Nonlinear process monitoring using kernel principal component analysis

In this paper, a new nonlinear process monitoring technique based on kernel principal component analysis (KPCA) is developed. KPCA has emerged in recent years as a promising method for tackling nonlinear systems. KPCA can efficiently compute principal components in high-dimensional feature spaces by means of integral operators and nonlinear kernel functions. The basic idea of KPCA is to first map the input space into a feature space via nonlinear mapping and then to compute the principal components in that feature space. In comparison to other nonlinear principal component analysis (PCA) techniques, KPCA requires only the solution of an eigenvalue problem and does not entail any nonlinear optimization. In addition, the number of principal components need not be specified prior to modeling. In this paper, a simple approach to calculating the squared prediction error (SPE) in the feature space is also suggested. Based on T² and SPE charts in the feature space, KPCA was applied to fault detection in two example systems: a simple multivariate process and the simulation benchmark of the biological wastewater treatment process. The proposed approach effectively captured the nonlinear relationship in the process variables and showed superior process monitoring performance compared to linear PCA.     

Learning a data-dependent kernel function for KPCA-based nonlinear process monitoring

Kernel principal component analysis (KPCA)-based process monitoring methods have recently shown to be very effective for monitoring nonlinear processes. However, their performances largely depend on the kernel function and currently there is no general rule for kernel selection. Existing methods simply choose the kernel function empirically or experimentally from a given set of candidates. This paper proposes a kernel function learning method for KPCA to learn a kernel function tailored to specific data and explores its potential for KPCA-based process monitoring. Motivated by the manifold learning method maximum variance unfolding (MVU), we obtain the kernel function by optimizing over a family of data-dependent kernels such that the nonlinear structure in input data is unfolded in the kernel feature space and gets more likely to be linear there. Using the optimized kernel, the nonlinear principal components of KPCA which are linear principal components in the kernel feature space can effectively capture the variation in data, and thus the data under normal operating conditions can be more precisely modeled by KPCA for process monitoring. Simulation results on an simple nonlinear system and the benchmark Tennessee Eastman (TE) demonstrate that the optimized kernel functions lead to significant improvement in the performance over the popular Gaussian kernels when used in the KPCA-based process monitoring.     

基于稀疏核主元分析的在线非线性过程监控

核主元分析(KPCA)适合非线性过程的监控,但存在计算量大、实时性差等缺点。提出一种基于稀疏KPCA(SKPCA)的过程监控方法,先使用SKPCA对正常建模数据进行加权,少数权值大的数据基本能代表全部正常数据的信息,因此稀化了建模数据,然后根据稀化后的正常数据建立过程的KPCA模型,并提出监控指标,大大减少了计算量,提高了监控的实时性,最后以化工分离过程为对象,就KPCA与SKPCA的监控效果和实时性进行了详细的对比研究,结果表明了基于SKPCA监控方法的优越性。     

递归核PCA及其在非线性过程自适应监控中的应用

PCA、PLS作为常用的多变量统计监控算法,一般适用于线性、定常的过程。针对实际工业过程的时变、非线性特性,提出了一种递归核PCA（RKPCA）方法用于非线性过程的自适应监控。RKPCA算法通过将递归奇异值分解推广到核空间,给出了核形式描述的递归KPCA算法,运算复杂度比KPCA明显降低,保证非线性监控模型能够在线更新。在Alstom工业燃气发生装置上的自适应监控表明,所提出的RKPCA算法能够及时跟踪非线性过程的时变特征,保证了监控模型的有效性。     

Fault identification for process monitoring using kernel principal component analysis

In this research, we develop a new fault identification method for kernel principal component analysis (kernel PCA). Although it has been proved that kernel PCA is superior to linear PCA for fault detection, the fault identification method theoretically derived from the kernel PCA has not been found anywhere. Using the gradient of kernel function, we define two new statistics which represent the contribution of each variable to the monitoring statistics, Hotelling's T²and squared prediction error (SPE) of kernel PCA, respectively. The proposed statistics which have similar concept to contributions in linear PCA are directly derived from the mathematical formulation of kernel PCA and thus they are straightforward to understand. The main contribution of this work is that we firstly suggest a fault identification method especially applicable to process monitoring using kernel PCA. To demonstrate the performance, the proposed method is applied to two simulated processes, one is a simple nonlinear process and the other is a non-isothermal CSTR process. The simulation results show that the proposed method effectively identifies the source of various types of faults.     

Improved dynamic kernel PCA based on local preserving projections and its application for electric submersible pump fault diagnosis

Dynamic kernel principal component analysis (DKPCA) has been frequently implemented for nonlinear and dynamic process monitoring of complex industrial processes. However, traditional DKPCA focuses only on the global structural analysis of data sets and strongly neglects the local information, which is equally essential for process detection and identification. In this paper, an improved DKPCA, referred to as the local DKPCA (LDKPCA), is proposed based on local preserving projections (LPP) for nonlinear dynamic process fault diagnosis. The method combines the advantages of LPP and DKPCA by utilizing the local structure feature to maintain the geometric structure of the data in a unified framework. To achieve a highly comprehensive feature extraction, the local characteristics are fused in DKPCA to produce an optimization objective. The neighbouring points of the new objective function projection in the feature space are still maintained in proximity, and the variance information is retained simultaneously. For the purpose of fault detection, two statistics, known as the T² and squared prediction error (SPE) statistics, are constructed, based on the LDKPCA model, and used to monitor the latent variable space and the residual space, respectively. In addition, the sensitivity analysis is brought in for fault identification of the two statistics. Based on the experimental analysis using the shaft breakage data of an offshore oilfield electric submersible pump (ESP), the proposed method outperforms the conventional DKPCA in terms of fault monitoring performance. The experimental results demonstrate the potential of the method in nonlinear dynamic process fault diagnosis.     

On-line batch process monitoring using dynamic PCA and dynamic PLS models

Producing value-added products of high-quality is the common objective in industries. This objective is more difficult to achieve in batch processes whose key quality measurements are not available on-line. In order to reduce the variations of the product quality, an on-line batch monitoring scheme is developed based on the multivariate statistical process control. It suggests using the past measured process variables without real-time quality measurement at the end of the batch run. The method, referred to as BDPCA and BDPLS, integrates the time-lagged windows of process dynamic behavior with the principal component analysis and partial least square respectively for on-line batch monitoring. Like traditional MPCA and MPLS approaches, the only information needed to set up the control chart is the historical data collected from the past successful batches. This leads to simple monitoring charts, easy tracking of the progress in each batch run and monitoring the occurrence of the observable upsets. BDPCA and BDPLS models only collect the previous data during the batch run without expensive computations to anticipate the future measurements. Three examples are used to investigate the potential application of the proposed method and make a comparison with some traditional on-line MPCA and MPLS algorithms.     

基于互信息的PCA方法及其在过程监测中的应用

主元分析(PCA)是一种经典的特征提取方法,已被广泛用于多变量统计过程监测,其算法的本质在于提取过程数据各变量之间的相关性。然而,传统PCA算法中定义的相关性矩阵局限于计算变量间的线性关系,无法衡量两个变量间相互依赖的强弱程度。为此,提出一种新的基于互信息的PCA方法(MIPCA)并将之应用于过程监测。与传统PCA所不同的是,MIPCA通过计算两两变量间的互信息来定义相关性,将原始相关性矩阵取而代之为互信息矩阵,并利用该互信息矩阵的特征向量实现对过程数据的特征提取。在此基础上,可以建立相应的统计监测模型。最后,通过实例验证MIPCA用于过程监测的可行性和有效性。     

Improved process monitoring using the CUSUM and EWMA-based multiscale PCA fault detection framework

Process monitoring techniques are of paramount importance in the chemical industry to improve both the product quality and plant safety. Small or incipient irregularities may lead to severe degradation in complex chemical processes, and the conventional process monitoring techniques cannot detect these irregularities. In this study to improve the performance of monitoring, an online multiscale fault detection approach is proposed by integrating multiscale principal component analysis (MSPCA) with cumulative sum (CUSUM) and exponentially weighted moving average (EWMA) control charts. The new Hotelling's T² and square prediction error (SPE) based fault detection indices are proposed to detect the incipient irregularities in the process data. The performance of the proposed fault detection methods was tested for simulated data obtained from the CSTR system and compared to that of conventional PCA and MSPCA based methods. The results demonstrate that the proposed EWMA based MSPCA fault detection method was successful in detecting the faults. Moreover, a comparative study shows that the SPE-EWMA monitoring index exhibits a better performance with lower values of missed detections ranging from 0% to 0.80% and false alarms ranging from 0% to 21.20%.     

Statistical monitoring of dynamic processes based on dynamic independent component analysis

Most multivariate statistical monitoring methods based on principal component analysis (PCA) assume implicitly that the observations at one time are statistically independent of observations at past time and the latent variables follow a Gaussian distribution. However, in real chemical and biological processes, these assumptions are invalid because of their dynamic and nonlinear characteristics. Therefore, monitoring charts based on conventional PCA tend to show many false alarms and bad detectability. In this paper, a new statistical process monitoring method using dynamic independent component analysis (DICA) is proposed to overcome these disadvantages. ICA is a recently developed technique for revealing hidden factors that underlies sets of measurements followed on a non-Gaussian distribution. Its goal is to decompose a set of multivariate data into a base of statistically independent components without a loss of information. The proposed DICA monitoring method is applying ICA to the augmenting matrix with time-lagged variables. DICA can show more powerful monitoring performance in the case of a dynamic process since it can extract source signals which are independent of the auto- and cross-correlation of variables. It is applied to fault detection in both a simple multivariate dynamic process and the Tennessee Eastman process. The simulation results clearly show that the method effectively detects faults in a multivariate dynamic process.     

基于核Fisher判别分析方法的非线性统计过程监控与故障诊断

化工过程中大量的生产数据反应了生产过程的内在变化和系统的运行状况,基于数据驱动的统计方法可以有效地对生产过程进行监控。对于复杂的化工和生化过程,其过程变量之间的相关关系往往具有很强的非线性特性,传统的线性统计过程监控方法显得无能为力。本文提出了基于核Fisher判别分析的非线性统计过程监控方法,首先利用非线性核函数将数据从原始空间映射到高维空间,在高维空间中利用线性的Fisher判别分析方法提取数据最优的Fisher特征矢量和判别矢量来实现过程监控与故障诊断,能有效地捕获过程变量之间的非线性关系,通过对流化催化裂化（FCCU）过程的仿真表明该方法的有效性。     

基于核独立元分析的间歇过程在线监控

针对间歇过程独特的数据特点,提出了一种基于核独立元分析(kernelICA)的局部在线建模监控方法。核独立元分析通过规范相关性将比较函数扩展到一个再生的核希尔伯特空间,并用核的方法在此空间对比较函数进行计算和寻优。对含有多种分布的过程源数据,核独立元分析是一种比独立元分析(ICA)更有效的特征提取方法。对于按批次方向展开的间歇过程历史建模数据,在每一个时间间隔点应用核独立元分析算法提取独立元用于建模,并计算I2和SPE统计量及相应的控制限。此方法不需要对未来测量值进行估计,更重要的是解决了核独立元分析不能直接处理间歇过程高维历史建模数据的难题。仿真结果验证了所提出方法的可行性和有效性,并显示出比传统MICA更好的监控效果。     

基于KECA的化工过程故障监测新方法

针对化工过程数据复杂、非线性的特点,提出一种基于核熵成分分析(KECA)的化工过程故障监测算法。首先,KECA算法按照Renyi熵值的大小选取特征值及特征向量,相比传统的KPCA监测算法,其保留主元个数更少,可以有效减少运算量。同时,仿真研究表明KECA算法选取的主元具有角度结构特性,据此,提出一种新的统计量--CS(Cauchy-Schwarz)统计量,其对应到核特征空间中即为向量间的角度余弦值,可以较好表述不同概率密度分布之间的相似度。最后,将KECA和KPCA算法分别应用于TE(Tennessee Eastman)过程,结果表明KECA在故障检测延迟与检出率相比KPCA都有很大的优势。     

A non-Gaussian pattern matching based dynamic process monitoring approach and its application to cryogenic air separation process

Principal component analysis (PCA) based pattern matching methods have been applied to process monitoring and fault detection. However, the conventional pattern matching approaches do not specifically take into account the non-Gaussian dynamic features in chemical processes. Furthermore, those techniques are more focused on fault detection instead of fault diagnosis. In this study, a non-Gaussian pattern matching based fault detection and diagnosis method is developed and applied to monitor cryogenic air separation process. First, independent component analysis (ICA) models are built on the normal benchmark and monitored data sets along sliding windows. The IC subspaces from the benchmark and monitored data are then extracted to evaluate the non-Gaussian patterns and detect process faults through a mutual information based dissimilarity index. Further, a difference subspace between the two IC subspaces is computed to characterize the divergence of the dynamic and non-Gaussian patterns between the benchmark and monitored data. Subsequently, the mutual information between the IC difference subspace and each process variable direction is defined as a new non-Gaussian contribution index for fault identification and diagnosis. The presented approach is applied to a simulated cryogenic air separation plant and the monitoring results are compared against those of PCA based pattern matching techniques and ICA based monitoring method. The application study demonstrates that the developed non-Gaussian pattern matching approach can effectively monitor the complex air separation process with superior fault detection and diagnosis capability.     

一种改进的动态核主元分析故障检测方法

针对复杂工业系统动态非线性故障检测过程精度低和计算量大的问题,提出了一种改进的动态核主元分析故障检测方法，该方法首先利用不可区分度剔除相关程度较小或者不相关变量,减少数据量，然后通过观测值扩展对筛选后的新数据构建增广矩阵，并对矩阵使用核主元分析提取变量数据的非线性空间相关特征，最后通过监测T ²和SPE 两种统计量诊断出系统发生故障及识别故障变量。仿真实验证明，该方法能对风力发电机故障进行有效监测和诊断，与KPCA方法相比，改进的动态核主元分析方法对微小故障更为敏感。     

Improved performance of process monitoring based on selection of key principal components

Conventional principal component analysis (PCA) can obtain low-dimensional representations of original data space, but the selection of principal components (PCs) based on variance is subjective, which may lead to information loss and poormonitoring performance. To address dimension reduction and information preservation simultaneously, this paper proposes a novel PC selection scheme named full variable expression.On the basis of the proposed relevance of variables with each principal component, key principal components can be determined. All the key principal components serve as a low-dimensional representation of the entire original variables, preserving the information of original data spacewithout information loss.A squaredMahalanobis distance,which is introduced as themonitoring statistic, is calculated directly in the key principal component space for fault detection. To test the modeling and monitoring performance of the proposed method, a numerical example and the Tennessee Eastman benchmark are used.     

基于互信息的分散式动态PCA故障检测方法

对现代大型复杂动态过程来讲,不同测量变量会存在不同的序列相关性,而且变量间的相互影响会体现在不同的采样时刻上。为此,结合利用分散式建模的优势,提出一种基于互信息的分散式动态过程故障检测方法。该方法在对每个测量变量都引入多个延时测量值后,利用互信息为每个变量区分出与其相关的测量值,并建立起相应的变量子块。这种变量分块方式使每个变量子块都能充分地获取与之相对应的自相关性与交叉相关性信息,较好地处理了数据的动态性问题。然后,利用主元分析（PCA）算法对每一变量子块进行统计建模从而建立起适于大规模动态过程的多模块化的故障检测模型。最后,通过实例验证该方法用于动态过程监测的可行性和有效性。     

基于主元分析和核密度估计的多变量统计过程监控及在工厂聚丙烯催化剂反应器的应用

Abstract Data-driven tools, such as principal component analysis （PCA） and independent component analysis （ICA） have been applied to different benchmarks as process monitoring methods. The difference between the two methods is that the components of PCA are still dependent while ICA has no orthogonality constraint and its latentvariables are independent. Process monitoring with PCA often supposes that process data or principal components is Gaussian distribution. However, this kind of constraint cannot be satisfied by several practical processes. To ex-tend the use of PCA, a nonparametric method is added to PCA to overcome the difficulty, and kernel density estimation （KDE） is rather a good choice. Though ICA is based on non-Gaussian distribution intormation, .KDE can help in the close monitoring of the data. Methods, such as PCA, ICA, PCA.with .KDE（KPCA）, and ICA with KDE,（KICA）, are demonstrated and. compared by applying them to a practical industnal Spheripol craft polypropylene catalyzer reactor instead of a laboratory emulator.     

Decentralized fault diagnosis using multiblock kernel independent component analysis

In this paper, a multiblock kernel independent component analysis (MBKICA) algorithm is proposed. Then a new fault diagnosis approach based on MBKICA is proposed to monitor large-scale processes. MBKICA has superior fault diagnosis ability since variables are grouped and the non-Gaussianity is considered compared to standard kernel methods. The proposed method is applied to fault detection and diagnosis in the continuous annealing process. The proposed decentralized nonlinear approach effectively captures the nonlinear relationship and non-Gaussianity in the block process variables, and shows superior fault diagnosis ability compared to other methods.