采用遗传相关向量机的工业机器人焊缝跟踪预测

为了克服传统焊缝跟踪方法精度低的问题,提出基于遗传相关向量机(G-RVM)的工业机器人焊缝跟踪预测方法,其中应用遗传算法对相关向量机参数进行优化。相关向量机通过构建回归函数以解决焊缝跟踪问题。为验证所设计控制器的有效性,进行了焊缝的跟踪实验,并设计了实验条件。实验结果表明,基于遗传相关向量机的焊缝跟踪误差小于支持向量机法所得数据。可见采用遗传相关向量机的控制更能够适应实际焊接过程的变化。     

基于改进LSSVM算法的数控机床热误差建模研究

为了提高数控机床热误差的预测精度,提出了基于改进最小二乘支持向量机和预测误差校正相结合的方法。首先引入提升小波分解原始数据克服数控机床误差的耦合性;然后采用最小二乘支持向量机对分解后的信号做预测,接着用误差校正方式修正预测结果;最后通过实验数据对比分析得到,基于提升小波的最小二乘支持向量机(LWT-LSSVM)法比最小二乘支持向量机(LSSVM)法的建模预测精度高8.51%,证明此建模方法有效可行。     

基于LSSVM的直线电机进给定位误差补偿研究

误差补偿是提高机械加工精度的有效途径,在分析测试数控直线电机进给定位精度的基础上,提出基于最小二乘支持向量机的直线电机进给定位精度误差回归建模和预测方法.最小二乘支持向量机具有完备的统计学习理论基础和学习功能,它用核函数建立预测模型,再用已知数据为学习样本训练学习机,用检验样本进行验证、预测系统未来误差.采用径向基核函数的最小二乘支持向量机对不同速度加速度下的定位误差进行了预测,并进行误差补偿.研究结果表明,采用LSSVM的方法可以较大地提高直线电机进给的定位精度.     

基于LSSVM和GA的镍铬合金薄膜制备工艺参数优化

为提高磁控溅射镍铬合金薄膜的性能和确定最佳工艺参数条件,建立了一种粒子群-最小二乘支持向量机(PSO-LSSVM)和遗传算法(GA)联合优化算法。以功率、压强、基底温度及偏压为因素设计表面粗糙度和沉积速率的正交试验;采用极差分析法得到工艺参数组合,并分析了该方法存在的不足;通过最小二乘支持向量机分别建立对表面粗糙度和沉积速率的预测模型。为提高预测模型的精度,提出了高斯核函数与多项式核函数的加权混合核函数,采用粒子群算法对权重因子、惩罚因子、高斯核带宽以及多项式阶次进行多目标寻优,得到LSSVM非线性回归预测函数。将回归预测函数作为遗传算法的适应度函数,并确定四种相关因素的约束条件,分别得到了两种性能的最佳工艺参数。     

基于PSO-LSSVM的网络化制造系统安全检测

为了解决传统网络化制造系统安全检测技术检测精度低的难题,提出基于粒子群优化最小二乘支持向量机(PSO.LSSVM)的网络化制造系统安全检测方法.首先,确定网络化制造系统安全检测特征,并对特征进行预处理.然后,建立基于粒子群优化最小二乘支持向量机的网络化制造系统安全检测模型.最后,通过实例证明该方法的有效性及优越性.分别采用人工神经网络、支持向量机与PSO-LSSVM方法进行对比分析,实验结果表明,PSO-LSSVM对网络化制造系统的安全检测性能优于神经网络与支持向量机的安全检测性能.     

相对变换主元分析的变压器油击穿电压预测

针对变压器油击穿电压在线测量困难,数据噪声和孤立点影响支持向量机预测精度的问题,提出了基于相对变换(RT)主元分析(PCA)的变压器油击穿电压预测方法。首先,通过相对变换将原始数据空间变换到相对空间,抑制噪声和孤立点对模型精度的影响;然后在相对空间进行主元分析,降低相对空间维数,使提取的主元特征更具有代表性和更大的变化度;最后,将提取的主元作为最小二乘支持向量机(LSSVM)的输入,建立变压器油击穿电压的最小二乘支持向量机预测模型。与LSSVM、RT-LSSVM和PCA-LSSVM的对比实验结果表明,本文提出的方法具有较优的预测精度和泛化能力。     

基于提升小波和递推LSSVM的实时故障诊断方法

提出了一种基于提升小波(LW)与递推最小二乘支持向量机(RISSVM)相集合的实时故障诊断方法(LW-RLSSVM).该方法首先通过提升小波变换对数据实时去噪,再通过实时算法训练最小二乘支持向量机分类器.由于采用了递推算法,节省了存储空间和运算时间,同时增加了诊断模型的适应性.为验证所提方法的有效性,将LW-RLSSVM应用于TE过程和青霉素发酵过程.实验结果表明,LW-RLSSVM集合方法能有效实现实时故障诊断,在诊断速度和适应性方面,优于基于第一代小波与LSSVM相集合(W-LSSVM)的故障诊断方法;在诊断精度等方面,该方法优于LSSVM、RLSSVM等方法.     

基于PSO-LSSVM模型的扩散硅压力传感器的温度补偿

针对扩散硅压力传感器温度漂移的问题,文中提出了一种基于粒子群算法优化最小二乘支持向量机(PSO-LSSVM)的温度补偿模型。通过对扩散硅压力传感器做二维标定实验,利用AD590集成温度传感器监测实验环境温度,建立PSO-LSSVM模型。最小二乘支持向量机的惩罚因子和核函数的选取会直接影响到模型的预测精度,PSO-LSSVM模型利用粒子群算法优化最小二乘支持向量机模型的惩罚因子和核函数的参数,改善了传统的最小二乘支持向量机模型对参数选取耗时耗力且未必找到全局最优解的缺陷。实验结果表明,经该模型补偿后的零点温度系数和灵敏度温度系数都减小了一个数量级,且预测值与标定值的均方误差的数量级达到10~(-6),实现了温度补偿并改善了预测精度。     

基于能量诱导型PSO算法与LSSVM模型的钢铁企业高炉煤气消耗量预测

针对钢铁企业高炉煤气消耗量存在的波动大、随机性强、难以预测等特点,引入能量剩余函数,提出了一种与粒子自身能量相关的能量诱导型粒子群(Energy Guided Particle Swarm Optimization,EGPSO)算法。利用其对最小二乘支持向量机(Least Square Support Vector Machine,LSSVM)的参数进行优化,最后采用优化后的最小二乘支持向量机模型(EGPSO-LSSVM)进行高炉煤气消耗量预测。仿真实验表明:改进后的预测模型在平均绝对百分比误差、均方误差、均方百分比误差三项指标上均优于普通BP神经网络模型和普通最小二乘支持向量机模型,可以为高炉煤气资源的合理使用提供依据。     

基于递归神经网络的LS-SVM硬件实现与实验研究

在标准支持向量机(SVM)学习神经网络的基础上,将最小二乘支持向量机(LS-SVM)与递归神经网络相结合, 提出一种新的最小二乘支持向量机学习神经网络.该网络直接采用Lagrange乘子进行训练,消除了标准SVM神经网络中的线性部分,可用于进行分类和回归学习.并且其拓扑结构更适合于用简单的硬件模拟电路实现.对两种网络的稳定性进行了证明,并设计了相应的硬件电路,最后通过Simulink、Pspice仿真和硬件电路实验证明了所提出的方法是有效的.     

基于最小二乘支持向量机的N型热电偶非线性校正及应用

本文针对具有小样本数据的N型热电偶在应用中存在的问题，提出了一种基于最小二乘支持向量机对热电偶进行非线性校正的方法，并与以往采用的BP网络、RBF网络和ANFIS校正方法进行了比较。结果表明，采用最小二乘支持向量机的校正精度高于以上3种校正方法；同时以阳极焙烧过程中料箱温度为对象进行了仿真和实际应用研究，取得了满意的结果。     

激光跟踪仪测角误差的现场评价

激光跟踪仪是基于角度传感和测长技术相结合的球坐标测量系统,其长度测量采用激光干涉测长方法,可直接溯源至激光波长,因此,激光跟踪仪的长度测量精度远高于角度测量精度,相对而言,测角误差就成为评价跟踪仪测量精度的重要指标。为了对现场测量激光跟踪仪的测角误差进行快速有效地评价,采用跟踪仪多站位对空间中测量区域内若干个被测点进行测量,与传统基于角度交汇原理的多站位冗余测量不同,利用各站位所观测的高精度测长值建立误差方程,并通过测长方向的矢量位移对跟踪仪测长误差进行约束,获得被测点三维坐标在跟踪仪水平角和垂直角方向上的改正值,以此来评价激光跟踪仪的测角误差。通过Leica激光跟踪仪AT901-LR进行了多站位测角误差评价实验,在现场测量条件下,跟踪仪水平和垂直方向测角误差约为0.003 mm/m(1σ),符合跟踪仪的测量误差特性。     

光束平差在激光跟踪仪系统精度评定中的应用

对自主研制的激光跟踪仪的精度评定进行研究,以期解决大尺寸空间坐标测量系统的空间坐标精度难于评定的问题.考虑现场环境条件、仪器状态和操作者技能等因素对测量精度影响都很大,提出了基于光束平差原理对激光跟踪仪系统进行精度评定的方法.通过Matlab软件对激光跟踪仪的精度评定进行了仿真,仿真结果显示光束平差法能客观地反映激光跟踪仪的测量精度.另外,使用Faro生产的激光跟踪仪进行了实物实验,实验结果显示其水平角精度σH为1.97″,垂直角精度σV为2.61″,测距精度σD为3.75×10-6,对比Faro生产的激光跟踪仪精度(σH =2.0″;σV =2.0″;σD=4 μm)可证明采用光束平差法评定自主研发的激光跟踪仪测量精度是正确、可行的.该方法为探索激光跟踪仪新的应用技术、开展面向对象的测量不确定评定奠定了基础.     

Influence of measurement noise and laser arrangement on measurement uncertainty of laser tracker multilateration in machine tool volumetric verification

This paper aims to present different techniques and factors that affect the measurement accuracy of a commercial laser tracker responsible for capturing checkpoints used in machine tool volumetric verification. This study was conducted to uncover various sources of error affecting the measurement uncertainty of the laser tracker, additional sources of error that further contributed to the uncertainty, and the factors influencing these techniques. We also define several noise reduction techniques for the measurements.The improvement in the accuracy of captured points focuses on a multilateration technique and its various resolution methods both analytically and geometrically. Similarly, we present trilateration and least squares techniques that can be used for laser tracker self-calibration, which is an essential parameter in multilateration.This paper presents the influence of the spatial distribution of laser trackers (LTs) in measurement noise reduction by multilateration, which produces an improvement in volumetric error machine tool reduction. A study of the spatial angle between LTs, the distance and the visibility of the point to be measured are presented using a synthetic test. All of these factors limit the scope of multilateration. Similarly, a comparison of self-calibration techniques using the least squares and trilateration methods with which to determine the relative position of the laser tracker employees is presented. We also present the influence of the relationship between the radial and angular measurement noise self-calibration processes as it relates to the volumetric error reduction achieved by the machine tool with multilateration. All studies were performed using synthetic tests generated using a synthetic data parametric generator.     

The detection of rotary axis of NC machine tool based on multi-station and time-sharing measurement

At present, the detection of rotary axis is a difficult problem in the errors measurement of NC machine tool. In the paper, a method with laser tracker on the basis of multi-station and time-sharing measurement principle is proposed, and this method can rapidly and accurately detect the rotary axis. Taking the turntable measurement for example, the motion of turntable is measured by laser tracker at different base stations. The redundant equations can be established based on the large amount of measured data concerning the distance or distance variation between measuring point and base station. The coordinates of each measuring point during turntable rotation can be accurately determined by solving the equations with least square method. Then according to the error model of rotary axis, the motion error equations of each measuring point can be established, and each error of turntable can be identified. The algorithm of multi-station and time-sharing measurement is derived, and the error separation algorithm is also deduced and proved feasible by simulations. Results of experiment show that a laser tracker completes the accuracy detection of the turntable of gear grinding machine within 3 h, and each error of the turntable are identified. The simulations and experiments have verified the feasibility and accuracy of this method, and the method can satisfy the rapid and accurate detecting requirements for rotary axis of multi-axis NC machine tool.     

基于多嵌入维数的时用水量LSSVM组合预测

为解决供水系统调度所需混沌时用水量高精度预测等问题,将最小二乘支持向量机(LSSVM)组合预测模型应用到城市时用水量预测中。在分析不同嵌入维数和预测方法对模型预测精度影响程度的基础上,提出了基于多嵌入维数的LSSVM组合预测模型。采用互信息法和G-P方法求取多个嵌入维数,并建立了不同相空间模型,通过LSSVM算法对上述多个预测模型进行了组合预测,既综合了各不同嵌入维数各预测方法下的信息,又对单一模型下的预测偏差进行了融合,以有效地提高预测精度;最后在某地进行了时用水量序列的仿真实验。研究结果表明,该模型预测精度平均误差小于2%,明显优于各单一模型的预测结果,证实了该组合模型的有效性和实用性。     

Modelling,kinematic parameter identification and sensitivity analysis of a Laser Tracker having the beam source in the rotating head

This paper presents a new kinematic model, a parameter identification procedure and a sensitivity analysis of a laser tracker having the beam source in the rotating head. This model obtains the kinematic parameters by the coordinate transformation between successive reference systems following the Denavit–Hartenberg method. One of the disadvantages of laser tracker systems is that the end-user cannot know when the laser tracker is working in a suitable way or when it needs an error correction. The ASME B89.4.19 Standard provides some ranging tests to evaluate the laser tracker performance but these tests take a lot of time and require specialized equipment. Another problem is that the end-user cannot apply the manufacturer’s model because he cannot measure physical errors. In this paper, first the laser tracker kinematic model has been developed and validated with a generator of synthetic measurements using different meshes with synthetic reflector coordinates and known error parameters. Second, the laser tracker has been calibrated with experimental data using the measurements obtained by a coordinate measuring machine as nominal values for different strategies, increasing considerably the laser tracker accuracy. Finally, a sensitivity analysis of the length measurement system tests is presented to recommend the more suitable positions to perform the calibration procedure.     

基于LSSVM的大型造船厂24 h用气流量预测建模

针对大型造船厂气动系统的用气特点,建立了基于LSSVM 24 h用气流量预测模型。模型以上海某造船厂每天实际用气流量为样本,根据实际预测控制需要进行预测。仿真结果表明,基于LSSVM 的24 h用气流量预测模型能够实时反映系统当前状态,预测结果基本上吻合实际用气量,能够满足大型离心空压机组启、停控制需求。