基于EEMD-SVM的刀具磨损状态研究

针对刀具磨损监测中信号的非平稳特性和小样本建模中神经网络容易陷入局部值的问题,提出基于多传感器信号,运用集合经验模态分解(ensemble empirical mode decomposition,EEMD)和支持向量机(support vector machine,SVM)相结合的算法,实现对刀具磨损多状态的识别。首先对振动信号进行集合经验模态分解,将其分解为若干个本征模态函数(intrinsic mode function,IMF)之和,然后计算得到三向切削力信号的均值和各本征模态函数分量的能量百分比值作为磨损状态分类特征,最后运用支持向量机和Elman神经网络对刀具在不同磨损状态下的特征数据样本进行训练和识别。实验结果证明该方法能很好地实现对刀具磨损状态的识别,与Elman神经网络相比,支持向量机具有更高的识别率,更适合小样本情况下刀具磨损状态的分类识别。     

支持向量机在电机转子故障诊断识别中的应用

提出了一种基于支持向量机的鼠笼式电机转子断条故障检测方法,通过对电机转子断条故障进行实验模拟,获取了采样信号,利用支持向量机(SVM)对故障样本进行训练,使得支持向量机(SVM)具有分类功能.最后,采用支持向量机(SVM)对电动机各种转子断条故障进行诊断分类,取得较满意的结果.     

基于支持向量机的印品缺陷分类方法

目的研究印品图像的各类形状缺陷,建立基于支持向量机(Support vector machine,SVM)的印品形状缺陷分类模型。方法对印品进行符合人眼视觉特性的缺陷识别,并对提取缺陷进行特征分析。将特征数据导入支持向量机进行训练学习,SVM分类器对缺陷图像进行测试。结果分类器对点缺陷和面缺陷的识别率为100%,对线缺陷的分类准确率达93.94%。结论基于SVM的缺陷分类方法能较好地满足印品质量检测的需求。     

基于模糊聚类和模糊支持向量机的湿法炼锌净化除钴过程建模

提出了一种基于模糊C-均值(FCM)聚类和模糊支持向量机(SVM)方法相结合的湿法炼锌净化除钴过程建模方法。该方法针对样本空间影响支持向量机泛化性能和样本数量影响计算复杂度的问题,首先采用模糊聚类将学习样本分类,然后在各个类的样本空间内采用模糊支持向量机进行训练,并对各支持向量机模型的输出加权作为过程模型的输出。以净化除钴过程生产数据进行实验验证的结果表明,该方法明显减少了模型的训练时间,模型具有精度高、泛化性能好等特点,可以用于净化过程的优化控制。     

基于DEMD和模糊熵的滚动轴承故障诊断方法研究

提出一种基于微分的经验模式分解(DEMD)模糊熵和支持向量机(SVM)相结合的滚动轴承故障诊断方法。首先对信号进行基于微分的经验模式分解,得到若干具有物理意义的本征模函数(IMF)分量,再利用相关度准则对固有模式分量进行筛选,计算所选分量的模糊熵,组成故障特征向量,然后将其作为支持向量机的输入来识别滚动轴承的状态。并将该方法与基于EMD模糊熵和SVM相结合的方法进行比较,实验结果表明该方法对机械故障信号能够更有效准确地进行识别分类。     

基于SVM的南极望远镜驱动非预期故障诊断方法

针对南极望远镜驱动系统的非预期故障检测存在先验信息不足、故障特征难确定和故障样本少等问题,提出一种基于支持向量机(support vector machine, SVM)的非预期故障检测方法。以南极望远镜驱动系统为实验平台故障植入,采集的数据中心化和标准化预处理。基于KNN(K-nearest neighbor)、K-means、BP(back propagation)神经网络和SVM算法建立4种非预期故障检测分类器,将各个算法参数调优,再根据数据特征预测分类。实验结果表明：在相同的实验条件下,基于SVM算法的非预期故障检测分类器性能优于其他3种分类器性能。将该类方法应用于半实物仿真平台,验证该算法可行、有效。     

不均衡数据下基于SVM的故障检测新算法

针对传统支持向量机(SVM)算法在数据不均衡情况下无法有效实现故障检测的不足,提出一种基于过抽样和代价敏感支持向量机相结合的故障检测新算法。该算法首先利用边界人工少数类过抽样技术(BSMOTE)实现训练样本的均衡。为减少人工增加样本带来的噪声影响,利用K近邻构造一个代价敏感的支持向量机(CSSVM)算法,利用每个样本的代价函数消除噪声样本对SVM算法分类精度的影响。将该算法应用在轴承故障检测中,并同传统的SVM算法,不同类代价敏感SVM-C算法,SVM和SMOTE相结合的算法进行比较,试验结果表明当样本不均衡时,建议算法的故障检测性能较其它算法有显著提高。     

齿轮箱故障非线性特征测度及状态TWSVM辨识研究EI北大核心CSCD

针对齿轮箱振动的非线性,利用非线性特征测度的方法提取齿轮箱振动信号的故障特征。并利用双子支持向量机(TWSVM)对齿轮箱故障类别的辨识性能进行研究。TWSVM努力构造两个非平行的超平面来实现分类,它比支持向量机(SVM)针对多分类问题具有更好的样本不均衡适应性,并且分类性能优势明显。对齿轮箱故障类别辨识的实验表明,与传统的SVM和BP神经网络算法相比较,TWSVM具有更高的分类准确率。     

齿轮箱故障非线性特征测度及状态TWSVM辨识研究

针对齿轮箱振动的非线性,利用非线性特征测度的方法提取齿轮箱振动信号的故障特征。并利用双子支持向量机(TWSVM)对齿轮箱故障类别的辨识性能进行研究。TWSVM努力构造两个非平行的超平面来实现分类,它比支持向量机(SVM)针对多分类问题具有更好的样本不均衡适应性,并且分类性能优势明显。对齿轮箱故障类别辨识的实验表明,与传统的SVM和BP神经网络算法相比较,TWSVM具有更高的分类准确率。     

基于一类超球面支持向量机的机械故障诊断研究

针对机械故障诊断中故障类样本不易获取以及样本分布不均的问题,提出了基于一类超球面支持向量机(SVM)的故障诊断方法,该方法只需要对正常类样本进行训练.试验分析了异常类样本缺失对一类超球面支持向量机性能的影响,并提出模型参数优化选择方法,以提高分类模型的推广能力.分析了不同训练结果的分类能力,并对一类超球面支持向量机与一类超平面支持向量机的分类结果进行比较,验证了前者的正确性和有效性.     

Nanomachining by rubbing at ultrasonic frequency under controlled shear force

This study proposes a new method of proximal-probe machining that uses a rubbing process by introducing concentrated-mass (CM) cantilevers. At the second resonance of the CM cantilever vibration, the tip site of the cantilever becomes a node of the standing deflection wave because of the sufficient inertia of the attached concentrated mass. The tip makes a cyclic motion that is tangential to the sample surface, not vertical to it, as in a tapping motion. This lateral tip motion that is selectively excited by CM cantilevers was effective for the material modification of a sample due to the friction between the tip and the sample. Imaging and nanomachining under controlled shear force were demonstrated by means of the modified cantilever and a normal atomic force microscope. We were able to write a micron-sized letter "Z" having a line width of 30-100 nm on a resin surface.     

Combined AFM nano-machining and reactive ion etching to fabricate high aspect ratio structures

In this paper, a new combined method of sub-micron high aspect ratio structure fabrication is developed which can be used for production of nano imprint template. The process includes atomic force microscope (AFM) scratch nano-machining and reactive ion etching (RIE) fabrication. First, 40 nm aluminum film was deposited on the silicon substrate by magnetron sputtering, and then sub-micron grooves were fabricated on the aluminum film by nano scratch using AFM diamond tip. As aluminum film is a good mask for etching silicon, high aspect ratio structures were finally fabricated by RIE process. The fabricated structures were studied by SEM, which shows that the grooves are about 400 nm in width and 5 microm in depth. To obtain sub-micron scale groove structures on the aluminum film, experiments of nanomachining on aluminum films under various machining conditions were conducted. The depths of the grooves fabricated using different scratch loads were also studied by the AFM. The result shows that the material properties of the film/substrate are elastic-plastic following nearly a bilinear law with isotropic strain hardening. Combined AFM nanomachining and RIE process provides a relative lower cost nano fabrication technique than traditional e-beam lithography, and it has a good prospect in nano imprint template fabrication.     

Nanoscale wear and machining behavior of nanolayer interfaces

An atomic force microscope was used to subnanometer incise a nanomultilayer to consequently expose individual nanolayers and interfaces on which sliding and scanning nanowear/machining have been performed. The letter reports the first observation on the nanoscale where (i) atomic debris forms in a collective manner, most-likely by deformation and rupture of atomic bonds, and (ii) the nanolayer interfaces possess a much higher wear resistance (desired for nanomachines) or lower machinability (not desired for nanomachining) than the layers.     

基于AFM的纳米级表面加工中切屑形成的影响因素

以原子力显微镜（AFM）为加工工具进行了纳米级加工实验,对不同加工条件下的材料去除过程和切屑形态进行了研究．切屑形态通过扫描电子显微镜（SEM）进行观察,分析了不同垂直载荷、循环次数和针尖加工方向下铝铜被加工表面的切屑形成过程．实验结果表明：低栽下切屑呈细小断屑,散布在加工区域周围;随着垂直载荷的增加,切屑逐渐变成连续的带状切屑．不同循环次数、针尖加工面时切屑形成都有很大影响．在此基础上,对比分析了相同实验条件下,不同力学性能材料的切屑形成过程．最后,通过检测被加工表面得出被加工表面质量与切屑的数量和形态之间的关系,提出了改善被加工表面质量的方法,以帮助人们更好地理解基于AFM的纳米级加工技术．     

High-rate tunable ultrasonic force regulated nanomachining lithography with an atomic force microscope

This paper describes a high-rate tunable nanomachining-based nanolithography technique using an atomic force microscope (AFM). Controlled vibration between the cantilever tip and the sample is introduced to increase the lithographical speed and controllability of the nanomachining process. In this approach, an ultrasonic z?vibration of the sample and the resulting ultrasonic force from the nonlinear force-distance interaction between the sample and the cantilever tip are utilized to regulate fabrication depth. A high frequency in-plane circular vibration is introduced between the tip and the sample to control the width of the fabricated features, and to improve the speed of nanolithography. Features (e.g.?slots) with dimensions spanning from tens of nanometers to hundreds of nanometers are fabricated in one scan. A lithography speed of tens of microns per second can be achieved, which is significantly higher than other known mechanical-modification-based nanolithography methods. The patterns, that are machined on a thin PMMA film, are transferred to silicon substrate through a reactive ion etching process, which provides a cost-effective tunable approach for the fabrication of nanostructures.     

考虑刀具磨损影响的CFRP复合材料钻削轴向力预测

在碳纤维增强树脂(CFRP)复合材料钻削过程中,随着刀具磨损量的累积,轴向力会逐渐增加,轴向力过大会导致CFRP复合材料一系列的加工缺陷。为实现在CFRP复合材料钻削过程中随刀具磨损量的累积轴向力变化的有限元分析及预测,建立了CFRP复合材料钻削仿真模型,通过对ABAQUS仿真软件二次开发,利用Python语言开发子程序,将考虑磨损量累积的轴向力预测模型导入仿真软件,运用ABAQUS软件对CFRP复合材料钻削中轴向力进行研究,实现了随着刀具磨损量累积轴向力变化的预测功能。随后通过CFRP复合材料钻削试验,分析了轴向力随钻削孔数的变化规律,以验证轴向力的预测结果。结果表明:3D钻削有限元模型能够良好地预测实际加工过程中刀具未磨损时轴向力的大小,其误差为9.10%;在考虑磨损量累积后,轴向力预测模型能够较准确地预测实际加工过程轴向力的大小,其最大误差不超过10%。      

Evaluation and analysis of cutting speed,wire wear ratio,and dimensional deviation of wire electric discharge machining of super alloy Udimet-L605 using support vector machine and grey relational analysis

The current study investigates the behavior of wire electric discharge machining (WEDM) of the super alloy Udimet-L605 by employing sophisticated machine learning approaches.The experimental work was designed on the basis of the Taguchi orthogonal L27 array,considering six explanatory variables and evaluating their influences on the cutting speed,wire wear ratio (WWR),and dimensional deviation (DD).A support vector machine (SVM) algorithm using a normalized poly-kernel and a radial-basis flow kernel is recommended for modeling the wire electric discharge machining process.The grey relational analysis (GRA) approach was utilized to obtain the optimal combination of process variables simultaneously, providing the desirable outcome for the cutting speed, WWR,and DD.Scanning electron microscope and energy dispersive X-ray analyses of the samples were performed for the confirmation of the results.An SVM based on the radial-basis kernel model dominated the normalized polykernel model.The optimal combination of process variables for a mutually desirable outcome for the cutting speed,WWR,and DD was determined as T_on1,T_off2,I_P1, WT3,SV1,and WF3.The pulse-on time is the significant variable influencing the cutting speed,WWR,and DD.The largest percentage of copper (8.66%) was observed at the highest cutting speed setting of the machine compared to 7.05% of copper at the low cutting speed setting of the machine.The full text can be downloaded at https://link.springer.com/article/10.1007/s40436-017-0192-7     

An investigation on the optimum machinability of NiTi based shape memory alloy

In the present work, the machinability of nickel–titanium (Nitinol) shape memory alloy has been discussed. Nitinol is known as a difficult-to-machine alloy due to its high hardness, which requires a large amount of cutting force, resulting in high rate of tool wearing. Therefore, researchers have made an effort to ameliorate the machinability of this material to achieve a finer surface quality. The previous studies found that the cutting speed will remarkably influence the surface properties of machined nickel–titanium alloy in turning process. Tool wear and cutting force are at minimum values in a particular range of cutting speeds so that it leads to diminishing machining barriers such as burr formation and chip-breaking. Lower cutting force and consequently lower temperature and stresses in the machining process improve the mechanical properties as well as reducing hardness, distortion, and residual stress. The machining process was optimized by applying a numerical approach through ANSYS/LS-DYNA R15 software. The obtained results demonstrated the optimum cutting speed in the machining process, which are in good agreement with experiments.     

Performance of alumina-based ceramic inserts in high-speed machining of nimonic 80A

The study of machining forces and cutting tool wear during the machining is important for designing and selection of machining system and improving the productivity. This study reports the machinability of Nimonic 80A superalloy with alumina-based ceramic inserts. The objective is to analyze the reason for higher cutting forces generated during machining and tool wear mechanism on machining parameters. The cutting forces and tool wear are found to be mainly influenced by the cutting speed. The main causes of tool failure while machining Nimonic 80A are adhesion and abrasion. The role of tool wear is more dominant on the surface finish at lower cutting speed. Also, with an increase in cutting speed, thermally activated wear quietly increases at tool surfaces. The mechanistic approach is used to model the main cutting force. Developed cutting force model agrees well with experimental cutting force values.     

基于粒子群算法优化支持向量机的铝热连轧机轧制力预报

为了提高热轧带材的轧制力预报精度,提出了粒子群算法和支持向量机结合的方法来预报轧制力。根据轧制原理用支持向量机建立轧制力预报的模型,通过粒子群算法优化支持向量机参数来提高预报精度。为了进一步提高轧制力预报精度,还提出了支持向量机网络与数学模型相结合的方法,对某“1+4”铝热连轧厂现场采集的5052铝合金轧制数据进行离线仿真,仿真结果可以看出支持向量机网络与数学模型结合的方法预报轧制力,提高了轧制力预报速度并使其轧制力预报精度控制在7%以内。