基于ICA_FNN的软传感器建模过程原始特征选择

针对软传感器建模过程中辅助变量通常是多因素的混杂信号,在原始特征空间很难进行原始特征约简的问题,提出一种结合独立成分分析(ICA)和虚假最近邻点法(FNN)的原始特征选择法.利用独立成分分析法(ICA)将原始特征空间的混杂信号映射到新的独立特征子空间；然后再利用FNN计算每个原始特征剔除前后在独立特征子空间里的相似性测度,进而判断它对主导变量的影响能力,由此选择出原始特征.仿真结果表明,该方法具有优秀的原始特征选择能力.因此,该研究为选择出软传感器模型的原始特征提供了新方法.     

基于特征子空间虚假邻点判别的铝电解槽况诊断模型

造成铝电解异常槽况的因素较多,彼此相关性强,建立槽况诊断模型时计算量巨大。利用核主元分析法虽然可以对非线性数据进行降维,但得到的主元仍然是原始变量在特征空间的线性组合,既无明确的物理含义,又无法对原始特征进行约简达到减少传感器种类的目的。提出一种基于特征子空间虚假邻点判别的槽况诊断方法,首先考察各原始变量置零前后在核空间主元投影上的相似度,根据其对槽况的解释能力进行原始变量选择;再将约简后的原始变量输入概率神经网络,对各类异常槽况进行诊断。通过取自某厂170KA大型预焙槽的268组样本进行检验:在原始特征约简44.4%的情况下分类精度达到95%以上,表明该方法不但可对原始特征进行有效约简,而且槽况分类精度和训练时间均优于同类模型。     

基于KCCA虚假邻点判别的非线性变量选择

特征变量选择技术是非线性系统建模过程中降低信息冗余和提高精度的有效方法。提出一种结合核典型相关法(kernel canonical correlation analysis,KCCA)与虚假最近邻法的变量选择法。首先引入核方法,将非线性原始数据映射到线性空间,再采用典型相关法有效合理地消除因子之间的多重共线性,受混沌相空间虚假最近邻点法的启示,通过计算原始数据在KCCA子空间中投影的距离,判断其对主导变量的解释能力,由此进行变量的选择。该方法用氢氰酸生产工艺工程中的非线性模型验证,并与全参数模型进行比较,结果显示该方法有良好的变量选择能力。因此,该研究为非线性系统建模的变量选择方法提供了一种新方法。     

基于KPCA子空间虚假邻点判别的非线性建模的变量选择

特征变量选择技术是非线性系统建模过程中降低信息冗余和提高精度的有效方法。提出一种结合核主成分分析法(Kernel principal components analysis,KPCA)与虚假最近邻点法(False nearest neighbor,FNN)的变量选择法。引入核方法,将非线性原始数据映射到线性空间,再采用主成分分析法有效合理地消除因子之间的多重共线性,受混沌相空间虚假最近邻点法的启示,通过计算原始数据在KPCA子空间中投影的距离,判断其对主导变量的解释能力,由此进行变量的选择该方法用氢氰酸生产工艺工程中的非线性模型验证,并与全参数模型进行比较,结果显示该方法有良好的变量选择能力。因此,该研究为非线性系统建模的变量选择方法提供一种新方法。     

基于ReliefF算法和相关度计算结合的故障特征降维方法及其应用

在对旋转机械进行故障诊断时,通常要从时域、频域或时频域提取故障特征参数,组成原始的故障特征向量,然而在众多的故障特征当中并不是每个特征对于故障分类都是敏感且有效的。为此,本研究提出了基于ReliefF算法和相关度计算结合的故障特征降维方法。采用ReliefF加权特征选择算法对原始各特征的分类能力进行评价,选择出分类能力较强的特征;再通过特征相关度算法剔除其中分类能力相近的冗余特征,将剩余的分类能力较强的特征组成最终的降维特征向量用于故障分类和诊断,实现原始特征的降维。通过液压泵和滚动轴承的故障诊断实验,并与传统的主元分析(PCA)方法对比,结果表明该方法能够用较少的降维后的信号特征获得更高的故障正确识别率。     

基于局部平衡判别投影的旋转机械故障诊断方法

针对旋转机械故障数据的多类别、高维复杂特性导致的分类困难问题,提出一种基于局部平衡判别投影(LBDP)的故障数据集降维方法。从时域、频域和时频域多个角度提取转子振动信号的混合特征,构建原始高维故障特征集;通过LBDP选择出其中最能反映故障本质的敏感特征子集;将得到的低维特征子集输入到K近邻分类器(KNN)中进行故障模式辨识。通过一个双转子系统的振动信号集合验证了所提出方法的有效性,证明了该方法能够全面地提取出局部判别信息,使故障类别之间的差异性更清晰。     

基于KICA子空间虚假邻点判别的软传感器变量选择方法

针对软传感器建模过程中,高维冗余的非线性辅助变量造成的维度灾难问题,提出一种结合核独立成分分析法(Kernel independent component analysis,KICA)与虚假最近邻点法(False nearest neighbors,FNN)的非线性辅助变量选择方法。主要利用核函数将原始非线性数据映射到线性子空间,并采用独立成分分析消除因子之间的多重共线性,再运用虚假最近邻点法,计算原始数据在KICA子空间中投影的距离,依次判断各辅助变量对主导变量的解释能力,由此进行非线性变量选择。以某企业氢氰酸(Hydrocyanic acid,HCN)生产工艺过程中的转化率为软传感器预测目标,仿真结果表明该方法可有效降低辅助变量的维数、同时提高模型的预测精度。     

核保局鉴别分析人脸识别算法

针对人脸图像的高维、非线性特性,本文结合核函数和鉴别保局投影分析,提出了一种新的非线性子空间特征提取方法——核保局鉴别分析(kernel locality preserving discriminant analysis,KLPDA)。该算法首先利用核函数将输入空间内线性不可分的人脸样本通过非线性映射投影到高维特征空间,以增强样本的可分性,然后在该特征空间内通过保局鉴别分析算法寻求保持样本局部近邻结构的鉴别特征。实验结果指出,在ORL和UMIST人脸图像库上KLPDA的识别结果相较于其对应的线性方法(LPDA)分别高出6.16%和3.48%,而相较于其他几种典型的核方法(KPCA、KFDA和KCLPP)的最高识别率也分别高出1.28%和2.60%,验证了本文提出方法的有效性。     

PSO与LS混合算法在故障特征选择中的应用

为了提高分类器的分类精度和泛化能力,提出一种基于粒子群算法(Particle Swarm Optimization,PSO)和拉普拉斯分值(Laplace Score,LS)的混合式故障特征选择方法。该方法首先采用过滤式的特征选择方法(LS)对原始特征集进行筛选,然后利用PSO在经过精简的特征子空间里进行随机搜索,搜索过程中以支持向量机的分类准确率为适应度函数,选择出最优特征子集。用双跨转子实验台的一组故障特征数据集进行验证。实验结果表明,该方法可以有效地筛选出规模较小且最有辨别力的特征子集,能显著提高分类器的分类准确率及效率。     

基于KPCA和PSOSVM的异步电机故障诊断

针对异步电机故障振动信号具有较强的非线性特征,而传统的线性分析方法易造成振动信号非线性成分的丢失这一情况,提出一种核主元分析和粒子群支持向量机相结合的异步电机故障诊断方法。利用核函数实现输入空间到高维特征空间的非线性映射以及对映射数据的主元分析,得到原始样本的非线性主元,实现特征提取和数据压缩,将获得的核主元特征通过支持向量机进行模式识别。采用距离比值法和粒子群算法分别对核主元分析和支持向量机的参数进行双重优化选择。实验结果表明,该方法能有效提取故障信号的非线性特征,具有较强的非线性模式识别能力,相比主元分析和支持向量机方法,分类效果更好,实时性更强,可快速有效实现异步电机故障诊断。     

飞行时间质谱仪射频四极杆驱动器的研制

针对自制垂直引入式飞行时间质谱仪（orthogonal extraction time-of-flight mass spectrometer, O-TOFMS）的需要,运用直接数字频率合成 (direct digital frequency synthesis,DDS)技术,研制了射频四极杆的高压射频驱动电路装置。该驱动器频率可调范围为0.5~2 MHz,幅度最高达到1 000 Vp-p。该射频四极杆驱动器（radio frequency quadrupole driver,RFQ Driver）可用于分子离子反应器（molecule ion reactor,MIR）和RFQ驱动,其结构简洁、成本低、性能稳定可靠。     

从光电容积图中提取脉搏率、呼吸率和心率的新方法

光电容积描记图(PPG)是一种利用光学技术无创检测人体心血管脉搏波的方法。PPG信号来源于MIMIC数据库,它含有许多生理信息。本文提出了将集合经验模态分解(EEMD)、倒谱、快速傅里叶变化和过零点检测相结合的方法,从PPG中可靠地估算脉搏率(PR)、心率(HR)和呼吸频率(RR)。首先,将PPG信号通过EEMD分解为有限个固有模态函数(IMF)。因为EEMD有自适应滤波特性,所以估算不同的生理参数时,可以用不同的IMF分量来重构信号。其次,估算脉搏率时,舍去低频含有伪迹的IMF,再通过过零点检测可以得到瞬时脉搏率。然后,估算心率时,用1 Hz~1.67 Hz(60次/分钟~100次/分钟)的IMF来重构信号,再求倒谱,选取反映心脏活动的频带来得到心率。最后,估算呼吸率时,用0.05 Hz~0.75 Hz(3次/分钟~45次/分钟)的IMF来重构信号,然后对呼吸信号求快速傅里叶变化得到频谱图,寻找频谱图中的峰值得到呼吸率。对来自MIMIC数据库的53个成人PPG信号进行了仿真。仿真结果表明使用该综合信号处理方法提取的生理参数与实际生理参数一致,且该方法有运算量小,精确度高的优点(误差不超过1.17%)。     

UPLC-MS/MS测定二苯甲烷二异氰酸酯

建立了二苯甲烷二异氰酸酯（MDI）的超高效液相色谱-串联质谱（UPLC-MS/MS）联用的分析方法。实验用甲醇作衍生试剂,将二苯甲烷二异氰酸酯衍生成二苯甲烷二氨基甲酸甲酯（MDC）,通过测定二苯甲烷二氨基甲酸甲酯来确定二苯甲烷二异氰酸酯的含量。实验采用Waters Acquity BEH C18超高效液相色谱柱,甲醇-0.1%甲酸作为梯度洗脱液,二苯甲烷二氨基甲酸甲酯在1.5 min内与其他化合物进行完全分离,经四极杆质谱选择离子监测模式检测。线性范围为2~100 μg/L,检出限为1 μg/L,相关系数0.999 4。考察了水、甲酸、甲酸铵流动相体系,以及进样量大小对样品的分离效果,发现甲酸流动相体系的分离效果优于其他两种流动相体系,同时发现在进样量大时出现平头峰。并且在优化条件下,对实际样品中游离的二苯甲烷二异氰酸酯进行了测定。     

Effect of the molecular weight on deformation states of the polystyrene film by AFM single scanning

Nanobundles patterns can be formed on the surface of most thermoplastic polymers when the atomic force microscope (AFM)‐based nanomechanical machining method is employed to scratch their surfaces. Such patterns are reviewed as three‐dimensional sine‐wave structures. In the present study, the single‐line scratch test is used firstly to study different removal states of the polystyrene (PS) polymer with different molecular weights (MWs). Effects of the scratching direction and the scratching velocity on deformation of the PS film and the state of the removed materials are also investigated. Single‐wear box test is then employed to study the possibility of forming bundle structures on PS films with different MWs. The experimental results show that the state between the tip and the sample plays a key role in the nano machining process. If the contact radius between the AFM tip and the polymer surface is larger than the chain end‐to‐end distance, it is designated as the “cutting” state that means the area of both side ridges is less than the area of the groove and materials are removed. If the contact radius is less than the chain end‐to‐end distance, it is designated as the “plowing” state that means the area of both side ridges is larger than the area of the groove and no materials are removed at all. For the perfect bundles formation on the PS film, the plowing state is ideal condition for the larger MW polymers because of the chains’ entanglement. SCANNING 35:308‐315, 2013. © 2012 Wiley Periodicals, Inc.     

Rapid one-of-a-kind production

To shorten the product development cycle time and reduce the cost at the same time, which is very critical for the survival of those small and mid-sized one-of-a-kind production (SMOKP) companies, various computer communication, simulation and computer aided engineering and management techniques have been widely used in these SMOKP companies. However, according to our study, the applications of these advanced computer technologies in these SMOKP companies are lack of system integration and synergy, which result in communication errors, reworks, duplications and hence a longer product development lead time and a higher cost. To solve this problem, an inter/intranet based computer aided rapid product development system is presented in this paper. This system consists of a number of inter/intranet based computer models or sub-systems for supporting rapid development of a one-of-a-kind product. It also employs a novel concurrent product development strategy called “prototype based incremental product development strategy” and an integrated product data structure called “product production structure”. Some industrial implementations of the proposed system, product development strategy and product data structure are also reported in the paper.     

Rapid one-of-a-kind production

To shorten the product development cycle time and reduce the cost at the same time, which is very critical for the survival of those small and mid-sized one-of-a-kind production (SMOKP) companies, various computer communication, simulation and computer aided engineering and management techniques have been widely used in these SMOKP companies. However, according to our study, the applications of these advanced computer technologies in these SMOKP companies are lack of system integration and synergy, which result in communication errors, reworks, duplications and hence a longer product development lead time and a higher cost. To solve this problem, an inter/intranet based computer aided rapid product development system is presented in this paper. This system consists of a number of inter/intranet based computer models or subsystems for supporting rapid development of a one-of-a-kind product. It also employs a novel concurrent product development strategy called “prototype based incremental product development strategy” and an integrated product data structure called “product production structure”. Some industrial implementations of the proposed system, product development strategy and product data structure are also reported in the paper.     

汽车ESC系统DIL/HIL仿真平台设计

全面准确地标定与测试是开发车辆电子稳定控制(electronic stability control,ESC)系统的关键环节。车辆控制过程是典型的强非线性系统,建立精确的车辆模型比较困难,驾驶员模型也很难准确模拟紧急工况下驾驶员的真实反应。在仿真系统中将难以精确建模的液压制动系统、ESC压力调节单元、转向及电子节气门装置采用真实硬件,并在试验平台中嵌入真实驾驶员,基于Matlab/Simulink/dSPACE环境和车辆动力学软件veDYNA,开发硬件在环(hardware-in-the-loop,HIL)和驾驶员在环(driver-in-the-loop,DIL)的仿真试验平台,并在此平台上对所设计的ESC系统进行仿真试验研究。试验结果表明,所设计的DIL/HIL仿真平台可以大幅提高ESC系统的开发效率和测试准确性。     

The efficacy of back propagation neural network with delta bar delta learning in predicting the wear of carbide inserts in face milling

Face milling is a process predominantly affected by dynamic variation of cutting forces, thermo-mechanical shocks and vibration leading to catastrophic tool failure along with gradual wear of the inserts. Keeping in view the industrial importance of this process, it is necessary to devise suitable methods to predict in advance the onset of tool failure without grossly impairing the machining set-up and the job. Hence, the applicability of back propagation neural network with delta bar delta learning rule for faster convergence has been studied with the above objective. The multi sensor based tool condition monitoring strategy shows that the learning rate adaptation scheme combined with the selection of suitable process parameters drastically reduces the training time of the artificial neural network without dispensing with the prediction accuracy.     

复合ICA-SVM机械状态模式分类

提出了一种新颖的、基于独立分量分析(ICA)的复合神经网络，用于不同机械状态模式的特征提取。利用支持向量机(SVM)进行最终分类。与通常的基于经验风险最小化(ERM)原理的神经网络方法相比，基于结构风险最小化(SRM)原理的支持向量机分类方法具有更好的推广能力。而借助多个独立分量分析网络，隐藏于多通道振动观测信号中的不变特征得到有效提取，从而实现了支持向量机分类器在分类能力和推广性两者间的合理平衡。     

同位素稀释质谱法测定贻贝中多氯联苯的国际比对研究(CCQM-P67)

A method developed to participate the pilot study of international comparison CCQM-P67 was introduced. Isotope dilution mass spectrometry (IDMS) was used in the detection of five PCB congeners in freeze-dried muscle tissue. Soxhlet Extraction is used for extraction of mussel, duplicate clean-up of GPS and SPE column were used, C-13 labeled PCB congeners were added as internal standard to avoid the error from sample preparation and diction. High resolution Mass spectrometer was used to give good result of detection.