Multiphase batch process monitoring based on higher-order cumulant analysis

In this paper, a two-step phase partitioning strategy is proposed. Firstly, the number of phases is automatically determined according to the intra-class and inter-class similarity of feature space data, thus avoiding excessive manual intervention. Secondly, the phases are partitioned by step-wise adding the kernel entropy extended load matrix (KEELM), avoiding the wrong division of phases caused by unstable state of working condition conversion. A process monitoring model based on multiway kernel entropy independent component analysis (MKEICA) is constructed in each sub-phase to deal with complex batch processes with nonlinear and non-Gaussian properties. A new statistics index based on the idea of high order cumulant analysis (HCA) is constructed in each sub-phase for process monitoring. Compared with the traditional second-order statistics, it can obtain high-order statistical information. Finally, the proposed method is applied to the penicillin simulation platform process and compared with the traditional multiway kernel independent components analysis (MKICA) and HCA methods to verify the effectiveness of the method that is mentioned above.     

基于判别核窗宽的掌纹识别方法

提出了一种新的判别 核窗宽方法,进而研究了基于判别核窗宽的KPCA和LPP在掌纹识别中的应用。首先根据训练 样本和类标签计 算类内核窗宽和类间核窗宽;在分类密集区选择较小窗宽,在分类稀疏区选择较大窗宽,可 以有效提取数 据的关联特征;然后运用基于判别核窗宽的KPCA和LPP方法提取低维特征向量,计算特征向 量间的余弦 距离进行掌纹匹配;最后运用PolyU掌纹图像库,对本文算法进行测试。实验结果表明,与 传统算法相比, 本文算法的识别率最高,识别时间小于0.6s,验证了方法的有效性 。     

基于支持向量机的高炉向凉、向热炉况预测

高炉冶炼过程中炉温是影响技术经济指标的关键参数,保持合理的炉温是高炉稳定顺行的关键因素。采用某炼铁厂在线采集的数据,通过核主元分析对建模数据进行预处理,根据相关系数选定模型参数,确定参数对炉温的滞后时间,基于支持向量机建立了高炉向凉、向热预测诊断模型。通过实例验证,该模型具有很高的精度。     

基于核主元分析的净化除钴过程数据预处理研究

在净化除钴过程中,复杂的现场生产使工艺参数具有高维性、强耦合性、非线性等特点,对除钴效果分析以及过程建模产生较大影响。针对该问题,本文通过对现场工艺参数的特性分析,提出了一种基于核主元分析（KPCA）的数据预处理方法,消除噪音并有效提取工业数据信息。采用工业生产数据对该方法进行对比仿真实验,结果表明,基于KPCA的数据预处理方法改善了模型的精度,对净化除钴优化控制具有重要意义。     

改进的RBF文本分类算法

分析了文本分类系统的一般模型及现有技术,在应用了核主成分分析的特征降维方法进行处理后,提出了一种基于样本中心的径向基( RBF)神经网络文本分类算法,并且引入了聚类算法的核心思想,来改进误差反向传播(BP)神经网络分类算法收敛速度较慢的缺点.实验结果表明, RBF网络与BP网络相比,具有较高的运算速度和较强的非线性映射能力,在收敛速度和准确程度上也有更好的分类效果.     

转子故障特征数据分类的KPCA-BFDA方法

对非线性转子系统故障特征数据的分类方法进行了研究.在提出一种偏费歇判别分析法(biased fisher discriminatory analysis,简称BFDA)的前提下,进一步提出将核主成分分析法(kernel principal component analysis,简称KPCA)与偏费歇判别分析法相结合的数据集降维方法,该方法中的核主成分分析步骤用于构造剔除数据集冗余信息的降维数据集,偏费歇判别分析步骤用于进一步降低数据集维数并提高不同类别数据子集间的分离程度.对实例数据与典型故障数据的分类结果表明,提出的偏费歇判别分析法在具备费歇判别分析降维可分性能的基础上,具有更低计算复杂度的特点.提出的核主成分分析结合偏费歇判别分析的算法,其对应的降维结果能直接应用于线性分类器,且取得了较好的分类效果.     

基于核主元分析与神经网络的传感器故障诊断新方法

提出综合利用核函数主元分析(KPCA)和神经网络的方法实现非线性系统内传感器故障的检测和识别,克服了以往核函数主元分析法只能给出故障检测结果,却无法对故障进行识别的缺陷,并给出了在特征空间中计算残差的简单方法.最后,对一个简单的多变量过程进行了故障检测和识别,验证了这一诊断策略的有效性.     

基于贡献率法的非线性工业过程在线故障诊断（英文）

Over past decades, kernel principal component analysis(KPCA) appeared quite popularly in data-driven process monitoring area. Enormous work has been done to show its simplicity, feasibility, and effectiveness. However, the introduction of kernel trick makes it impossible to directly employ traditional contribution plots for fault diagnosis. In this paper, on the basis of revisiting and analyzing the existing KPCA-relevant diagnosis approaches, a new contribution rate based method is proposed which can explain the faulty variables clearly. Furthermore, a scheme for online nonlinear diagnosis is established. In the end, a case study on continuous stirred tank reactor(CSTR) benchmark is applied to access the effectiveness of the new methodology, where the comparisons with the traditional linear method are involved as well.     

一种基于KPCA-LSSVM的可用带宽在线预测算法

针对目前端到端可用带宽预测方面研究工作较少的现状,提出一种基于核主成分分析KPCA(Kernel Principle Component Analysis)和最小二乘支持向量机LSSVM(Least Squares Support Vector Machine)的可用带宽在线预测算法ABOP。在采集网络状态样本数据并对其进行相空间重构的基础上,采用KPCA对数据进行降维降噪处理,最后基于LSSVM对可用带宽进行在线预测。为减小计算开销,提出一种递推计算的方法加快模型更新速度,并采用粒子群优化算法对模型参数进行多步更新,确保了在线预测的时效性。仿真表明,提出的ABOP算法具有较高的预测精度和较快的预测速度,能够满足可用带宽在线预测的要求。     

Mine-hoist fault-condition detection based on the wavelet packet transform and kernel PCA

A new algorithm was developed to correctly identify fault conditions and accurately monitor fault development in a mine hoist. The new method is based on the Wavelet Packet Transform (WPT) and kernel PCA (Kernel Principal Component Analysis, KPCA). For non-linear monitoring systems the key to fault detection is the extracting of main features. The wavelet packet transform is a novel technique of signal processing that possesses excellent characteristics of time-frequency localization. It is suitable for analysing time-varying or transient signals. KPCA maps the original input features into a higher dimension feature space through a non-linear mapping. The principal components are then found in the higher dimension feature space. The KPCA transformation was applied to extracting the main nonlinear features from experimental fault feature data after wavelet packet transformation. The results show that the proposed method affords credible fault detection and identification.