基于相对变换PLS的故障检测方法

针对偏最小二乘方法(partial least squares,PLS)在无量纲标准化处理后导致的特征值大小近似相等,难以获得代表性的潜变量等问题,提出了一种基于相对变换PLS( relative-transformation PLS,RTPLS)的故障检测方法.该方法引入马氏距离相对变换理论,通过计算采样数据之间的马氏距离,将原始空间数据变换到相对空间.然后在相对空间进行PLS分解,提取有代表性的潜变量,建立故障检测模型,实现采样数据的在线检测.通过对TE (Tennessee Eastman)过程故障和轧钢机系统力传感器故障的仿真实验验证了所提出方法的有效性和实用性.理论分析和仿真实验均表明,基于RTPLS的故障检测方法能有效地消除量纲的影响,提取具有更大的变化度和代表性的隐变量,增加故障检测的精度和实时性.     

基于改进重构贡献图的故障定位方法

针对重构贡献(RBC)方法仅适合单变量故障的定位及贡献图中出现拖尾效应(SE)的问题,本文提出一种基于改进重构贡献图(MRBCP)的故障定位方法。采用概率主元分析(PPCA)建立监视模型和统一量度的监视统计量,克服PCA方法中不同量度的监视统计量造成的诊断结果不一致的缺点。对于故障样本,以变量的重构监视统计量为贡献统计量,通过组合最大化思想对故障变量进行逐次定位。在历史故障信息未知的情况下,能够进行多变量故障的定位,然后在定位出的故障变量中进行贡献图分析,进一步对故障变量实现准确定位,从而避免了拖尾效应。通过数值案例和TE过程——实际化工过程的真实模拟过程进行实验,并与基本RBC方法、基于PCA的MRBCP方法进行比较,结果表明了所提方法的有效性。     

基于方差分析的电子产品质量控制图的研究

为提高电子产品的质量,降低生产成本,使用控制图是个很好的质量控制方法.针对航空电子产品生产过程中统计过程控制方法的局限性,介绍了方差分析及控制图的概念,并考虑到不同环境下样本抽样数据之间的差异,结合方差分析原理对(x)-R控制图的构造方法及上下控制限的确定方法进行了研究.通过实例分析验证了修正后的(x)一R控制图对航空...     

基于数字仿真建模的船闸液压系统故障诊断

液压系统是船闸闸门运行的重要动力机构。系统回路之间耦合性强，故障机理各不相同，导致其故障诊断困难重重。针对以上问题，提出了一种基于数字仿真建模的船闸液压系统故障诊断方法，通过AMEsim仿真平台模拟液压元件故障状态，生成故障数据，结合偏最小二乘法（PLS）建立过程变量与故障类型的回归方程，对液压系统故障类型进行判别分析。结果表明，偏最小二乘法能够实现过程变量数据融合，提取故障特征，完成故障类型的检测、分离与识别。     

基于改进多尺度核主元分析的化工过程故障检测与诊断方法研究

针对化工过程数据的多尺度性和非线性特性,提出了改进多尺度核主元分析法.先利用小波变换分析测量数据的多尺度特性,然后采用核主元分析算法进行在线故障检测,对检测到的故障采用核函数梯度算法实现在线故障诊断,根据每个监控变量对统计量T~2和SPE的贡献程度,绘制贡献图,用于故障的分离.在监控过程中为解决核矩阵计算困难,引入特征向量选择方法.TE过程的仿真结果表明它能有效实现故障检测、故障诊断,与主元分析方法相比,显示出更高的过程监控能力.     

递推PLS方法及其在过程监控中的应用

针对处理批量数据时传统的偏最小二乘(PLS)模型无法在线更新的问题,提出了结合遗忘因子法的递推PLS算法,进而更新Qa控制限,有效地克服了传统PLS中Qa无法反映系统时变性的缺点。并对异丁烯酸甲酯(MMA)聚合反应过程中的数据进行了仿真分析。研究结果表明,递推PLS方法大大降低了监控过程中的误报率和漏报率,提高了监控系统的性能。     

基于主元分析的生产过程质量控制与诊断

针对生产过程中监控变量的波动、故障及过程扰动对生产稳定性及可靠性影响情况下,常规统计过程控制图无法准确识别缺陷的问题,结合主元分析模型设计质量监控图以对生产过程中影响质量的因素进行控制与诊断。通过介绍主元分析模型,利用主元子空间和残差子空间反应的数据变化测度,结合统计控制图技术设计集成生产过程控制模型对生产过程质量进行监控,提出了基于主元分析模型的生产过程控制与诊断流程,对生产中的扰动进行诊断与识别。     

基于子空间混合相似度的过程监测与故障诊断

针对现代工业过程多变量、过程数据通常同时包含高斯性和非高斯性分布的特点,提出了一种基于混合子空间的系统性能监控与故障诊断方法.首先使用小波去噪、PCA和ICA方法来进行过程检测,然后将基于PCA特征子空间距离相似度和基于ICA子空间余弦相似度的方法结合,建立故障诊断库,计算混合相似度,确定各类故障的诊断阈值.最后对在线的数据进行监控,判断过程是否正常.当有故障发生时,利用混合子空间相似度确定故障类型.该方法可以充分利用过程数据中的高斯和非高斯信息.通过对Tennessee Eastman(TE)过程的仿真研究,验证了该方法的可行性与有效性,与变量贡献图方法相比可以更加有效地监测出故障原因.     

基于小批量制造过程的动态质量控制限及其简便计算方法

讨论了一种基于t分布动态控制限的小批量生产质量控制方法，并给出了该控制限的简便计算过程。控制方法通过分析抽样样本数量与控制图虚发警报概率之间的函数对应关系，得到一组能使虚发警报概率保持相对稳定的动态控制限，建立了基于t分布的控制界限值随样本数量变化的数学模型，并利用上侧分位数的初等数学表达式近似表示t分布，从而简化计算过程。实际计算结果表明，当样本组数小于5时，控制方法的计算误差仅为正态逼近法的20％。控制方法形式简单，不需使用正态分布函数，适合小批量生产过程的质量控制。     

MWCVA方法的带钢热连轧过程故障检测研究

带钢热连轧过程中采集的数据具有较强的自相关性、互相关性以及时变性，典型变量分析方法（CVA）虽然能够解决自相关性和互相关性的问题，但是由于过程数据存在时变性，导致以前构建的监控模型不再适用于现在的数据，因此提出了一种基于滑动窗口的典型变量分析方法（MWCVA）。该方法首先通过初始窗口构建CVA模型和统计量，解除了数据之间的自相关性和互相关性，然后通过滑动窗口更新过程数据，不断更新CVA模型和统计量，解决了时变性导致检测结果不准确的问题，最后通过控制限判断是否发生故障，在监测系统内部状态空间的同时监测外部状态空间的变化，更加全面地进行故障检测。通过带钢热连轧过程（HSMP）案例的仿真研究，对比CVA、MWCVA的检测效果，证明了MWCVA对故障识别的精度高达100%，误报率不足0.5%。     

Batch process fault detection and identification based on discriminant global preserving kernel slow feature analysis

As an attractive nonlinear dynamic data analysis tool, global preserving kernel slow feature analysis (GKSFA) has achieved great success in extracting the high nonlinearity and inherently time-varying dynamics of batch process. However, GKSFA is an unsupervised feature extraction method and lacks the ability to utilize batch process class label information, which may not offer the most effective means for dealing with batch process monitoring. To overcome this problem, we propose a novel batch process monitoring method based on the modified GKSFA, referred to as discriminant global preserving kernel slow feature analysis (DGKSFA), by closely integrating discriminant analysis and GKSFA. The proposed DGKSFA method can extract discriminant feature of batch process as well as preserve global and local geometrical structure information of observed data. For the purpose of fault detection, a monitoring statistic is constructed based on the distance between the optimal kernel feature vectors of test data and normal data. To tackle the challenging issue of nonlinear fault variable identification, a new nonlinear contribution plot method is also developed to help identifying the fault variable after a fault is detected, which is derived from the idea of variable pseudo-sample trajectory projection in DGKSFA nonlinear biplot. Simulation results conducted on a numerical nonlinear dynamic system and the benchmark fed-batch penicillin fermentation process demonstrate that the proposed process monitoring and fault diagnosis approach can effectively detect fault and distinguish fault variables from normal variables.     

Graphical enhancement to support PCA-based process monitoring and fault diagnosis

Principal component analysis (PCA) for process modeling and multivariate statistical techniques for monitoring, fault detection, and diagnosis are becoming more common in published research, but are still underutilized in practice. This paper summarizes an in-depth case study on a chemical process with 20 monitored process variables, one of which reflects product quality. The analysis is performed using the PLS - Toolbox 2.01 with MATLAB, augmented with software which automates the analysis and implements a statistical enhancement that uses confidence limits on the residuals of each variable for fault detection rather than just confidence limits on an overall residual. The newly developed graphical interface identifies and displays each variable's contribution to the faulty behavior of the process; and it aids greatly in analyzing results. The case study analyzed within shows that using the statistical enhancement can reduce the fault detection time, and the automated graphical interface implements the enhancement easily.     

Multivariate fault isolation of batch processes via variable selection in partial least squares discriminant analysis

In recent years, multivariate statistical monitoring of batch processes has become a popular research topic, wherein multivariate fault isolation is an important step aiming at the identification of the faulty variables contributing most to the detected process abnormality. Although contribution plots have been commonly used in statistical fault isolation, such methods suffer from the smearing effect between correlated variables. In particular, in batch process monitoring, the high autocorrelations and cross-correlations that exist in variable trajectories make the smearing effect unavoidable. To address such a problem, a variable selection-based fault isolation method is proposed in this research, which transforms the fault isolation problem into a variable selection problem in partial least squares discriminant analysis and solves it by calculating a sparse partial least squares model. As different from the traditional methods, the proposed method emphasizes the relative importance of each process variable. Such information may help process engineers in conducting root-cause diagnosis.     

Optimization of dimensionality of symptom space in machine condition monitoring

With the modern tools of metrology we can measure almost all variables in the phenomenon field of a working machine, and some of measuring quantities can be symptoms of machine condition. On this basis we can form the symptom observation matrix for condition monitoring. From the other side we know that contemporary complex machines can have many modes of failure/damage, so called faults. The paper presents the method of extraction of fault information from the symptom observation matrix by means of singular value decomposition, in the form of generalized fault symptoms. However, at the beginning of monitoring we do not know the sensitivity of potential symptoms to the given machine faults and to its overall condition. Hence, some method of symptom observation matrix optimization leading to redundancy minimization is presented first time in this paper. This gives the possibility to assess the diagnostic contribution of every primary measured symptom. Also in the paper some possibility to assess symptom limit value, based on symptom reliability is considered. These concepts are illustrated by symptom observation matrix processing with the special program and the data are taken directly from the machine vibration condition monitoring area.     

A fault diagnosis approach for roller bearing based on VPMCD under variable speed condition

Essentially the fault diagnosis of roller bearing is a process of pattern recognition. However, existing pattern recognition method failed to capitalize on the nature of multivariate associations between the extracted fault features. Targeting such limitation, a new pattern recognition method – variable predictive model based class discriminate (VPMCD) is introduced into roller bearing fault identification. The VPMCD consider that all or part of the feature values will exhibit interactions in nature and these associations will have different performances between different classes, which is always true in practice when faults occur in roller bearings. Target to the characteristics of non-stationary and amplitude-modulated and frequency-modulated (AM–FM) of vibration signal picked up under variable speed condition, a fault diagnosis method based upon the VPMCD, order tracking technique and local mean decomposition (LMD) is put forward and applied to the roller bearing fault identification. Firstly, LMD and order tracking analysis method are combined to extract the fault features of roller bearing vibration signals under variable speed condition; Secondly, the feature values are regard as the input of VPMCD classifier; finally, the working condition and fault patterns of the roller bearings are identified automatically by the output of VPMCD classifier. The analysis results from experimental signals with normal and defective roller bearings indicate that the proposed fault diagnosis approach can distinguish the roller bearing status-with or without fault and fault patterns under variable speed condition accurately and effectively.     

基于特征贡献率的机械故障分类方法

为提高往复压缩机、航空发动机等复杂机械故障分类的准确率,依据特征参数对不同故障的敏感度存在差异的特性,提出一种狄利克雷过程混合模型(Dirichlet process mixture model,简称DPMM)与贝叶斯推断贡献(Bayesian inference contribution,简称BIC)相结合的分析方法。采用DPMM方法自学习机械振动信号高维特征的统计分布模型,并依据BIC理论计算得到各特征参数对模型的贡献率,通过对比观测数据与各类故障数据特征贡献率间的差异实现故障分类。试验结果表明,该方法的平均分类准确率比基于高斯混合模型(Gaussian mixture model,简称GMM)的故障诊断方法的平均分类准确率提高19.29%,比基于Relief算法的故障诊断方法的平均分类准确率提高32.71%,且该方法的时效性高,泛化性能强,能够更有效地进行复杂机械故障分类。     

A data-driven multiplicative fault diagnosis approach for automation processes

This paper presents a new data-driven method for diagnosing multiplicative key performance degradation in automation processes. Different from the well-established additive fault diagnosis approaches, the proposed method aims at identifying those low-level components which increase the variability of process variables and cause performance degradation. Based on process data, features of multiplicative fault are extracted. To identify the root cause, the impact of fault on each process variable is evaluated in the sense of contribution to performance degradation. Then, a numerical example is used to illustrate the functionalities of the method and Monte-Carlo simulation is performed to demonstrate the effectiveness from the statistical viewpoint. Finally, to show the practical applicability, a case study on the Tennessee Eastman process is presented.     

基于小波变阈值去噪递推偏最小二乘的加工质量预测方法

为准确预测在噪声干扰下的加工质量,基于小波去噪和递推偏最小二乘方法,提出了小波变阈值去噪递推偏最小二乘方法.该方法针对小波硬软阈值去噪的不足,利用小波多尺度去噪,建立了变阈值计算公式,基于两小波域的维纳滤波,实现在偏最小二乘建模前对噪声的小波多尺度变阈值处理;同时,针对递推偏最小二乘算法中的"数据饱和"现象,基于滑动窗口的原理,通过引入折息因子控制遗忘程度,构建了多调节参数的递推偏最小二乘算法.通过该方法构建了加工质量预测模型,进行加工质量的预测,最后,结合具体实例分析,验证了该方法的有效性.     

基于支持向量机和顺序前项选择算法的PNN风电机组液压变桨的故障诊断

针对液压变桨距系统的强耦合、非线性,以及液压变桨距故障发生原因复杂、故障单一造成的定位问题,该文提出基于支持向量机和顺序前项选择算法的概率神经网络诊断方法。首先,选取SCADA数据的特征值为输入,桨距角为输出,利用支持向量机进行模型的回归,得出桨距角输出的预测值;接着,将测量值与预测值带入顺序前项选择算法,挖掘和发现特征与故障之间的关系,评估各特征之间的重要性,并选出最好的一组特征集合;最后,建立变桨距概率诊断模型,将所选的数据送到故障诊断模型进行训练,再用所选数据进行测试,定位出变桨距系统的故障原因。实验分析表明:基于支持向量机和顺序前项选择算法的概率神经网络液压变桨距故障诊断方法可以有效地分辨出不同故障,并且诊断的精确度得到了提高。