基于知识迁移的深度学习无线电信号聚类方法

现有的无线电信号调制识别方法在先验数据不足时通常很难对无类标信号进行有效识别。针对这个问题,本文提出了一种基于知识迁移的深度学习无线电信号聚类方法(DTC)。该方法基于样本对比,分析样本间的相似性,并利用卷积神经网络(CNN)提取无线电信号的特征,同时设计了一种预训练框架,通过迁移同领域数据集的知识,有效提升了CNN特征提取能力,实现了引导聚类方向、提升聚类性能的目标。实验结果表明,该方法在多个公开数据集上的聚类性能都显著优于现有的聚类方法。与现有方法相比,DTC在RML 2016.10A和RML 2016.04C数据集上的聚类精度分别提升了30.34%和28.04%。     

一种基于分子结构设计理论的聚类分析方法

针对常用的无监督聚类分析方法中存在的问题,提出了一种基于分子结构设计理论的聚类分析方法。该方法借鉴分子结构设计理论模型,将故障样本空间看作分子系统,将故障样本看作分子系统中的原子,以故障样本之间的差异度作为分子势能的度量指标,在故障样本间"相互作用势"的影响下,以样本间"势能"最小为依据,调整故障样本在映射平面上的位置,从而获得最佳的聚类效果。开展了不同状态滚动轴承振动测试实验,聚类结果表明,相比于SOM聚类方法,该方法将聚类有效性指标DB值降低49.04%。将该方法应用于柴油机故障振动数据的聚类中,实验结果表明聚类效果良好,能够有效地将不同故障的数据区分开,验证了该方法的可行性和有效性。     

基于经验模态分解样本熵的肌电信号识别

针对表面肌电信号非线性、噪声强等特点,设计一种快速有效的表面肌电信号手部多运动模式识别方法,用于肌电假手的实时控制。提出了一种基于经验模态分解样本熵和聚类分析的表面肌电信号多运动模式识别方法。该方法对动作持续阶段的信号首先进行经验模态分解,将其分解为多个平稳的固有模态函数（ IMF）,再依据频率有效度选取若干个包含肌电信号有效信息的IMF分量求和后,计算其样本熵。以尺侧腕伸肌和尺侧腕屈肌两路肌电信号对应的EMD样本熵作为特征向量,设计了主轴核聚类算法的聚类分类器进行模式识别。成功识别了展拳、握拳、腕上翻和腕下翻4种动作,平均识别率达到93%。该方法取得了较高的识别率,抗干扰能力强,计算量少,可用于肌电假手的控制。     

基于复杂网络社团聚类的故障模式识别方法研究

复杂网络是近年兴起的一种新的理论,已迅速渗透到从自然科学到工程科学等多个领域。本文从复杂网络社团结构的本质特性出发,将故障样本抽象为网络节点,样本与样本之间的联系抽象为边,建立故障数据网络模型。利用复杂网络节点关联度的概念选取社团初始聚类中心,利用欧氏距离函数实现网络社团初始划分,设计社团区分准则函数,并引入模块性合并指标变化进行同类社团合并,最终实现准确的社团聚类与故障诊断。滚动轴承故障诊断实例验证了方法的有效性。     

基于模糊c均值聚类算法的控制图模式识别

控制图模式识别能够区分制造过程中的一般因素与异常因素,提高制造过程中的产品质量,减少成本,提高效益。利用蒙特卡洛方法产生样本;采用一维离散小波变换处理原始数据;利用模糊c均值聚类算法进行控制图模式识别。识别准确率99.43%,其标准差为0.002 8。这表明基于该方法的控制图模式识别准确率高,稳定性好,较现有的控制图模式识别方法具有简易、高效等特点。     

采用小波框架的纺织品缺陷分类方法

针对纺织品缺陷分类,提出了一种基于判别小波框架的分类方法.该方法采用小波框架来描述纺织品图像的多尺度纹理特性,并设计与纺织品缺陷纹理相对应的小波框架函数来替代标准小波函数,来更有效地描述各类缺陷纹理的内在结构差异.在判别特征提取训练方法框架下,通过将小波框架函数和分类器两者的设计相联合,来实现缺陷分类错误概率的最小化.对8类纺织品缺陷的466个样本,以及434个无缺陷样本进行了分类实验,获得了95.8%的分类准确率.     

基于聚类分析的光谱重建样本分析

目的 解决目前光谱重建中因数量大而出现的冗余和繁重等问题,证明聚类算法可以很好地应用在光谱选择样本分析中,并可以实现较高的重构色度精度和物理精度。方法 采用主成分分析法进行仿真实验,首先探究主成分个数,再确定聚类个数,然后比较聚类方法和3种常用的样本选择方法,最后分析比较光源种类对重构结果的影响。结果 通过实验确定主成分个数为6且聚类个数为20时,在A光源下使用KFCM算法的重构效果最好,此时平均色差为0.35ΔE00,平均RMSE为0.0078,平均GFC为99.94%。结论 聚类方法可以应用于光谱成像过程中训练样本选择过程,且有助于提高光谱重构的运算速度和精度。     

模式识别技术在结构可靠度分析中的应用

提出了一种重构功能函数的新方法。该方法采用模式识别中的最小错分样本数准则以实现对样本点较好的分类效果,从而得到功能函数的一个较好逼近。然后通过迭代算法修正每步得到的设计点以实现在真实设计点附近较高精度地逼近真实功能函数。它克服了一般响应面方法需用到功能函数值的不足。数值算例表明:该方法具有较高的精度且所需样本数量较少。     

基于自然语言理解的需求聚类和需求优选方法研究

为了实现大数量软件需求的优选,提出了一种基于自然语言理解的需求聚类和需求优选方法。该方法首先基于自然语言处理技术对需求进行深层次语言理解,找到相同语义的需求并只保留一条。随后,采用基于超图分割的需求聚类方法对大量需求进行聚类。最后,提出了需求优选目标函数,该目标函数根据需求聚类结果为需求优先级打分,并在综合考虑需求与类别的相似性、类别的权重以及聚类标准的权重的基础上,最终给出一个合理的需求优选结果。在大规模真实的需求集合上进行了实验,实验结果表明:基于自然语言处理技术的需求聚类性能优异;需求聚类对后续的需求优选起有着非常重要的作用;基于超图分割的需求聚类,以及在此基础上提出的需求优选函数相对于基线方法有着明显的提高。     

基于灰度-梯度共生矩阵和模糊核聚类的振动图形识别方法

以往复机械振动参数图形为对象,提出了基于灰度-梯度共生矩阵和模糊核聚类的振动图形识别方法。利用灰度-梯度共生矩阵直接提取振动参数图形中的特征信息,将得到的纹理特征参量作为样本输入空间,通过Mercer核把输入样本映射到高斯特征空间后,在高维特征空间中进行聚类,从而实现往复机械故障智能诊断。实验结果表明,该方法可以获得较高的诊断精度,具有一定的可行性和有效性。     

Node placement for triangular mesh generation by Monte Carlo simulation

A new approach of node placement for unstructured mesh generation is proposed. It is based on the Monte Carlo method to position nodes for triangular or tetrahedral meshes. Surface or volume geometries to be meshed are treated as atomic systems, and mesh nodes are considered as interacting particles. By minimizing system potential energy with Monte Carlo simulation, particles are placed into a near‐optimal configuration. Well‐shaped triangles or tetrahedra can then be created after connecting the nodes by constrained Delaunay triangulation or tetrahedrization. The algorithm is simple, easy to implement, and works in an almost identical way for 2D and 3D meshing. Copyright © 2005 John Wiley & Sons, Ltd.     

考虑非正定相关性的测量不确定度蒙特卡洛评定方法

当采用蒙特卡洛法评定测量不确定度考虑输入量相关性时,需基于Nataf逆变换产生服从任意边缘概率分布的相关多维随机变量。为了解决Nataf逆变换过程中输入量相关系数矩阵非正定时,无法产生线性变换矩阵的问题,提出了基于Barzilai-Borwein梯度法的迭代修正算法。进而探讨了输入量服从非正态分布且相关的蒙特卡洛法实施步骤。最后,采用提出的迭代修正算法并基于Nataf逆变换的蒙特卡洛方法,对高速轮轨试验台轮轨纵向蠕滑率不确定度进行了评定,验证了该算法的可行性及有效性。     

A spatial variable selection method for monitoring product surface

Two-dimensional (2-D) data maps are generated in certain advanced manufacturing processes. Such maps contain rich information about process variation and product quality status. As a proven effective quality control technique, statistical process control (SPC) has been widely used in different processes for shift detection and assignable cause identification. However, charting algorithms for 2-D data maps are still vacant. This paper proposes a variable selection-based SPC method for monitoring 2-D wafer surface. The fused LASSO algorithm is firstly employed to identify potentially shifted sites on the surface; a charting statistic is then developed to detect statistically significant shifts. As the variable selection algorithm can nicely preserve shift patterns in spatial clusters, the newly proposed chart is proved to be both effective in detecting shifts and capable of providing diagnostic information for process improvement. Extensive Monte Carlo simulations and a real example have been used to demonstrate the effectiveness and usage of the proposed method.     

Monte Carlo approach for solving the radiative transfer equation over mountainous and heterogeneous areas

An algorithm based on the Monte Carlo method is developed to solve the radiative transfer equation in the reflective domain (0.4-4 mum) of the solar spectrum over rugged terrain. This algorithm takes into account relief, spatial heterogeneity, and ground bidirectional reflectance. The method permits the computation of irradiance components at ground level and radiance terms reaching an airborne or satelliteborne sensor. The Monte Carlo method consists of statistically simulating the paths of photons inside the Earth-atmosphere system to reproduce physical phenomena while introducing neither analytical modeling nor assumption. The potentialities of the code are then depicted over different types of landscape, including a seashore, a desert region, and a steep mountainous valley.     

Optimization of kinoform lenses with the Monte Carlo method

For the optimization of non-Fourier-type computer-generated phase holograms (kinoform lenses), a method based on the Monte Carlo procedure is suggested. This method can be regarded as analogous to the iterative Fourier transform algorithm method that is widely used for the optimization of Fourier-type computer-generated phase holograms (kinoforms).     

An efficient analytical Bayesian method for reliability and system response updating based on Laplace and inverse first-order reliability computations

This paper presents an efficient analytical Bayesian method for reliability and system response updating without using simulations. The method includes additional information such as measurement data via Bayesian modeling to reduce estimation uncertainties. Laplace approximation method is used to evaluate Bayesian posterior distributions analytically. An efficient algorithm based on inverse first-order reliability method is developed to evaluate system responses given a reliability index or confidence interval. Since the proposed method involves no simulations such as Monte Carlo or Markov chain Monte Carlo simulations, the overall computational efficiency improves significantly, particularly for problems with complicated performance functions. A practical fatigue crack propagation problem with experimental data, and a structural scale example are presented for methodology demonstration. The accuracy and computational efficiency of the proposed method are compared with traditional simulation-based methods.     

Computerized Simulation on the Structure of Cryolite-Alumina Melts

Monte Carlo calculations were performedon liquid ionic solution of saturated aluminain cryolite melt containing a total of 205 ions.The local structure and radial distributionfunctions of the melt were computerized by MonteCarlo method at 1283 K. It has been found thatthere exist the complex ionic clusters suchas AlOF_3~(2-), AlOF_4~(3-), AlF_4~-, AlF_5~(2-), AlOF_2~- etc.,as well as oxygen-bridge and flourine-bridgeionic clusters. In the melt studied, Al~(3+) ionstend to have preferential even coordinationnumber of 4F~- ions.     

From the interface energy to the solubility limit of aluminium in nickel from first-principles and Kinetic Monte Carlo calculations

We developed a way to predict the solubility limit of solute atoms in a binary alloy using a Kinetic Monte Carlo algorithm. The idea is to use the interface energies calculated by first-principles calculations to parameterize the pair interaction energies used in the Kinetic Monte Carlo algorithm. In order to validate this method, it was tested on a very well known case: the Ni-Al alloy. We found that the calculations are in very good agreement with the previously calculated phase diagrams.     

Identification of hazardous road locations of traffic accidents by means of kernel density estimation and cluster significance evaluation

This paper proposes a procedure which evaluates clusters of traffic accident and organizes them according to their significance. The standard kernel density estimation was extended by statistical significance testing of the resulting clusters of the traffic accidents. This allowed us to identify the most important clusters within each section. They represent places where the kernel density function exceeds the significance level corresponding to the 95th percentile level, which is estimated using the Monte Carlo simulations. To show only the most important clusters within a set of sections, we introduced the cluster strength and cluster stability evaluation procedures. The method was applied in the Southern Moravia Region of the Czech Republic.