements. This paper describes a method for measuring the overlap by presenting the lidar with a virtual cloud through the use of an imaging system. Results show good agreement with horizontal hard target lidar measurements and with geometric overlap calculated for the ideal aberration-free case.     

Stable near-end solution of the lidar equation for clear atmospheres

A stable variant of the near-end solution has been developed for inversion of lidar signals measured in clear atmospheres. The inversion is based on the use of reference values of the extinction coefficient obtained with a nephelometer at the lidar measurement site. The inversion method, based on a combination of the optical depth and boundary point solutions, is illustrated by simulated and experimental data.     

Comparison of Continuous-Wave CO(2) Lidar Calibration by Use of Earth-Surface Targets in Laboratory and Airborne Measurements

Backscatter of several Earth surfaces was characterized in the laboratory as a function of incidence angle with a focused continuous-wave 9.1-mum CO(2) Doppler lidar for use as possible calibration targets. Some targets showed negligible angular dependence, while others showed a slight increase with decreasing angle. The Earth-surface signal measured over the complex Californian terrain during a 1995 NASA airborne mission compared well with laboratory data. Distributions of the Earth's surface signal shows that the lidar efficiency can be estimated with a fair degree of accuracy, preferably with uniform Earth-surface targets during flight for airborne or space-based lidar.