基于支持向量数据描述的非高斯过程故障重构与诊断

提出一种基于支持向量描述（SVDD）的统计过程监控与故障重构及诊断算法,避免了PCA、PLS等传统统计过程监控方法假设过程数据服从高斯分布的不足。鲁棒故障重构算法通过迭代保证重构后的数据对应的SVDD监控统计量最小化。诊断算法根据故障集中的不同故障重构后监控统计量是否恢复正常,确定实际发生的过程故障。CSTR过程的仿真研究表明了所提出方法的有效性。     

基于改进KFDA与DE优化SOM的故障诊断模型及其化工过程诊断

针对化工过程故障诊断数据存在高维度、故障特征不易区分、自组织映射（self-organizing map,SOM）网络易陷入局部最优等问题,提出了一种基于改进核Fisher判别分析（kernel Fisher discriminant analysis,KFDA）与差分进化算法（differential evolution,DE）优化SOM神经网络相结合的故障诊断方法。该方法首先利用欧氏距离对类间距进行加权处理,以避免因类间距离过大造成投影后的数据存在重叠的问题,使故障数据样本获得较好的投影效果,优化分类性能;然后,利用DE算法对SOM神经网络的权值向量进行动态调整,有效避免了由于“死神经元”的出现陷入局部最优的问题;最后,通过对田纳西-伊斯曼（tennessee-eastman,TE）过程和对二甲苯（paraxylene,PX）歧化工艺过程的故障数据进行诊断测试。结果表明,与传统SOM网络相比,提出的KFDA-DE-SOM算法具有较高的分类诊断精度,可有效应用于化工过程的故障诊断。     

基于加权深度支持向量数据描述的工业过程故障检测

传统支持向量数据描述（SVDD）方法本质上采用浅层学习框架,难以有效监控非线性工业过程的复杂故障。针对此问题,提出一种基于加权深度支持向量数据描述（WDSVDD）的故障检测方法。该方法一方面在深度学习框架下重新定义SVDD优化目标函数,构建基于深度特征的深度SVDD监控模型（DSVDD）,并利用核密度估计法计算监控指标的统计控制限;另一方面,考虑到深度特征的故障敏感度差异特性,在DSVDD监控模型中设计特征加权层,分别从静态和动态信息分析角度给出权重因子的计算方法,利用权重因子突出故障敏感特征的影响以提高故障检测率。应用于一个典型化工过程的测试结果表明,所研究的方法能够比传统SVDD方法更有效地监控过程中复杂故障的发生。     

基于SVDD的冷水机组传感器故障检测及效率分析

传感器是制冷空调系统的重要组成部分,起着测量数据和监控状态的作用。传感器故障,尤其是输出偏差会引起测量值不准,影响控制策略,导致系统能耗增加。依据模式识别理论,故障检测可处理为一种单分类问题。据此采用一种单分类模式识别工具——支持向量数据描述(SVDD),针对冷水机组进行了偏差故障条件下的传感器故障检测工作。收集冷水机组实测正常运行数据,基于训练集建立SVDD模型,进行冷水机组传感器故障检测;在测试集中引入不同幅值水平的偏差故障,分析检测效率。结果表明:基于SVDD的冷水机组传感器故障检测效果明显,但对于不同传感器的不同幅值偏差故障,故障识别程度并不一致。     

基于密度权重支持向量数据描述的冷水机组故障检测

虚警率（FAR）是评价冷水机组故障检测性能的关键指标，用户无法接受过高的FAR。为了降低支持向量数据描述（SVDD）在冷水机组故障检测时的FAR，将密度权重集成到SVDD中，提出了一种基于密度权重支持向量数据描述（DW-SVDD）的冷水机组故障检测方法，该方法考虑了样本数据在真实空间中的密度分布情况。使用ASHRAE RP-1043冷水机组实验数据对提出的方法进行验证，并将检测结果与传统SVDD的冷水机组故障检测方法进行比较。结果表明，提出的方法将FAR从10.5%降低到7%，同比下降超过了30%，同时对4个劣化等级下的7种典型冷水机组故障有着优良的检测性能。     

基于MSPCA-KECA的冷水机组故障监测及诊断

针对冷水机组同类型不同等级故障的变量变化存在差异会造成误诊断的问题，提出一种基于多尺度主元分析-核熵成分分析（MSPCA-KECA）的故障诊断策略。MSPCA提取故障特征，其输出作为KECA分类器的输入，实现故障的实时监测与自动诊断。首先，改进的MSPCA算法通过将小波多尺度分析与主元分析相结合，筛选故障信息可能存在的尺度直接重构并采用PCA提取故障特征，获取不同类型故障之间差异的同时也保留了同类型但不同等级故障之间的相似性，提高故障诊断的可靠性。之后建立KECA非线性分类器并引入一种新的监测统计量--散度测度统计量，使降维后不同特征信息之间呈现显著的角度差异，易于分类。最后，采用支持向量数据描述（SVDD）算法确定新统计量的控制限，以克服无法获知统计量分布的问题。通过对冷水机组数据的仿真研究，验证了MSPCA-KECA方法的可行性及有效性。     

基于光谱数据融合和人工神经网络的汽车灯罩鉴别

针对法庭科学领域对物证快速、无损、准确的检验需求，采用红外光谱原始数据和导数数据相结合的光谱数据融合技术对汽车灯罩样本进行分析。对收集的44个汽车灯罩样本采集红外谱图，采用自动基线校正、峰面积归一化、Savitzky?Golay 算法平滑对谱图进行预处理，并对处理后的数据进行一阶求导，结合人工神经网络（ANN）算法构建分类模型。在径向基函数神经网络（RBF）模型中，结合主成分分析对光谱原始数据、一阶导数数据和融合的数据进行分类，分类准确率分别为81.2 %、84.1 %和90.9 %；在多层感知器神经网络（MLP）模型中，结合主成分分析对光谱原始数据、一阶导数数据和融合的数据进行分类，分类准确率分别为84.1 %、86.4 %和97.7 %，且在对44个汽车灯罩样本的12种品牌进行分类时，分类准确率也达到97.7 %，实验结果理想。结果表明，基于红外光谱原始数据和导数数据相结合的光谱数据融合技术能够实现对汽车灯罩样本的准确分析，且满足快速、无损、准确的检验要求，可以为光谱融合技术在法庭科学领域中物证的检验提供一定参考。     

基于Fisher判别⁃支持向量机的汽车灯罩显微激光拉曼光谱研究

为实现对司法鉴定工作中经常遇到的汽车灯罩类物证进行数据化、可视化的无损高效率鉴别,采用PCA主成分分析前处理结合FDA-SVM（RBF）组合分析鉴别物证的方法,对获取的“奥迪”“别克”等18个品牌的173组拉曼红外光谱数据进行了实验和理论分析。借助Pearson相关性分析和PCA主成分分析的结果选择特征位移,分别建立基于Fisher判别分析和SVM支持向量机的数据分类模型。结果表明,FDA模型和SVM（RBF）模型对灯罩样本的综合区分准确率分别为97 %和51.85 %,SVM模型对“奔驰”“别克”等8个品牌的区分准确率达到了100 %,FDA与SVM模型互相补充的FDA-SVM（RBF）模型可对不同品牌灯罩拉曼红外光谱进行准确区分,分类效果较好。该方法高效、准确,对侦查破案中借助灯罩物证鉴定缩小侦察范围有一定的参考意义。     

基于KSFDA-SVDD的非线性过程故障检测方法

慢特征分析(SFA)是一种无监督的线性学习算法,没有考虑过程数据的类别信息和非线性特征。针对此问题,提出一种基于核慢特征判别分析(KSFDA)和支持向量数据描述(SVDD)的非线性过程故障检测方法KSFDA-SVDD。该方法首先利用核技巧将数据从原始空间映射到高维空间,然后通过最大化正常工况数据和故障模式数据之间伪时间序列的时间变化同时最小化正常工况数据内部伪时间序列的时间变化计算判别矩阵,最后利用SVDD描述采用判别矩阵降维后的正常工况数据的分布域,构建监控统计量检测过程故障。在连续搅拌反应器(CSTR)过程上的仿真结果表明所提出方法的故障检测性能优于传统的KPCA方法。     

基于MFDA的井下排水过程监控算法研究

针对现有泵房监控系统运行可靠性差、兼容性不强、扩展能力受限的问题,分析了排水系统特点和控制规律,总结了影响排水系统工作性能的因素,提出了基于HHT和FDA的MFDA排水过程故障监测算法,对监测到的主要过程量进行HHT变换,在多个尺度上建立FDA模型,该算法可较好地识别出故障工况,提高了故障监测准确率。     

A combined canonical variate analysis and Fisher discriminant analysis (CVA–FDA) approach for fault diagnosis

This paper proposes a combined canonical variate analysis (CVA) and Fisher discriminant analysis (FDA) scheme (denoted as CVA–FDA) for fault diagnosis, which employs CVA for pretreating the data and subsequently utilizes FDA for fault classification. In addition to the improved handling of serial correlations in the data, the utilization of CVA in the first step provides similar or reduced dimensionality of the pretreated datasets compared with the original datasets, as well as decreased degree of overlap. The effectiveness of the proposed approach is demonstrated on the Tennessee Eastman process. The simulation results demonstrate that (i) CVA–FDA provides better and more consistent fault diagnosis than FDA, especially for data rich in dynamic behavior; and (ii) CVA–FDA outperforms dynamic FDA in both discriminatory power and computational time.     

A support vector clustering‐based probabilistic method for unsupervised fault detection and classification of complex chemical processes using unlabeled data

A new support vector clustering (SVC)‐based probabilistic approach is developed for unsupervised chemical process monitoring and fault classification in this article. The spherical centers and radii of different clusters corresponding to normal and various kinds of faulty operations are estimated in the kernel feature space. Then the geometric distance of the monitored samples to different cluster centers and boundary support vectors are computed so that the distance–ratio–based probabilistic‐like index can be further defined. Thus, the most probable clusters can be assigned to the monitored samples for fault detection and classification. The proposed SVC monitoring approach is applied to two test scenarios in the Tennessee Eastman Chemical process and its results are compared to those of the conventional K‐nearest neighbor Fisher discriminant analysis (KNN‐FDA) and K‐nearest neighbor support vector machine (KNN‐SVM) methods. The result comparison demonstrates the superiority of the SVC‐based probabilistic approach over the traditional KNN‐FDA and KNN‐SVM methods in terms of fault detection and classification accuracies. © 2012 American Institute of Chemical Engineers AIChE J, 59: 407–419, 2013     

Hyperplane distance neighbor clustering based on local discriminant analysis for complex chemical processes monitoring

The collected training data often include both normal and faulty samples for complex chemical processes. However, some monitoring methods, such as partial least squares (PLS), principal component analysis (PCA), independent component analysis (ICA) and Fisher discriminant analysis (FDA), require fault-free data to build the normal operation model. These techniques are applicable after the preliminary step of data clustering is applied. We here propose a novel hyperplane distance neighbor clustering (HDNC) based on the local discriminant analysis (LDA) for chemical process monitoring. First, faulty samples are separated from normal ones using the HDNC method. Then, the optimal subspace for fault detection and classification can be obtained using the LDA approach. The proposed method takes the multimodality within the faulty data into account, and thus improves the capability of process monitoring significantly. The HDNC-LDA monitoring approach is applied to two simulation processes and then compared with the conventional FDA based on the K-nearest neighbor (KNN-FDA) method. The results obtained in two different scenarios demonstrate the superiority of the HDNC-LDA approach in terms of fault detection and classification accuracy.     

Localized Fisher discriminant analysis based complex chemical process monitoring

Complex chemical process is often corrupted with various types of faults and the fault‐free training data may not be available to build the normal operation model. Therefore, the supervised monitoring methods such as principal component analysis (PCA), partial least squares (PLS), and independent component analysis (ICA) are not applicable in such situations. On the other hand, the traditional unsupervised algorithms like Fisher discriminant analysis (FDA) may not take into account the multimodality within the abnormal data and thus their capability of fault detection and classification can be significantly degraded. In this study, a novel localized Fisher discriminant analysis (LFDA) based process monitoring approach is proposed to monitor the processes containing multiple types of steady‐state or dynamic faults. The stationary testing and Gaussian mixture model are integrated with LFDA to remove any nonstationarity and isolate the normal and multiple faulty clusters during the preprocessing steps. Then the localized between‐class and within‐class scatter mattress are computed for the generalized eigenvalue decomposition to extract the localized Fisher discriminant directions that can not only separate the normal and faulty data with maximized margin but also preserve the multimodality within the multiple faulty clusters. In this way, different types of process faults can be well classified using the discriminant function index. The proposed LFDA monitoring approach is applied to the Tennessee Eastman process and compared with the traditional FDA method. The monitoring results in three different test scenarios demonstrate the superiority of the LFDA approach in detecting and classifying multiple types of faults with high accuracy and sensitivity. © 2010 American Institute of Chemical Engineers AIChE J, 2011     

Fault Diagnosis in Chemical Process Based on Self-organizing Map Integrated with Fisher Discriminant Analysis

Fault diagnosis and monitoring are very important for complex chemical process. There are numerous methods that have been studied in this field, in which the effective visualization method is still challenging. In order to get a better visualization effect, a novel fault diagnosis method which combines self-organizing map （SOM） with Fisher discriminant analysis （FDA） is proposed. FDA can reduce the dimension of the data in terms of maximizing the separability of the classes. After feature extraction by FDA, SOM can distinguish the different states on the output map clearly and it can also be employed to monitor abnormal states. Tennessee Eastman （TE） process is em- ployed to illustrate the fault diagnosis and monitoring performance of the proposed method. The result shows that the SOM integrated with FDA method is efficient and capable for real-time monitoring and fault diagnosis in complex chemical process.     

基于Fisher判别分析和核回归的质量监控和估计

A novel systematic quality monitoring and prediction method based on Fisher discriminant analysis (FDA) and kernel regression is proposed. The FDA method is first used for quality monitoring. If the process is under normal condition, then kernel regression is further used for quality prediction and estimation. If faults have occurred, the contribution plot in the fault feature direction is used for fault diagnosis. The proposed method can effectively detect the fault and has better ability to predict the response variables than principle component regression (PCR) and partial least squares (PLS). Application results to the industrial fluid catalytic cracking unit (FCCU) show the effectiveness of the proposed method.     

Nonlinear Real‐time Process Monitoring and Fault Diagnosis Based on Principal Component Analysis and Kernel Fisher Discriminant Analysis

The aim of this paper is to propose a novel real‐time process monitoring and fault diagnosis method based on the principal component analysis (PCA) and kernel Fisher discriminant analysis (KFDA). There is a need to develop this method in order to overcome the inherent limitations of the current kernel FDA method. The idea of the method is to initially reduce dimensionality using PCA and then to map the score data in the reduced original space to the high‐dimensional feature space via a nonlinear kernel function. Following this, the optimal Fisher feature vector and discriminant vector are extracted to perform process monitoring. If faults occur, the method uses the degree of similarity between the optimal discriminant vector presented and the optimal discriminant vector of the faults in the historical dataset to perform a diagnosis. The proposed method can effectively capture nonlinear relationships in process variables. In comparison with kernel FDA, the PCA plus kernel FDA method is more efficient and has a more rapid response when used to undertake online monitoring and fault diagnosis. In this study, the method is evaluated by applying it to the fluid catalytic cracking unit (FCCU) process. As a consequence, its effectiveness is demonstrated.     

基于核Fisher包络分析的间歇过程故障诊断

随着间歇过程越来越受重视,其过程监控和故障诊断技术也成为研究热点。以核Fisher判别分析为基础,提出了基于核Fisher的正常工况与故障包络面模型,给出了基于该模型的在线故障诊断流程。此方法利用了Fisher判别分析对类别的划分特点,分别针对正常工况数据和各故障类型数据建立包络曲面模型。与多向Fisher判别分析相比,该方法按批次方向将数据展开,能够解决生产周期不一致问题,在线故障诊断时也不需要预报完整的生产轨迹,并且加入核函数来处理复杂的非线性。最后在青霉素发酵过程的仿真平台上对该方法进行测试,与改进多向Fisher判别分析方法进行对比,该方法获得了满意的诊断效果：能够及早诊断出故障的发生,并在有效识别已有故障的同时还具有对新故障的自学习能力。