Efficient faulty variable selection and parsimonious reconstruction modelling for fault isolation

Reconstruction-based fault isolation, which explores the underlying fault characteristics and uses them to isolate the cause of the fault, has attracted special attention. However, it does not explore how the specific process variables change and which ones are most significantly disturbed under the influences of abnormality; thus, it may not be helpful to understanding the specifics of the fault process. In the present work, an efficient faulty variable selection algorithm is proposed that can detect the significant faulty variables that cover the most common fault effects and thus significantly contribute to fault monitoring. They are distinguished from the general variables that are deemed to follow normal rules and thus are uninformative to reveal fault effects. To further reveal the fault characteristics, the selected significant faulty variables are then chosen to obtain a parsimonious reconstruction model for fault isolation in which relative analysis is performed on these selected faulty variables to explore the relative changes from normal to fault condition. The faulty variable selection can not only focus more on the responsible variables but also exclude the influences of uninformative variables and thus probe more effectively into fault effects. It can also help in finding a more interesting and reliable model representation and better identify the underlying fault information. Its feasibility is illustrated with simulated faults using data from the Tennessee Eastman (TE) benchmark process.     

一种新颖的深度因果图建模及其故障诊断方法

为了实现复杂工业过程故障检测和诊断一体化建模, 提出了一种新颖的深度因果图建模方法. 首先, 利用循环神经网络建立深度因果图模型, 将Group Lasso稀疏惩罚项引入到模型训练中, 自动地检测过程变量间的因果关系. 其次, 利用模型学习到的条件概率预测模型对每个变量建立监测指标, 并融合得到综合指标进行整体工业过程故障检测. 一旦检测到故障, 对故障样本构建变量贡献度指标, 隔离故障相关变量, 并通过深度因果图模型的局部因果有向图诊断故障根源, 辨识故障传播路径. 最后, 通过田纳西?伊斯曼过程进行仿真验证, 实验结果验证了所提方法的有效性.     

A new fault diagnosis method of multimode processes using Bayesian inference based Gaussian mixture contribution decomposition

This paper presents a novel Bayesian inference based Gaussian mixture contribution (BIGMC) method to isolate and diagnose the faulty variables in chemical processes with multiple operating modes. The statistical confidence intervals of traditional principal component analysis (PCA) based T² and SPE diagnostics rely upon the assumption that the operating data follow a multivariate Gaussian distribution approximately and therefore may not be able to determine the faulty variables in multimode non-Gaussian processes accurately. As an alternative solution, the proposed BIGMC method first identifies the multiple Gaussian modes corresponding to different operating conditions and then integrates the Mahalanobis distance based variable contributions across all the Gaussian clusters through Bayesian inference strategy. The derived BIGMC index is of probabilistic feature and includes all operation scenarios with posterior probabilities as weighting factors. The Tennessee Eastman process (TEP) is used to demonstrate the utility of the proposed BIGMC method for fault diagnosis of multimode processes. The comparison of the single-PCA and multi-PCA based contribution approaches shows that the BIGMC method can effectively identify the leading faulty variables with superior diagnosis capability.     

基于区间预测模型的污水处理厂传感器故障检测

污水处理厂配备许多传感器用于监测出水水质。传感器的正常工作与否对保证出水水质至关重要。给出了一种污水处理出水变量传感器故障检测方法。该方法根据入水和出水数据,采用径向基函数神经网络构造出水变量预测模型;使用参数线性集员辨识算法得到网络输出权值的集合描述,从而使预测模型能够给出出水变量的置信区间;以此置信区间为基础获得传感器的故障检测策略。由于置信区间描述了出水变量的存在范围,当传感器测量值超出置信区间,则可推断传感器发生故障。此外,在设计传感器故障检测策略时还考虑了污水处理过程异常的影响。实验结果证实所提方法的有效性。     

超立方体并行计算机的一个新型故障诊断算法

提出了超立方体并行计算机的一个新型系统级故障诊断算法.与现有诊断算法相比,该算法能够在系统中存在较多故障处理器的情况下,正确定位全部故障处理器（代价是至多误诊断三个无故障处理器）.另外,该算法的时间复杂度与最好的现有算法相当.     

Combining mutation and fault localization for automated program debugging

This paper proposes a strategy for automatically fixing faults in a program by combining the ideas of mutation and fault localization. Statements ranked in order of their likelihood of containing faults are mutated in the same order to produce potential fixes for the faulty program. The proposed strategy is evaluated using 8 mutant operators against 19 programs each with multiple faulty versions. Our results indicate that 20.70% of the faults are fixed using selected mutant operators, suggesting that the strategy holds merit for automatically fixing faults. The impact of fault localization on efficiency of the overall fault-fixing process is investigated by experimenting with two different techniques, Tarantula and Ochiai, the latter of which has been reported to be better at fault localization than Tarantula, and also proves to be better in the context of fault-fixing using our proposed strategy. Further experiments are also presented to evaluate stopping criteria with respect to the mutant examination process and reveal that a significant fraction of the (fixable) faults can be fixed by examining a small percentage of the program code. We also report on the relative fault-fixing capabilities of mutant operators used and present discussions on future work.     

Fault detection and diagnosis method for batch process based on ELM-based fault feature phase identification

Because of the multiplicity of operation phases in batch process, which have specific control objects, different dominant process variables and distinct process correlation characteristics, the faults may also have phase characteristic. To conduct fault diagnosis for batch process more precisely, this paper proposes a fault detection and diagnosis method based on fault feature phase identification results. Firstly, extreme learning machine is used to identify fault feature phases between the faulty data set and the normal data set. Then, focusing on the different data nature implied in different fault feature phases, several ‘short stages’ are partitioned for the whole batch. After that, different multiway fisher discriminant analysis (MFDA) models are developed for these ‘short stages,’ respectively. The proposed method can deepen the search space analyzed by fault diagnosis into specific fault feature phases, which not only overcome the disadvantage of too many models in MFDA, but also overcome the disadvantage of low diagnosis accuracy and high false recognition rate of traditional MFDA method. Simulation results show the feasibility and validity of the proposed method.     

Fault diagnosis of rolling element bearings via discriminative subspace learning: Visualization and classification

Rolling element bearings play an important role in ensuring the availability of industrial machines. Unexpected bearing failures in such machines during field operation can lead to machine breakdown, which may have some pretty severe implications. To address such concern, we extend our algorithm for solving trace ratio problem in linear discriminant analysis to diagnose faulty bearings in this paper. Our algorithm is validated by comparison with other state-of art methods based on a UCI data set, and then be extended to rolling element bearing data. Through the construction of feature data set from sensor-based vibration signals of bearing, the fault diagnosis problem is solved as a pattern classification and recognition way. The two-dimensional visualization and classification accuracy of bearing data show that our algorithm is able to recognize different bearing fault categories effectively. Thus, it can be considered as a promising method for fault diagnosis.     

Bayesian filtering of the smearing effect: Fault isolation in chemical process monitoring

The isolation of faulty variables is a crucial step in the determination of the root causes of a process fault. Contribution plots, with their corresponding control limits, are the most popular tools used for isolating faulty variables. However, the isolation results may be misled by the smearing effect. In addition, the control limits of the contributions cannot be used to isolate faulty variables, as the control limits are obtained from normal operating data, which lack any information about the faults. In chemical processes, process faults rarely show random behavior; on the contrary, they will be propagated to different variables due to the actions of the process controllers. During the evolution of a fault, the task of isolating faulty variables needs to be concerned with the faulty variables identified at a previous time-point; in addition, the current decisions should influence the isolation results for the next sample when a fault constantly occurs. In the presented work, an unsupervised data-driven fault isolation method was developed based on Bayesian decision theory. Two fault scenarios of the Tennessee Eastman (TE) process were illustrated using visual comparative analysis to demonstrate how the different faulty variables were isolated when the fault evolved. In the industrial application, the proposed approach successfully located the faulty variables that were individually responsible for the simultaneous occurrence of multiple sensor faults and a process fault.     

基于核规范变量分析的非线性故障诊断方法

提出一种基于核规范变量分析（KCVA）的非线性过程故障诊断方法．该方法使用核函数完成非线性空间到高维线性空间的映射,避免了高维空间中的数据处理和非线性映射函数的使用．在线性空间中使用规范变量分析（CVA）来辨识状态空闻模型,从数据中提取状态信息．3个监测量（Tr^2,Ts^2,Q）用来进行故障检测,同时使用贡献图分离故障变量,并判断故障原因．在CSTR系统上的仿真结果表明,KCVA方法比主元分析法（PCA）和CVA方法能更灵敏地检测到故障的发生,更有效地监控过程变化．     

Fault diagnosis using contribution plots without smearing effect on non-faulty variables

Isolating faulty variables to provide additional information about a process fault is a crucial step in the diagnosis of a process fault. There are two types of data-driven approaches for isolating faulty variables. One is the supervised method, which requires the datasets of known faults to define a fault subspace or an abnormal operating region for each faulty mode. This type of approach is not practical for an industrial process, since the known event lists might not exist for some industrial processes. The counterpart is to isolate faulty variables without a priori knowledge, using, for example, a contribution plot, which is a popular tool in the unsupervised category. However, it is well known that this approach suffers from the smearing effect, which may mislead the faulty variables of the detected faults. In the presented work, a contribution plot without the smearing effect on non-faulty variables was derived based on missing data analysis. Two benchmark examples, the continuous stirred tank reactor (CSTR) and the Tennessee Eastman (TE) process, were provided to compare the fault isolation performances of the alternatives using the missing data approach.     

一种基于数据重构的KPCA故障识别方法

针对核主元分析(KPCA)方法只能实现故障检测,但无法实现故障变量识别的问题,提出一种基于数据重构的KPCA故障变量识别方法。采用改进的数据重构方法对各参数进行重构,然后利用故障识别指数对监控参数进行故障变量识别。通过对某型涡扇发动机进行实验的结果表明,该方法能够准确识别故障变量,从而有助于维护人员分析故障原因,初步确定可能的故障源,大大缩短故障定位及排故的时间,可预防重大事故的发生。     

基于T-PLS贡献图方法的故障诊断技术

多变量统计过程监控对于复杂工业过程是一种有效的故障检测和诊断技术. 最小二乘(或称潜空间投影)模型是多变量统计过程监控中常用的一种投影模型, 能够同时对过程数据和质量数据进行建模. 讨论了一种新的基于全潜空间投影模型的故障诊断技术. 全潜空间投影模型中有4个检测统计量. 提出了一种新的T2贡献图计算方法, 对于所有检测统计量, 得到了相应的贡献图算法. 为了确定一个变量是否发生了故障, 计算所有变量贡献图的控制限. 该技术可以将辨识到的故障变量分为与Y有关和与Y无关的两类. 基于Tennessee Eastman过程的案例研究表明了该技术的有效性.     

加权k最近邻重构分析的工业过程故障诊断

k--最近邻(k--nearest neighbor, k--NN)是一种有效的基于数据驱动的故障检测方法, 该方法在工业过程监视方面已经得到了广泛的应用. 但在过程中存在故障时, 精确地寻找故障根源和识别故障变量是故障诊断的重要目标, 也是保证工业过程安全生产的重要任务. 本文在k--NN故障检测技术的基础上, 提出了一种加权的k--NN重构方法, 对使控制指标减小最大(maximize reduce index, MRI)的过程变量依次进行重构, 进而确定发生故障的传感器. 根据理论分析并结合数值仿真对提出的方法进行了验证, 数值仿真先从精度方面验证了该方法能够有效地对故障传感器数值进行重构, 然后验证了该方法不仅适用于单一传感器 故障诊断, 对于同时发生或者因变量相关性而传播的传感器故障也具有很好的效果. 最后, 该方法被成功应用于TE(Tennessee Eastman)化工过程.     

Probabilistic fault diagnosis method based on the combination of nest-loop fisher discriminant analysis and analysis of relative changes

Bias of data location and increase in data variations are two typical disturbances, which in general, simultaneously exist in the fault process. Targeting their different characteristics, a nested-loop fisher discriminant analysis (NeLFDA) algorithm and relative changes (RC) algorithm are effectively combined for analyzing the fault characteristics. First, a prejudgment strategy is developed to evaluate the fault types and determine what changes are covered in the fault process. Two statistical indexes are defined, which conduct Monte Carlo based center fluctuation analysis and dissimilarity analysis respectively. Second, for the fault data containing those two faults simultaneously, a combined NeLFDA-RC algorithm is proposed for fault deviations modeling, which is termed as CNR-FD. Fault directions concerning bias of data location are extracted by the NeLFDA algorithm and then corresponding fault deviations are removed from the fault data. Then RC algorithm is performed on these fault data to extract directions concerning increase of data variations. These fault directions are used as reconstruction models to characterize each fault class. Particularly, the compromise between these two algorithms is determined by the Monte Carlo based center fluctuation analysis. For online applications, a probabilistic fault diagnosis strategy based on Bayes’ rule is performed to identify fault cause by discovering the right reconstruction models that can make the reconstructed monitoring statistics have the largest probabilities of belonging to normal condition. The motivation of the proposed algorithm is illustrated by a numerical case and the performance of the reconstruction models and the probabilistic fault diagnosis strategy are illustrated using pre-programmed faults from the Tennessee Eastman benchmark process and the real industrial process data from the cut-made process of cigarettes in some cigarette factory.     

基于自联想核回归的带离群值化工过程故障检测EI北大核心CSCD

基于数据驱动的故障检测模型通常要求训练数据必须是正常操作条件下的测量值.然而在实际工业生产过程中,即使在正常工况下,数据集中也难以避免存在离群值.此时若仍采用传统的基于多元统计分析的方法,其监测模型的控制限会受到严重影响,造成故障漏报.因此,为了确保当训练数据包含离群值时,监测模型仍然呈现较好的故障检测效果,本文提出了一种基于自联想核回归的故障检测方法.首先基于最小化β散度的鲁棒预白化算法对训练集进行白化计算,消除变量之间相关性对样本相似度度量的影响.然后通过自联想核回归算法重构正常工况下的验证数据,根据重构误差建立模型监测指标.为了消除离群值对故障样本重构的影响,构造截断函数来避免离群样本参与相似故障数据的重构,并对所有参与构建Q统计量的残差变量基于指数加权滑动平均方法自适应加权,得到新的监测统计量.将该方法运用于田纳西–伊斯曼过程并与其他方法进行比较,验证了本文所提故障检测算法的有效性.     

基于多尺度核主元分析的非线性过程监控方法研究

针对化工过程数据的多尺度性和非线性特性,提出了一种多尺度核主元分析方法(MSKPCA)监控过程的运行状态。使用小波变换在不同尺度下分解测量信号．然后借助于核函数对分解后的数据进行非线性变换,在变换后的线性空间中用主元分析(PCA)提取过程数据的主要特征,构造监控统计量T2和Q来检测故障。在此基础上,提出了一种贡献图方法．计算过程变量对故障的贡献量,用于故障变量的分离。在TE过程上的监控结果表明,MSKPCA可以比PCA和动态PCA更迅速地检测到过程故障,贡献图方法能够正确地分离故障变量。     

Variable-weighted Fisher discriminant analysis for process fault diagnosis

Variable-weighted Fisher discriminant analysis (VW-FDA) is proposed to improve the fault diagnosis performance of the conventional FDA. VW-FDA incorporates the variable weighting into FDA. The variable weighting is used to find out each weight vector for all faults. After all fault data are weighted by the corresponding weight vectors, the summed fault data can be constructed to magnify each fault’s local characteristics. Then, VW-FDA is performed on the summed fault data rather than the original fault data. It is helpful to extract discriminative features from overlapping fault data. Moreover, the partial F-values with the cumulative percent variation are used for exactly variable weighting, which is indispensable to VW-FDA. The proposed approach is applied to Tennessee Eastman process. The results demonstrate that VW-FDA shows better fault diagnosis performance than the conventional FDA.     

一种面向输入输出故障的变结构模型预测控制方法

针对输入执行机构故障及输出测量装置故障往往导致MPC（model predictive control）控制器无法实现控制目标的问题,通过对输入稳态与输出稳态关系的分析,提出将存在故障的输入或者输出从控制器的操作变量和被控输出中去除、改变控制器结构的变结构预测控制方法．由于输入故障变结构控制减少了控制器操作变量自由度导致输出稳态误差很大,故根据输出变量优先级重新计算输出设定点以保障重要输出优先满足控制要求．输出故障变结构控制采用结合输入变量稳态值目标跟踪的DMC（dynamic matrix control）算法,避免了输出传感器故障对系统的影响并且保障了被控输出的控制目标可达．利用Shell benchmark重油分馏塔模型仿真验证了本方法的有效性.     

Fault diagnosis of continuous annealing processes using a reconstruction-based method

The continuous annealing process line (CAPL) has complex process characteristics, such as strong correlation of a large number of process variables and interconnected multi-subsystems and multiple operation zones. Practitioners are concerned with typical process faults, such as strip-break and roll-slippage, whose effects are often confined in a specific zone. Considering the large-scale process characteristics and fault characteristics, a multi-block fault diagnosis method is proposed. A novel reconstruction-based block contribution (RBBC) is first proposed in order to diagnose the faulty block. The reconstruction-based variable contribution (RBVC) within a block is also proposed to determine the faulty variables. The proposed RBBC-RBVC hierarchical scheme is applied successfully to a real CAPL on two fault cases. A finite state machine is utilized to diagnose strip-break and reconstructed combined index is studied to diagnose roll-slippage.