基于统计模量和局部近邻标准化的局部离群因子故障检测方法

针对多工况过程数据的批次不等长、中心漂移、工况结构不同等特点,提出基于统计模量和局部近邻标准化的局部离群因子故障检测方法（SP-LNS-LOF）。首先计算每个训练样本的统计模量;然后使用局部近邻集标准化统计模量,得到标准样本;最后计算标准化样本的局部离群因子,并将其作为检测指标,将局部离群因子的分位点作为检测控制限,当在线样本的局部离群因子大于检测控制限时,判定其为故障;否则为正常。统计模量提取过程的主要信息,且消除批次不等长的影响;局部近邻标准化克服工况中心漂移和工况结构不同的困难;局部离群因子度量样本的相似度,实现故障样本和正常样本的分离。进行了半导体蚀刻过程故障检测仿真实验,实验结果表明SP-LNS-LOF检测出了全部21个故障,比主元分析（PCA）、核主元分析（kPCA）、基于k近邻的故障检测（FD-kNN）、局部离群因子（LOF）方法具有更高的检测率。理论分析和仿真实验说明SP-LNS-LOF方法适用于多工况过程故障检测,具有较高的故障检测效率,能保证多工况生产过程的安全性。     

Fault detection for multirate sampled-data systems with time delays

In this paper, problems of fault detection in multirate sampled-data (MSD) systems with time delays are studied. The background of our study is the increasing demands for fault detection in complex, distributed process control systems, where the plant, controllers, sensors and actuators are networked by standardized bus systems. The basic idea behand the problem solutions is to derive the parity relations of the MSD system with time delays by taking into account the multirate nature and the multiple time delays. Having introduced two operators, the fault detection problem can be formulated as an optimization problem in the framework of the well-known parity space method. Application and advantages of the approaches proposed in this paper are finally illustrated by simulation examples and by comparison with other indirect design approaches.     

执行器故障多率采样间歇过程的鲁棒 耗散迭代学习容错控制

针对一类具有干扰和执行器故障的多率采样间歇过程,提出一种具有鲁棒耗散性能的迭代学习容错控制算法.通过提升技术将多采样率过程用慢速率采样的状态空间模型来描述,并基于二维系统理论,把迭代学习控制过程转化为等价2D Roesser故障系统,再沿时间和迭代方向设计具有耗散性能的反馈容错控制器,并以线性矩阵不等式形式给出容错控制器存在的充分条件,同时确保多率采样间歇过程在正常和故障条件下的耗散性能.注塑过程的注射速度控制仿真验证了方法的有效性和可行性.     

基于SP-LNS-kNN的半导体生产过程故障检测方法研究*

半导体生产过程是典型的间歇过程,针对其过程数据的多模态、多阶段、模态结构不同和批次不等长等特点,提出了基于统计模量的局部近邻标准化和k近邻相结合的故障检测方法(SP-LNS-kNN)。首先计算样本的统计模量,其次对样本的统计模量使用其局部K近邻集进行标准化,最后计算样本与其前k近邻距离,得到平均累积距离D作为检测指标,进而对工业过程故障进行在线检测。统计模量保留了数据的主要信息,将二维样本数据简化为一维数据。局部近邻标准化可以有效降低中心漂移、模态结构差异明显的影响。SP-LNS-kNN不仅能够对大故障实现检测,并且能够提高对小模态的微弱故障的检测能力。使用SP-LNS-kNN对一个实际半导体生产过程数据进行故障检测实验,并将实验结果与PCA、kPCA、LOF和FD-kNN方法的结果进行对比分析,验证了本文方法的有效性。     

An optimal scheme for fast rate fault detection based on multirate sampled data

Multirate systems are abundant in industry. In this paper, the problem studied is designing a residual generator for fault detection based on multirate sampled data. The key new feature of such a residual generator is that it operates at a fast rate for prompt fault detection. The design is based on optimizing a performance index to obtain an optimal parity space based residual generator. The lifting technique is used to convert the time-varying multirate design problem into a time-invariant one with a causality constraint for implementability. A procedure for computing an explicit optimal, causal solution is proposed. The advantages of this design are shown through an example.     

Observer-Based Fast Rate Fault Detection for a Class of Multirate Sampled-Data Systems

This note deals with the problem of observer-based fast rate fault detection for a class of multirate sampled-data (MSD) systems. Applying a lifting technique, a linear time-invariant (LTI) representation with slow sampling period is firstly obtained for the MSD systems and, based on this, an observer-based fault detection filter is considered as a residual generator. Then an optimization fault detection approach for LTI systems is modified to the residual generation for the MSD systems and, by solving a discrete-time Algebraic Riccati equation, a family of optimal solutions with causality constraint can be obtained. An inverse lifting operation on the generated residual implements its fast rate. The residual evaluation problem is also considered. A numerical example is finally given to illustrate the effectiveness of the proposed design techniques     

基于K近邻证据融合的故障诊断方法

为了兼顾数据建模的准确性和诊断的实时性,提出一种K近邻诊断证据融合新方法.利用故障特征 的历史样本构建随机模糊变量(RFV)形式的故障样板模式,由KNN算法获取测试样本的K个近邻历史样本,并定义它们的RFV待检模式;经样板和待检模式的匹配获取K个诊断证据,再将各特征的K个诊断证据融合,并作出故障决策;使用RFV实现对故障数据的精准建模,利用K个历史样本丰富诊断信息,并增加诊断的时效性.诊断效果在电机转子试验台上得到了验证.     

基于\begin{document}$ \pmb k $\end{document}近邻主元得分差分的故障检测策略

针对具有非线性和多模态特征过程的故障检测问题, 本文提出一种基于k近邻主元得分差分的故障检测策略.首先, 通过主元分析(Principal component analysis, PCA)方法计算样本的真实得分.然后, 应用样本的k近邻均值计算样本估计得分.接下来, 通过上述两种得分计算样本的得分差分矩阵和残差矩阵, 其中残差矩阵由样本的估计得分计算得到,这区别于传统方法.最后, 在差分子空间和残差子空间中分别建立新的统计指标进行故障检测.值得注意的是本文的得分差分方法能够消除数据结构对过程故障检测的影响, 同时, 新的统计量能够提高过程的故障检测率.将本文方法在两个模拟例子和Tennessee Eastman (TE)过程中进行测试, 并与传统方法如PCA、KPCA、DPCA和~FD-kNN等进行对比分析, 测试结果证明了本文方法的有效性.     

Observer-based robust fault detection of multirate linear system using a lift reformulation

This paper presents observer-based robust fault detection for multirate sampled-data linear system. It is assumed that all plant inputs are updated with a unique sampling period and that the plant outputs are sampled uniformly at integer multiples of the input sampling period. A state observer is constructed, as in the single-rate case, and a residual generator is derived. The paper shows that the residual signals can be decoupled from noise and disturbances. An illustrative numerical example is provided.     

FDR验证试验的故障样本分配策略研究

故障率作为测试性验证试验故障样本分配的主要影响因素,针对一些情况下使得故障样本分配结果的合理性不足的问题,以故障检测率(Fault Detection Rate, FDR)作为验证指标,提出了一种考虑严酷度的样本故障模式选取方法。提出了基于模糊证据推理的故障模式严酷度排序解决方法。通过对相关标准中涉及的故障样本分配策略进行梳理,针对现行多因子综合加权比例分配方法不足之处,根据故障模式种类与验证样本量的数量关系,区分不同情况,借助预选样本集随机抽样、考虑严酷度的取整策略,以及动态概率调整,合理改善了故障率主体分配方案进行故障模式选取时样本分配集中不合理的状况。以某装备单元的FDR验证试验为例,验证了所提故障样本分配方法的可行性合理性。     

基于时空近邻标准化和局部离群因子的复杂过程故障检测

针对复杂过程数据的非线性、动态性和中心漂移等特征,提出了基于时空近邻标准化和局部离群因子的故障检测方法(TSNS–LOF).首先使用训练样本在时空两个方向上的近邻集来标准化训练样本;然后在标准样本集上计算样本的局部离群因子,并确定其上分位点作为检测控制限,进行在线故障检测.时空近邻标准化解决了复杂过程数据的非线性、动态性和中心漂移的问题;局部离群因子通过度量样本的相似度实现了故障样本和正常样本的分离.将TSNS–LOF应用于田纳西–伊斯曼过程(TE)过程进行故障检测实验,结果表明相对于主元分析、动态主元分析、k近邻、局部离群因子等方法, TSNS–LOF对故障预警更加及时且具有更高的故障检测率.理论分析和仿真实验说明TSNS–LOF方法适用于具有动态性或多模态特性或两者兼具的过程故障检测,能够更好地保障生产过程的安全性和产品的高质量.     

满足因果约束的多速率单值广义预测控制

针对输入更新频率和输出采样频率均不相同的多速率采样系统,提出一种多速率广义预测控制算法．通过采用在系统周期内根据不同的采样时刻构造扩展模型序列的方法,有效地解决了由于多速率采样而引起的因果约束问题．在此基础上,对算法的闭环稳定性进行了分析,仿真结果证明了算法的有效性．     

Sequential fusion estimation for multisensor multirate systems with measurement outliers

This paper investigates the suboptimal sequential fusion estimation problem for multisensor multirate networked systems with colored measurement noises under the interference of measurement outliers. The saturation function is used to constrain the innovation polluted by measurement outliers. Due to diverse physical restrictions, the sampling period of the sensor is assumed to be different from the update period of the system state, thereby better reflecting the engineering practice. The lifting technique is used to convert the multirate sampling system into a single-rate form. By solving the matrix difference equation, an upper bound of the filtering error covariance is obtained, and the filter gain is then derived, which can minimize the upper bound of the error covariance. Finally, a simulation example is given to demonstrate the effectiveness of the proposed sequential fusion method for multirate sampling systems under outlier interference.     

基于改进的BSMOTE和时序特征的风机故障采样算法

针对风机数据集的不平衡问题,提出了一种BSMOTE-Sequence采样算法,在合成新样本时综合考虑空间和时间特征,并对新样本进行清洗,从而有效减少噪声点的生成。首先,根据每个少数类样本的近邻样本的类别比例,将少数类样本划分为安全类样本、边界类样本和噪声类样本。然后,对每个边界类样本都遴选出空间距离、时间跨度最接近的少数类样本集,利用线性插值法合成新样本,并过滤掉噪声类样本以及类间重叠样本。最后,以支持向量机（SVM）、卷积神经网络（CNN）、长短期记忆（LSTM）人工神经网络作为风机齿轮箱故障检测模型,F1-Score、曲线下面积（AUC）和G-mean作为模型性能评价指标,在真实风机数据集上把所提算法与常用的多种采样算法进行对比,实验结果表明：相比已有算法,BSMOTE-Sequence算法所生成样本的分类效果更好,使得检测模型的F1-Score、AUC和G-mean平均提高了3%,该算法能有效地适用于数据具有时序规律且不平衡的风机故障检测领域。     

A method to design multirate controllers for plants sampled at alow rate

Classical discrete control systems allow control of the output of a plant at sampling instants. If control of the output between sampling instants is required, a multirate controller can be used. An analytical procedure is developed that allows feedforward control of the output at multirate instants and feedback control at sampling instants. The method proposed is a practical approach to the model-matching problem and takes into account the intersampling dynamics of the system     

双目系统全视域采样的SVM标定模型分析

目的 为改善摄像机间接标定采样不全、模型表达模糊问题,实现小视场下检测视域完备采样,提出一种基于双目系统全视域采样的支持向量机（SVM）标定方法。方法 该方法利用六角晶格标定板靶点序号可读特点为基础,采集整个双目系统有效视域中检测点的视差坐标、世界坐标并建立完备的样本集。选取SVM对该样本集进行训练,将SVM算法得到的模型参数代入其决策函数中进行求解,获得公式化的标定模型。由于六角晶格标定板的四角和中心分布了5个互为非中心对称的多边形,可在标定板部分区域被采集的情况下获取标定板位姿信息,进而读取采集的各靶点序号。通过上下移动标定板,利用HALCON算子获取图像中各靶点的序号,建立双目视觉系统检测区域的完备样本集。最后,利用SVM算法训练样本获得标定模型,可以明确表达出标定模型的数学形式。结果 与传统采样建立的模型进行对比分析,实验结果表明该方法建立模型的标定误差减小了24.51%,降低了标定模型在传统方法未采样区域的标定误差,证明了该方法的可行性。结论 提出一种基于双目系统全视域采样的支持向量机标定方法,通过非中心对称的多边形确定标定板上靶点的序号,实现双目视觉系统检测视域的完备采样。实验结果表明该方法提高了摄像机间接标定的精度,具有良好的适用性和鲁棒性,适用于小视域内双目视觉系统的间接标定。     

Stability analysis of explicit multirate methods

If stiff differential systems can be divided into stiff and nonstiff subsystems, each subsystem can be integrated with different step sizes. These methods are called multirate methods, and they have been successfully used in several practical problems, especially in real-time simulations.In this paper a new way to perform the stability analysis of multirate versions of linear multistep methods is presented. This stability analysis makes use of the multirate z-transform method.     

基于改进的局部近邻标准化和kNN的多阶段过程故障检测

针对多阶段过程数据具有多中心和各工序结构不同的特征问题,提出了一种基于改进的局部近邻标准化和k近邻的故障检测（ILNS-kNN）方法。首先寻找样本的前k个近邻样本的前K局部近邻集;其次使用局部近邻集的均值和标准差来标准化样本,获得标准样本;最后在标准样本集上计算样本的累积近邻距离作为检测指标进行故障检测。改进的局部近邻标准化（ILNS）将各阶段数据的中心平移到原点,并且调整各阶段数据的离散程度,使之近似相等,从而将多阶段过程数据融合为服从单一多元高斯分布的单阶段数据。进行了青霉素发酵过程故障检测实验。实验结果表明ILNS-kNN方法对所设置的六类故障的检测率高于97%。ILNS-kNN方法在保持对一般多阶段过程故障的检测能力的同时,能够实现对阶段方差差异显著的多阶段过程故障的检测,从而更好地保证多阶段生产过程的安全性和产品的高质量。     

An overview of multirate multisensor systems: Modelling and estimation

In this paper, research advances in modelling and estimation algorithms for multirate multisensor systems are reviewed. Multirate multisensor sampling schemes can be classified into two cases: uniform sampling and nonuniform sampling. A general method to solve an asynchronous estimation problem for multirate multisensor systems is to transform the asynchronous problem into a synchronous one. Then, synchronous estimation algorithms can be applied. Therefore, modelling synchronisation approaches are summarised for uniform and nonuniform sampling cases. Meanwhile, the corresponding estimation algorithms are reviewed according to different synchronised models. Last, future research topics are discussed.     

Grey box fault detection of heat exchangers

A grey-box model-based method for fault diagnosis is proposed in this paper. The method is based on a first principle model of the process unit, i.e. a heat exchanger, and on a grey-box model of the fault, i.e. the deterioration of the heat transfer surface by aging. During normal operating conditions the heat transfer coefficient is constant or slowly decreasing due to material settling on the heat transfer surface. In old heat exchangers big pieces of settled material can break off causing damage. When this happens, the heat transfer coefficients will rise sharply. In the proposed method a recursive least-squares estimator with forgetting factor is used to track the heat transfer coefficients. The settled material breakage fault is detected via detection of abrupt positive jump in the estimated heat transfer coefficients using a cumulative sum (CUSUM) test. The capability to detect faults in any industrial equipment is heavily dependent on the availability of suitable measurements. For heat exchangers the variables related to the in- and outflows of the equipment (flowrates and temperatures) are usually measured, but measurements along the equipment length are rarely available. Therefore, the possibilities of fault location in space are rather limited. However, simplified models can be used for fault detection in this case. Moreover, a fault detection method is proposed with the possibility of spatial fault location when measurements along the cold side are available. The proposed method is illustrated on simulated examples with different measurement situations.