A distribution-free method for process monitoring

Traditional multivariate statistical process control methods such as principal component analysis are limited to Gaussian process data when they used for process monitoring. However, the deficiency is not due to the method itself, but lies in the monitoring statistic construction and its confidence limit determination. This paper proposed a distribution-free method, which employs the one-class SVM to construct new monitoring statistics. Thus two new statistics are developed separately in two subspaces of the PCA model: the principal component subspace and the residual subspace. When some fault has been detected, a novel fault reconstruction scheme is proposed. For fault identification, two new identification indices are constructed. The performance of the proposed method in fault detection, reconstruction and identification is evaluated through a case study of the Tennessee Eastman (TE) benchmark process.     

基于核主元分析与核密度估计的非线性过程故障监测与识别

在针对将核主元分析(kernel principal components analysis, KPCA)与基于高斯分布的控制限(control limits,CLS)相结合会降低其性能的问题,提出了一种基于核主元分析与核密度估计(kernel principal components analysiskernel density estimation, KPCA-KDE)相结合的非线性过程故障监测与识别方法.该方法采用核密度估计(kernel density estimation, KDE)技术来估计基于KPCA的非线性过程监控的CLS.通过研究KPCA和KPCA-KDE所有20个故障的检出率发现,与相应的基于高斯分布的方法进行比较, KDE具有较高的故障检出率;此外,基于KDE的检测延迟等于或低于其他方法.通过改变带宽和保留的主元数量进行故障检测, KPCA记录的FAR值较高,相反, KPCA-KDE方法仍然没有记录任何假报警.在田纳西伊斯曼过程(Tennessee Eastman, TE)上的应用表明,KPCA-KDE比基于高斯假设的CLS的KPCA在灵敏度和检测时间上都具有更好...     

基于独立分量分析的过程监控方法研究

为了解决多变量系统的各个变量之间往往相互影响,且一般不能严格服从高斯分布的问题,采用ICA方法时正常状态下观测的数据进行分析处理,从中提取出统计独立的独立分量,为简化后续分析,对得到的独立分量进行筛选、划分,并分别计算两类统计量:I2统计量和SPE统计量,确定其控制限,与在线数据进行对比,用于监控系统运行.通过一多变量过程仿真实例,证明了这种方法的可靠性,这为ICA应用于监控多变量系统的运行、检测故障的发生提供了有益的思路.     

独立分量分析结合数据划分的故障检测方法

针对多变量系统的各个变量之间往往相互影响,且一般不能严格服从高斯分布的特点,通过运用独立分量分析的方法对正常状态下观测的数据进行处理,从中提取出统计独立的独立分量,为简化后续分析,对得到的独立分量按照向量范数的概念划分为两部分:主部和余部,并分别计算两类统计量:I2统计量和SPE统计量,反映系统变化和偏差,确定其置信界,与在线数据进行对比,用于检测运行系统中发生的故障.最后给出的多变量过程仿真实例,快速准确地检测到了运行中发生的异常,证明了这种方法的可靠性.为广泛应用于监控多变量系统的运行,检测故障的发生提供了一个有效的工具.     

Integrating independent component analysis and local outlier factor for plant-wide process monitoring

We propose a novel process monitoring method integrating independent component analysis (ICA) and local outlier factor (LOF). LOF is a recently developed outlier detection technique which is a density-based outlierness calculation method. In the proposed monitoring scheme, ICA transformation is performed and the control limit of LOF value is obtained based on the normal operating condition (NOC) dataset. Then, at the monitoring phase, the LOF value of current observation is computed at each monitoring time, which determines whether the current process is a fault or not. The comparison experiments are conducted with existing ICA-based monitoring schemes on widely used benchmark processes, a simple multivariate process and the Tennessee Eastman process. The proposed scheme shows the improved accuracy over existing schemes. By adopting LOF, the monitoring statistic is computed regardless of data distribution. Therefore, the proposed scheme integrating ICA and LOF is more suitable for real industry where the monitoring variables are the mixture of Gaussian and non-Gaussian variables, whereas existing ICA-based schemes assume only non-Gaussian distribution.     

Statistical process monitoring with independent component analysis

In this paper we propose a new statistical method for process monitoring that uses independent component analysis (ICA). ICA is a recently developed method in which the goal is to decompose observed data into linear combinations of statistically independent components [1 and 2]. Such a representation has been shown to capture the essential structure of the data in many applications, including signal separation and feature extraction. The basic idea of our approach is to use ICA to extract the essential independent components that drive a process and to combine them with process monitoring techniques. I², I_e² and SPE charts are proposed as on-line monitoring charts and contribution plots of these statistical quantities are also considered for fault identification. The proposed monitoring method was applied to fault detection and identification in both a simple multivariate process and the simulation benchmark of the biological wastewater treatment process, which is characterized by a variety of fault sources with non-Gaussian characteristics. The simulation results clearly show the power and advantages of ICA monitoring in comparison to PCA monitoring.     

独立成分分析和支持向量机混合方法在过程监控中的应用

为克服传统过程监控方法需假设过程特征信号服从多元正态分布的缺陷,本文提出了一种将独立成分分析(ICA)与支持向量机结合的故障诊断方法。通过建立独立成分模型确定相应的统计量界限,筛选出需进一步检测的故障数据,再由支持向量机进行故障识别。将该方法用于化工聚合反应的过程监控与故障诊断中,仿真结果表明,这种混合故障诊断方法通过适当地调节统计量控制界限,不仅能够正确识别故障,而且能够纠正由误检数据引起的误报,提高故障诊断的准确率。     

Weighted preliminary-summation-based principal component analysis for non-Gaussian processes

For industrial chemical process, preliminary-summation-based principal component analysis (PS-PCA), an amended PCA method was recently provided for coping with both Gaussian and non-Gaussian characteristics. By summing the training and monitoring data respectively, PS-PCA is capable of resolving the issue of non-Gaussian processes and achieves higher fault detection rate than the traditional PCA. However, in the PS-PCA summation operation, all data samples are regarded as the same weight, which results in the fault information of newly-samples may be diluted, leading to significant detection delays. To address this challenge, in this paper, we propose a novel weighted PS-PCA (WPS-PCA) method that employs an exponential weighting scheme to put more emphasis on recent information. Subsequently, a mathematical argument demonstrates that when the number of variables is enough plentiful, the obtained summation combined with the generalized central limit theorem conforms to approximately a Gaussian distribution. The kurtosis relationships indicate this conversion will bring out well-pleasing feasibility for conventional PCA. Ultimately, the proposed technique verifies detection performance using the Tennessee Eastman process, which is compared with the existing PCA and PS-PCA schemes, in terms of the fault detection time and fault detection rate. The simulation studies reveal that the proposed method is efficient and superior.     

Non-Gaussian chemical process monitoring with adaptively weighted independent component analysis and its applications

Chemical process monitoring based on independent component analysis (ICA) is among the most widely used multivariate statistical process monitoring methods and has progressed very quickly in recent years. Generally, ICA methods initially employ several independent components (ICs) that are ordered according to certain criteria for process monitoring. However, fault information has no definite mapping relationship to a certain IC, and useful information might be submerged under the retained ICs. Thus, weighted independent component analysis (WICA) for fault detection and identification is proposed to process useful submerged information and reduce missed detection rates of I² statistics. The main idea of WICA is to initially build the conventional ICA model and then use the change rate of the I² statistic (RI²) to evaluate the importance of each IC. The important ICs tend to have higher RI²; thus, higher weighting values are then adaptively set for these ICs to highlight the useful fault information. Case studies on both simple simulated and Tennessee Eastman processes demonstrate the effectiveness of the WICA method. Monitoring results indicate that the performance of I² statistics improved significantly compared with principal component analysis and conventional ICA methods.     

A Dynamic Independent Component Analysis Approach To Fault Detection With New Statistics

This paper presents a fault detection method based on Dynamic Independent Component Analysis (DICA) with new statistics. These new statistics are statistical moments and first characteristic function that surrogate the norm operator to calculate the fault detection statistics to determine the control limit of the independent components (ICs). The estimation of first characteristic function by its series is modified such that the effect of series remainder on estimation is reduced. The advantage of using first characteristic function and moments, over second characteristic function and cumulants, as fault detection statistics is also presented. It is shown that the proposed method can detect a class of faults that the former methods cannot; in particular faults with small amplitude ICs that have either different probability density function or identical probability density function of the ICs, but different low order moments of the ICs compared with the normal performance. Simulation results are used to show the effectiveness of the proposed method.     

Dimension reduction method of independent component analysis for process monitoring based on minimum mean square error

This article proposes a novel dimension reduction method of independent component analysis for process monitoring based on minimum mean square error (MSE). Firstly, the order of the independent components (ICs) is ranked according to their importance estimated by MSE, and the mathematical proof is presented. Secondly, the top-n ICs are selected as dominant components and the dimension of ICs is reduced. The sum of the squared independent scores (I²) and the squared prediction error (SPE) are adopted as monitoring statistics. The control limits of I² and SPE are determined by the kernel density estimation (KDE). The proposed dimension reduction method is applied to fault detection in a simple multivariate process and the simulation benchmark of Tennessee Eastman process. Finally, two fault conditions of pulverizing system in power plant are analyzed by the proposed method. The experiments results verify the effectiveness of the proposed method.     

Fault detection based on robust characteristic dimensionality reduction

In this paper, a novel fault detection method is developed based on robust characteristic dimensionality reduction (RCDR). The time-constrained sparse representation (TCSR) method is firstly introduced by considering the space and time characteristics of industrial process monitoring data simultaneously. It can remove space-related outliers, time-related outliers and noises by solving an optimization problem. Then, a new RCDR method is proposed, which fully utilizes the constructed robust adjacency graph and considers the data characteristics. Its scatter matrices are specially designed by consideration of the data characteristics of fault detection. The within-class scatter matrix only characterizes normal data set with a classic covariance matrix, while the inter-class scatter matrix characterizes the separability between normal data and fault data through a pre-defined scatter matrix. It is worth mentioning that our method does not make Gaussian assumptions about the distribution of the fault data, and the number of projection directions is not limited as well. The TCSR is also embedded into our proposed dimensionality reduction method, enabling it to handle the fault detection problem under strong disturbances. Simulations on Tennessee Eastman process (TEP) and a case study of electric multiple unit (EMU) braking system of high-speed trains fully demonstrate the effectiveness and applicability of our proposed fault detection method.     

Fault detection using robust multivariate control chart

We introduce a new multivariate statistical process control chart for fault detection using robust statistics and principal component analysis. The proposed approach consists of two main steps. In the first step, a robust covariance matrix is determined using the minimum covariance determinant algorithm. In the second step, an eigen-analysis of the robust correlation matrix is performed to derive the robust control limits of the proposed multivariate chart. Our experimental results illustrate the much better fault detection performance of the proposed method in comparison with existing statistical monitoring and process controlling charts.     

Multiblock PLS-based localized process diagnosis

In this paper, we discuss a new fault detection and identification approach based on a multiblock partial least squares (MBPLS) method to monitor a complex chemical process and to model a key process quality variable simultaneously. In multivariate statistical process monitoring using MBPLS, four kinds of monitoring statistics are discussed. In particular, new definitions of the block and variable contributions to T² and Q statistics are proposed and derived in order to identify faults. Also, the relative contribution, which is the ratio of the contribution to the corresponding upper control limit, is considered to find process variables or blocks responsible for faults. As an application study, a large wastewater treatment process in a steel mill plant is monitored and the effluent chemical oxygen demand, which indicates the current process performance, is modeled based on the proposed MBPLS-based fault detection and diagnosis method.     

局部近邻标准化偏最小二乘的多模态间歇过程故障检测

本文针对多模态间歇过程数据多中心和模态方差差异明显的问题,提出了一种基于局部近邻标准化偏最小二乘方法.首先,采用统计模量方法处理间歇过程数据,再利用局部近邻标准化方法将统计模量后的训练数据进行高斯化处理,建立偏最小二乘监控模型,确定控制限;然后,同样对统计模量后的测试数据进行局部近邻标准化处理,再计算测试数据的高斯偏最小二乘监控指标,进行过程监视及故障检测.最后,通过数值实例和青霉素发酵过程验证方法有效性.实验结果表明所提方法解决了故障样本近邻集跨模态问题,对多模态数据具有更好的故障检测能力.     

Integrating independent component analysis and support vector machine for multivariate process monitoring

This study aims to develop an intelligent algorithm by integrating the independent component analysis (ICA) and support vector machine (SVM) for monitoring multivariate processes. For developing a successful SVM-based fault detector, the first step is feature extraction. In real industrial processes, process variables are rarely Gaussian distributed. Thus, this study proposes the application of ICA to extract the hidden information of a non-Gaussian process before conducting SVM. The proposed fault detector will be implemented via two simulated processes and a case study of the Tennessee Eastman process. Results demonstrate that the proposed method possesses superior fault detection when compared to conventional monitoring methods, including PCA, ICA, modified ICA, ICA–PCA and PCA–SVM.     

基于核典型相关性-熵成分分析的工业过程质量监测方法

工业过程多变量、数据高维度和非线性的特点使得对其质量监测及质量相关的故障诊断变得复杂.融合核熵成分分析(KECA)及典型相关分析(CCA)方法的思想,进行特征提取降维的同时确保所提取特征与质量变量的最大相关性,提出一种新的质量相关的工业过程故障检测方法.首先,采用KECA对输入数据进行核空间的映射及特征提取,同时融合CCA算法思想使得所提取特征与质量变量间关联最大化;然后,构建监测统计量并用Parzen窗估计其控制限,用于过程的故障检测;最后,运用所提方法对带钢热连轧工业过程实际生产数据进行分析,并与其他4种传统非线性算法对比分析,实验结果验证了所提方法的准确性、有效性及先进性.     

基于独立源分析的过程监测方法

多元统计过程控制要求观测数据服从正态分布,而实际的5-业过程数据大都不满足正态分布条件.独立源分析（ICA）近几年才发展起来的一种新的统计方法,可以克服对数据分布的依赖性．对此,以ICA算法为核心,引入一种新型的过程监测方法,应用ICA提取独立源,利用I^2图,Ic^2图和SPE图进行故障检测．最后以3水箱系统为例进行了实验研究,取得了很好的效果．     

基于局部近邻标准化和动态主元分析的故障检测策略

针对工业过程的动态和多模态特性,提出一种基于局部近邻标准化（LNS）和动态主元分析（DPCA）相结合的故障检测方法（LNS-DPCA）。首先,在训练数据集中寻找样本的K近邻集;然后,应用K近邻集的均值与标准差对当前样本进行标准化处理;最后,在新的数据集中应用DPCA方法确定T²和SPE控制限进行故障检测。LNS方法能够消除过程的多模态特征,使得标准化后数据近似服从多元高斯分布,且保持过程离群点偏离正常样本轨迹;而结合DPCA方法则能够提高对具有动态特性过程的监视性能。利用数值例子和青霉素发酵过程进行仿真,并将测试结果与主元分析法（PCA）、DPCA、K近邻故障检测（FD-KNN）等方法进行对比分析,验证了LNS-DPCA方法的有效性。     

基于Kullback-Leibler距离的闭环系统传感器微小故障诊断

在闭环控制系统中,当故障幅值较小时,由故障带来的影响会被控制量所掩盖.因此,闭环系统中的微小故障诊断实现更为复杂.本文针对闭环系统中的传感器故障,提出了基于Kullback-Leibler(KL)距离的微小故障在线检测与估计方法.本文首先介绍了KL距离的定义及其在多变量故障检测中的应用,然后提出了结合KL距离与快速移动窗口主成分分析(MWPCA)的在线微小故障检测与估计模型.在高斯分布的假设下,利用系统输入输出残差构造MWPCA的数据矩阵,然后通过在线更新数据矩阵主成分的均值与方差实现KL距离的在线更新,最终实现闭环系统中传感器的在线故障检测与估计.仿真实验表明,该方法能有效实现具有低故障—噪声比(FNR)特性的微小故障诊断.