半监督稀疏鉴别核局部线性嵌入的非线性过程故障检测

复杂过程往往受到运行状态复杂、工作条件恶劣等因素影响,过程数据具有很强的非线性、随机性和流形结构.近年来,核局部线性嵌入(kernel locally linear embedding,KLLE)已经成功应用于复杂过程故障检测.然而KLLE是一种无监督流形学习算法,能够保持样本的局部几何信息,忽视了总体数据样本集全局/非局部鉴别信息.针对上述问题,本文提出一种新的半监督稀疏鉴别核局部线性嵌入(semi-supervised sparse discriminantKLLE,SSDKLLE)算法并用于非线性工业过程故障检测.本文主要贡献如下:(1)把半监督学习与Fisher鉴别分析(fisher discriminant analysis,FDA)引入到KLLE,有效地利用了总体数据集几何鉴别信息,提高了算法对不同类别数据的分离性;(2)基于稀疏表示通过重构优化方法对信号自适应稀疏表达的优点,利用稀疏表示自适应选择最近邻样本以及数目,提高算法鲁棒性和局部保持性能;(3)引入局部邻域处理以及核技巧策略降低过程工况数据变化对监测算法的影响,提高非线性多工况过程监测方法的性能.基于UCI数据和TE平台的仿真实验结果验证了所提算法的有效性.

IGSA-KPCA邻域建模的多模过程故障检测方法

为提高多模过程故障检测的准确率,提出改进引力搜索算法-核主元分析邻域建模的故障检测方法。首先应用及时学习算法在参考数据集中找到待检数据的相关数据,再将相关数据和待检数据作为核主元分析检测模型的输入进行故障检测。核主元分析模型中的参数对故障检测性能有较大影响,提出改进引力搜索算法对模型中参数进行优化,提高检测性能。将所提方法应用于青霉素多模过程进行实验验证,仿真结果表明所提方法在多模过程故障检测中用时短、准确率高。

刍议多尺度主元分析对化工过程故障检测的作用及实践

化工过程中的数据具有非线性以及多尺度性特征,针对这一特征,提出了一种多尺度主元分析方法。采用小波变化对多尺度性数据进行分析,并借助核函数对非线性数据映射问题加以解决,有效的解决了化工过程中变量数据时间序列性动态问题。本文就在介绍多尺度主元分析方法的基础上,探讨其在化工过程故障检测中的具体应用。

基于故障子空间与PCA监测模型的故障可检测性研究

由于基于主元分析(Principal Component Analysis,PCA)的统计监控方法没有利用过程机理模型(First Principle Model)信息,因此在一定程度上限制了其故障诊断能力的发展。本文基于PCA的框架,采用故障子空间对故障进行描述,在PCA监测模型的基础之上,分析了主元空间和残差空间的故障可检测性问题,获得了故障可检测性的必要充分理论条件。通过对双效蒸发过程的仿真监测,证实了所获理论结果的有效性,表明了通过计算临界故障幅值就可事先对故障集内各故障的检测结果作定量的分析,从而事先了解各故障在PCA下的检测结果。

Fault Detection and Diagnosis of a Gearbox in Marine Propulsion Systems Using Bispectrum Analysis and Artificial Neural Networks

A marine propulsion system is a very complicated system composed of many mechanical components.As a result,the vibration signal of a gearbox in the system is strongly coupled with the vibration signatures of other components including a diesel engine and main shaft.It is therefore imperative to assess the coupling effect on diagnostic reliability in the process of gear fault diagnosis.For this reason,a fault detection and diagnosis method based on bispectrum analysis and artificial neural networks (ANNs) was proposed for the gearbox with consideration given to the impact of the other components in marine propulsion systems.To monitor the gear conditions,the bispectrum analysis was first employed to detect gear faults.The amplitude-frequency plots containing gear characteristic signals were then attained based on the bispectrum technique,which could be regarded as an index actualizing forepart gear faults diagnosis.Both the back propagation neural network (BPNN) and the radial-basis function neural network (RBFNN) were applied to identify the states of the gearbox.The numeric and experimental test results show the bispectral patterns of varying gear fault severities are different so that distinct fault features of the vibrant signal of a marine gearbox can be extracted effectively using the bispectrum,and the ANN classification method has achieved high detection accuracy.Hence,the proposed diagnostic techniques have the capability of diagnosing marine gear faults in the earlier phases,and thus have application importance.

Fault diagnosis of chemical processes based on partitioning PCA and variable reasoning strategy

Fault detection and identification are challenging tasks in chemical processes, the aim of which is to decide out of control samples and find fault sensors timely and effectively. This paper develops a partitioning principal component analysis(PPCA) method for process monitoring. A variable reasoning strategy is proposed and applied to recognize multiple fault variables. Compared with traditional process monitoring methods, the PPCA strategy not only reflects the local behavior of process variation in each model(each direction of principal components),but also improves the monitoring performance through the combination of local monitoring results. Then, a variable reasoning strategy is introduced to locate fault variables. Unlike the contribution plot, this method locates normal and fault variables effectively, and gives initiatory judgment for ambiguous variables. Finally, the effectiveness of the proposed process monitoring and fault variable identification schemes is verified through a numerical example and TE chemical process.

A local and global statistics pattern analysis method and its application to process fault identification

Traditional principal component analysis （PCA） is a second-order method and lacks the ability to provide higherorder representations for data variables. Recently, a statistics pattern analysis （SPA） framework has been incorporated into PCA model to make full use of various statistics of data variables effectively. However, these methods omit the local information, which is also important for process monitoring and fault diagnosis. In this paper, a local and global statistics pattern analysis （LGSPA） method, which integrates SPA framework and locality pre- serving projections within the PCK is proposed to utilize various statistics and preserve both local and global in- formation in the observed data. For the purpose of fault detection, two monitoring indices are constructed based on the LGSPA model. In order to identify fault variables, an improved reconstruction based contribution （IRBC） plot based on LGSPA model is proposed to locate fault variables. The RBC of various statistics of original process variables to the monitoring indices is calculated with the proposed RBC method. Based on the calculated RBC of process variables＇ statistics, a new contribution of process variables is built to locate fault variables. The simula- tion results on a simple six-variable system and a continuous stirred tank reactor system demonstrate that the proposed fault diagnosis method can effectively detect fault and distinguish the fault variables from normal variables.

Fault detection of large-scale process control system with higher-order statistical and interpretative structural model

Nonlinear characteristic fault detection and diagnosis method based on higher-order statistical(HOS) is an effective data-driven method, but the calculation costs much for a large-scale process control system. An HOS-ISM fault diagnosis framework combining interpretative structural model(ISM) and HOS is proposed:(1) the adjacency matrix is determined by partial correlation coefficient;(2) the modified adjacency matrix is defined by directed graph with prior knowledge of process piping and instrument diagram;(3) interpretative structural for large-scale process control system is built by this ISM method; and(4) non-Gaussianity index, nonlinearity index, and total nonlinearity index are calculated dynamically based on interpretative structural to effectively eliminate uncertainty of the nonlinear characteristic diagnostic method with reasonable sampling period and data window. The proposed HOS-ISM fault diagnosis framework is verified by the Tennessee Eastman process and presents improvement for highly non-linear characteristic for selected fault cases.

Application of Kernel Independent Component Analysis for Multivariate Statistical Process Monitoring

In this research, a new fault detection method based on kernel independent component analysis （kernel ICA） is developed. Kernel ICA is an improvement of independent component analysis （ICA）, and is different from kernel principal component analysis （KPCA） proposed for nonlinear process monitoring. The basic idea of our approach is to use the kernel ICA to extract independent components efficiently and to combine the selected essential independent components with process monitoring techniques. 12 （the sum of the squared independent scores） and squared prediction error （SPE） charts are adopted as statistical quantities. The proposed monitoring method is applied to Tennessee Eastman process, and the simulation results clearly show the advantages of kernel ICA monitoring in comparison to ICA monitoring.

Stochastic differential equation software reliability growth model with change-point

This paper presents software reliability growth models(SRGMs) with change-point based on the stochastic differential equation(SDE).Although SRGMs based on SDE have been developed in a large scale software system,considering the variation of failure distribution in the existing models during testing time is limited.These SDE SRGMs assume that failures have the same distribution.However,in practice,the fault detection rate can be affected by some factors and may be changed at certain point as time proceeds.With respect to this issue,in this paper,SDE SRGMs with changepoint are proposed to precisely reflect the variations of the failure distribution.A real data set is used to evaluate the new models.The experimental results show that the proposed models have a fairly accurate prediction capability.