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     

基于多动态核聚类的间歇过程在线监控

针对传统的多元统计监测方法不能有效检测工业过程中由于初始条件波动较大所引发的弱故障问题,提出一种基于多动态核聚类的核主元分析(DKCPCA)监控策略,实现多阶段间歇过程的弱故障在线监控.该方法首先针对过程中各阶段每一批次数据结合自回归移动平均时间序列模型(ARMAX)和核主成分分析(KPCA)方法分别建立动态核PCA模型,然后根据各批次模型间载荷的相似性采用分层次聚类方法进行聚类,最后将聚在一起的批次数据进行展开重新再建立动态核PCA模型,随着聚类数目的不同从而建立多个类模型.当在线应用时给出了多模型选择策略,以提高监测精度.将此方法应用于青霉素发酵过程的监控中,监测结果表明此方法取得了比DKPCA和MKPCA更好的监测性能.     

基于潜变量自回归算法的化工过程动态监测方法

从建立潜变量自回归（AR）模型的角度出发，提出了一种基于潜变量自回归（LVAR）算法的化工过程动态建模与监测方法，旨在提取动态潜变量的同时给出各潜变量的AR模型。LVAR算法在最小化潜变量的AR模型残差的约束下，通过同时搜寻投影变换向量与AR系数向量，实现了对动态潜变量的特征提取及其AR模型的建立。此外，LVAR算法通过先提取动态潜变量后提取静态成分信息的方式，有效地区分了采样数据中的自相关性与交叉相关性。在对比实验中，通过比较分析LVAR方法与其他三种典型的动态过程监测方法在经典化工过程对象上的故障监测结果，验证了LVAR方法在动态过程监测上的优越性与可靠性。     

多模态化工过程的全局监控策略

引言基于数据驱动的过程监控方法从20世纪80年代建立以来得到了蓬勃的发展,理论体系逐渐完善,功能模块不断丰富。特别是最近几年,来自人工智能,机器学习及信号处理领域的各种方法的引入为该领域注入了新的活力。目前,多数基于数据     

基于马氏距离局部离群因子方法的复杂化工过程故障检测

为了满足实际的生产需要,复杂化工过程往往包含多个运行模态。同时过程的复杂性使得同一模态下的数据分布是一种高斯分布和非高斯分布混合存在的不确定情况。数据的多模态分布特性以及同一模态下数据分布的不确定性使得传统多元统计监控(MSPM)方法很难给出令人满意的结果。针对这一问题,本文提出一种新的马氏距离局部离群因子(MDLOF)方法进行故障检测。通过利用马氏距离挖掘变量局部结构中包含的有用信息,并对样本的邻域密度加以考虑,形成对数据分布具有鲁棒性的基于密度的监控指标。最后通过数值仿真例子及Tennessee Eastman过程验证其有效性。     

基于核Fisher判别分析方法的非线性统计过程监控与故障诊断

化工过程中大量的生产数据反应了生产过程的内在变化和系统的运行状况,基于数据驱动的统计方法可以有效地对生产过程进行监控。对于复杂的化工和生化过程,其过程变量之间的相关关系往往具有很强的非线性特性,传统的线性统计过程监控方法显得无能为力。本文提出了基于核Fisher判别分析的非线性统计过程监控方法,首先利用非线性核函数将数据从原始空间映射到高维空间,在高维空间中利用线性的Fisher判别分析方法提取数据最优的Fisher特征矢量和判别矢量来实现过程监控与故障诊断,能有效地捕获过程变量之间的非线性关系,通过对流化催化裂化（FCCU）过程的仿真表明该方法的有效性。     

基于仿射传播聚类子集主元分析的间歇过程监测方法

针对间歇过程固有的多阶段特性,也为了克服传统阶段划分方法严格按照物理时刻顺序将采样点硬性分割而不能使其寻找数据特征最为相近的聚类中心的严重缺陷,提出基于仿射传播聚类(AP)的子集多向主元分析(subset-MPCA)监测新方法:采用全新的乱序聚类思想,将时间片矩阵打乱用AP进行无约束乱序聚类,使样本突破时间顺序的约束自由找寻与其特征最为相近的聚类中心,获得聚类子集,建立精确的子集MPCA监控模型。在线监控时,引入信息度传递实现实时采样点的阶段归属判断,解决阶段不等长批次的最佳模型选择问题。对青霉素仿真数据的实验表明,该方法较传统方法可有效降低故障的漏报和误报,有着更加可靠的监控性能。     

Knowledge based models for the analysis of complex separation processes

The paper describes a conceptual formulation to screen, scope and review design options of complex separation systems without the need for extensive regression experiments. It employs a supertask representation with new synthesis formulation that exploits thermodynamics and engineering insights of primary separation. The thermodynamic and separation efficiency is systematically assessed with the prepositions of general terms, ‘conceptual losses’. These terms explicitly account for the competing design drives of the problem and enables trade-offs to become clear in the optimal solution. In all cases, the approach guarantees simple mathematical models that one can solve to global optimality. Furthermore, the approach provides venues to interpret the results and explain the layouts selected by the optimisation. The approach is illustrated with two different examples and results are explained by reviewing important conceptual terms. Indeed, it provides a unique opportunity to understand and build confidence in the selected options.     

基于深度学习的复杂化工过程软测量模型研究与应用

针对复杂化工生产过程中的一些原材料消耗量难以直接测量的问题，提出了一种基于深度学习的软测量方法。该方法基于一段时间的历史数据，利用平稳小波变换提取历史数据中的多尺度信息，然后与每一个时间点的可观测数据进行合并得到完整的数据集，再划分出训练集和测试集，用带有注意力机制的深度学习算法进行训练和泛化，进而建立软测量模型。最后将提出的方法应用到对苯二甲酸（PTA）生产装置乙酸消耗的软测量中。通过与极限学习机 (extreme learning machine，ELM)、多层感知器 (multi-layer perceptron，MLP)以及普通长短期记忆网络(long short-term memory，LSTM)方法比较，结果表明，该模型的预测准确度较高，具有一定的有效性和适用性，同时对PTA生产装置的乙酸消耗量进行预测分析，从而提高产能和降低能耗。     

Framework of dynamic simulation for complex chemical processes

The current state of the art of a modelling and dynamic simulation system for complex chemical and biochemical processes is discussed. Process modelling activity involves modelling a physical plant and external tasks imposed on the plant, and details of both aspects are discussed. Typical software structure is concerned with a model builder, result analyser, translator, solution methods, model library and external software interface. Some of them are explained in moderate depth. Recent progress of functionality and numerical methods is presented. Numerical methods incorporating symbolic and structural techniques improve accuracy and efficiency. In order to illustrate benefits of employing dynamic simulation tools, one typical chemical process consisting of a mixing tank, tubular reactor and gas absorber is chosen and dynamic simulation is carried out. Taking into account the work in this paper, some suggestions for future development of a unified framework of a modelling package are made.     

Control strategies for complex chemical processes. Applications in polymerization processes

The proper implementation of advanced control schemes for complex chemical processes heavily rely on the availability of appropriate mathematical models. The main objective of this paper is to show through examples how process models can be inserted into advanced controllers to allow the successful control of the process when the controlled variables are not measured or are measured infrequently. The control strategies are illustrated with actual data obtained for two typical polymerization processes. In the first example, a control algorithm is designed and implemented experimentally for the simultaneous closed-loop control of composition and average molecular weight of a copolymer latex. In the second case, a control scheme is designed for the simultaneous control of polymer production, polymer composition and melting index (MI) in a solution ethylene Ziegler-Natta polymerization process.     

Systematic controllability analysis for chemical processes

Modern chemical industrial processes are becoming more and more integrated and consist of multiple interconnected nonlinear process units. These strong interactions profoundly complicate a system's inherent properties and further alter the plant‐wide process dynamics. This may lead to a poor control performance and cause plant‐wide operability problems. To ensure entire processes run robustly and safely, with considerable profitability, it is crucial to recognize the inherent characteristics that can jeopardize controllability and process behavior at the early design stage. With a focus on inherently safer designs, from a plant‐wide perspective, a systematic method for chemical processes controllability analysis is addressed in this study. In the proposed framework, based on open‐loop stability/instability and minimum/nonminimum‐phase behavior, the entire operating zone of the process can be categorized into distinct subregions with different inherent properties. Variations in the inherent characteristics of a plant‐wide process with the operation and design conditions, over the feasible operation region, can be probed and analyzed. An attempt of this framework is made to illustrate how to clarify the roots of the poor controllability that arise in the design and operation of a large scale chemical process, and the results can provide guidance for both deciding the optimal operation conditions and selecting the most suitable control structure. Singularity theory is also applied in the framework to improve the computational efficiency. The framework is illustrated with two case studies. One involves a reactor‐external heat exchanger network and the other a more complex plant‐wide process, comprising a reactor, an extractor, and a distillation column. © 2012 American Institute of Chemical Engineers AIChE J, 58: 3096–3109, 2012     

Statistical monitoring of dynamic processes based on dynamic independent component analysis

Most multivariate statistical monitoring methods based on principal component analysis (PCA) assume implicitly that the observations at one time are statistically independent of observations at past time and the latent variables follow a Gaussian distribution. However, in real chemical and biological processes, these assumptions are invalid because of their dynamic and nonlinear characteristics. Therefore, monitoring charts based on conventional PCA tend to show many false alarms and bad detectability. In this paper, a new statistical process monitoring method using dynamic independent component analysis (DICA) is proposed to overcome these disadvantages. ICA is a recently developed technique for revealing hidden factors that underlies sets of measurements followed on a non-Gaussian distribution. Its goal is to decompose a set of multivariate data into a base of statistically independent components without a loss of information. The proposed DICA monitoring method is applying ICA to the augmenting matrix with time-lagged variables. DICA can show more powerful monitoring performance in the case of a dynamic process since it can extract source signals which are independent of the auto- and cross-correlation of variables. It is applied to fault detection in both a simple multivariate dynamic process and the Tennessee Eastman process. The simulation results clearly show that the method effectively detects faults in a multivariate dynamic process.     

Recent progress on equation-oriented optimization of complex chemical processes

Process optimization in equation-oriented (EO) modeling environments favors the gradient-based opti-mization algorithms by their abilities to provide accurate Jacobian matrices via automatic or symbolic differentiation.However,computational inefficiencies including that in initial-point-finding for Newton type methods have significantly limited its application.Recently,progress has been made in using a pseudo-transient (PT) modeling method to address these difficulties,providing a fresh way for-ward in EO-based optimization.Nevertheless,research in this area remains open,and challenges need to be addressed.Therefore,understanding the state-of-the-art research on the PT method,its principle,and the strategies in composing effective methodologies using the PT modeling method is necessary for further developing EO-based methods for process optimization.For this purpose,the basic concepts for the PT modeling and the optimization framework based on the PT model are reviewed in this paper.Several typical applications,e.g.,complex distillation processes,cryogenic processes,and optimizations under uncertainty,are presented as well.Finally,we identify several main challenges and give prospects for the development of the PT based optimization methods.     

Multivariate time delay analysis based local KPCA fault prognosis approach for nonlinear processes☆

Currently, some fault prognosis technology occasionally has relatively unsatisfied performance especially for in-cipient faults in nonlinear processes duo to their large time delay and complex internal connection. To overcome this deficiency, multivariate time delay analysis is incorporated into the high sensitive local kernel principal com-ponent analysis. In this approach, mutual information estimation and Bayesian information criterion (BIC) are separately used to acquire the correlation degree and time delay of the process variables. Moreover, in order to achieve prediction, time series prediction by back propagation (BP) network is applied whose input is multivar-iate correlated time series other than the original time series. Then the multivariate time delayed series and future values obtained by time series prediction are combined to construct the input of local kernel principal component analysis (LKPCA) model for incipient fault prognosis. The new method has been exemplified in a sim-ple nonlinear process and the complicated Tennessee Eastman (TE) benchmark process. The results indicate that the new method has superiority in the fault prognosis sensitivity over other traditional fault prognosis methods. ? 2016 The Chemical Industry and Engineering Society of China, and Chemical Industry Press. Al rights reserved.     

Probabilistic monitoring of chemical processes using adaptively weighted factor analysis and its application

As a probabilistic statistical method, factor analysis (FA) has recently been introduced into process monitoring for the probabilistic interpretation and performance enhancement of noisy processes. Generally, FA methods employ the first several factors that are regarded as the dominant motivation of the process for process monitoring; however, fault information has no definite mapping relationship to a certain factor, and useful information might be suppressed by useless factors or submerged under retained factors, leading to poor monitoring performance. Weighted FA (WFA) for process monitoring is proposed to solve the problem of useful information being submerged and to improve the monitoring performance of the GT² statistic. The main idea of WFA is firstly building a conventional FA model and then using the change rate of the GT² statistics (RGT²) to evaluate the importance of each factor. The important factors tend to have larger RGT² values, and the larger weighting values are then adaptively assigned to these factors to highlight useful fault information. Case studies on both a numerical process and the Tennessee Eastman process demonstrate the effectiveness of the WFA method. Monitoring results indicate that the performance of the GT² statistic is improved significantly compared with the conventional FA method.     

基于深度残差网络的化工过程故障诊断

本工作提出了一种基于深度残差网络(DRN)的化工过程故障诊断方法,可从大量化工过程运行数据中自动提取故障特征。模型采用快捷连接缓解传统深度神经网络训练困难的问题,且使用批归一化(BN)方法,可有效缓解梯度消失/爆炸的问题。以田纳西?伊斯曼(TE)过程为实验对象对所提方法进行诊断性能评价实验,并与以往的基于传统深度学习模型的TE过程故障诊断方法进行对比,进一步探究了模型层数、BN技术和残差结构对故障诊断率的影响,最后,通过t分布随机邻域嵌入(t-SNE)方法对网络部分层的输出进行可视化。结果表明,模型对21种工况取得了94%的平均故障诊断率和0.30%的平均假阳率,表现出更加卓越的诊断性能。输出层的二维散点图显示了清晰的聚类,表明所提出的DRN模型能够对故障进行准确诊断。     

一种基于聚类方法的多阶段间歇过程监控方法

针对阶段不等长的多阶段间歇过程,提出了一种基于k-均值聚类方法的阶段分段策略,可以将不等长的阶段准确分类。首先,将间歇过程的三维训练数据按变量方向展开成二维矩阵,再通过k-均值聚类的方法按照相关性将数据聚成多类并运用主元分析（PCA）方法分别对每一类建立模型。在线监控时,通过计算样本与模型之间的相似系数以选择最合适的模型进行在线监控。此方法可以将不同批次在同一采样时刻的过程数据按照相关性分到多个阶段,更符合生产过程中常见的过程数据阶段不等长的情况。最后利用青霉素仿真验证了该方法的有效性。