深度融合特征提取网络及其在化工过程软测量中的应用

复杂化工过程的观测数据往往同时包含非线性和强动态特性,而传统的化工过程软测量方法无法准确提取观测数据的非线性动态特征,以至影响数据建模和质量预报的准确性。提出了一种基于变分自编码器的深度融合特征提取网络(deep fusion features extraction network, DFFEN)。在变分自编码器框架下,通过构建潜隐特征信息传递通道,提取非线性动态潜隐变量。并利用自注意力机制(self-attention)融合关键的隐层信息,优化因信息传递通道过长而导致的潜在特征被遗忘的问题。此外,在后端网络构建潜隐变量和关键质量变量之间的回归模型,以实现关键质量变量的预报。最后,通过数值案例和实际的合成氨过程验证了所提出的DFFEN模型的可行性和有效性。     

基于注意力动态卷积自编码器的发酵过程故障监测

发酵过程的状态监测对于及时发现各类异常故障起到了至关重要的作用。然而，由于发酵过程数据呈现非线性特性，导致在提取特征信息时存在困难，增加了故障监测的难度。为了解决上述问题，提出了一种基于注意力动态卷积自编码器（attention dynamic convolutional autoencoder, ADCAE）的发酵过程故障监测方法。首先，设计了一种动态卷积结构（dynamic convolution structure），该结构可以在浅层使用大尺寸卷积核提取低级特征，在深层使用小尺寸卷积核提取高级特征，从而拓宽了模型特征学习的尺度；其次，设计了一种通道卷积注意力（channel convolutional attention, CCA）模块，该模块能够从不同尺度提取输入的非线性特征，并且在通道向量转化为权重的过程中可以更好地提取局部特征，提高了对有效信息的关注能力；最后，将动态卷积结构与CCA模块融入卷积自编码器中，使模型能够有效地捕获变量中的非线性关系，从而更好地应对发酵过程中的故障监测问题。利用青霉素发酵过程仿真平台和大肠埃希菌实际生产数据对该方法的可行性进行了验证，结果表明该方...     

基于VAE-OCCA的质量相关故障检测方法研究

由于闭环反馈系统的存在，并不是所有故障均会导致质量发生恶化。质量变量通常难以获得或具有一定的延迟，传统的无监督方法不能在检测过程是否正常的同时判断故障对质量的影响。典型相关分析(canonical correlation analysis,CCA)是一种经典的有监督方法，可以考虑输入输出间的关系，已被用于质量相关故障检测。然而，过程数据存在着维度高、非线性等问题，流程系统的复杂性使得CCA对于隐藏特征的捕获更具挑战性。提出了一种变分自编码器-正交典型相关分析(variational automatic encoder-orthogonal CCA,VAE-OCCA)方法。首先，利用变分自编码器对输入数据进行无监督自适应学习，实现对高维非线性过程变量的特征提取；进而，基于典型相关分析方法考虑输入输出关系，利用得到的相关系数矩阵进行奇异值分解建立质量相关和质量无关监测统计量；最后，通过工业案例测试说明提出方法的有效性及优越性。     

基于VAE-DBN的故障分类方法在化工过程中的应用

针对化工过程高维数据的故障特征难以提取的难题,提出变分自动编码器(VAE)结合深度置信网络(DBN)的混合故障诊断方法. 在VAE的编码部分对隐变量空间Z添加约束,通过重参数化方法进行反向传播训练,可无监督地学习不同故障对应的隐变量特征,其作为DBN分类模型的输入特征训练网络,输入测试集进行故障诊断. 田纳西伊斯曼流程(TE)应用结果表明,VAE能提取原始数据更加抽象有效的特征,VAE?DBN分类准确.     

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。关键词:主成分分析;故障检测;缺失数据;过程系统     

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。     

基于DMVU-OCSVM的故障诊断方法

针对工业过程的非线性和动态特性,提出一种基于动态最大方差展开（DMVU）和单类支持向量机（OCSVM）的故障诊断方法DMVU-OCSVM。为了分析数据的动态特性和非线性,应用流形学习技术DMVU提取数据变量中的非线性动态流形特征。基于所提取的流形特征信息建立OCSVM统计模型,并构造非线性监控统计量实时检测过程故障。在连续搅拌反应器（CSTR）系统上的仿真结果说明,本文提出的方法能够比OCSVM更有效地检测过程故障。     

基于循环自动编码器的间歇过程故障监测

针对间歇过程数据非线性、动态性特征，提出一种基于循环自动编码器(recurrent autoencoder，RAE)的过程故障监测方法。采用长短时记忆(long short-term memory，LSTM)循环神经网络构建自动编码器建立监控模型，相比传统自动编码器，其能有效挖掘时序样本间的动态关联信息。该方法首先利用批次展开与变量展开相结合的三步展开方法将间歇过程数据展开成二维，并通过滑动窗采样得到模型输入序列；然后使用LSTM构建自动编码器，重构输入序列。进一步，利用重构误差构造平方预测误差(squared prediction error, SPE)统计量实现在线监测。最后将所提方法应用于青霉素发酵仿真和重组大肠杆菌发酵过程监测，结果表明，该方法能及时监测到故障，具有较好的监测性能。     

动态因子分析模型及其在过程监控中的应用

在复杂工业系统的监控中,因子分析（FA）方法不需要专业的机理知识,应用系统日常运行数据建立模型,充分考虑了模型误差的普遍意义,具有较大的推广价值。针对实际过程的动态特性,基于自回归（AR）方式扩展过程变量数据矩阵,本文提出一种动态因子分析（DFA）的数据建模方法,充分提取了变量的自相关信息和互相关信息。另一方面,将DFA引入过程监控中,构建统计量作为监控指标,分别衡量变量的特征信息和误差信息,从而实现对动态过程运行状况的监控与评估。在Tennessee-Eastman（TE）过程中的应用研究,反映了这种方法的优越性。     

一种改进的动态核主元分析故障检测方法

针对复杂工业系统动态非线性故障检测过程精度低和计算量大的问题,提出了一种改进的动态核主元分析故障检测方法,该方法首先利用不可区分度剔除相关程度较小或者不相关变量,减少数据量,然后通过观测值扩展对筛选后的新数据构建增广矩阵,并对矩阵使用核主元分析提取变量数据的非线性空间相关特征,最后通过监测T2和SPE两种统计量诊断出系统发生故障及识别故障变量。仿真实验证明,该方法能对风力发电机故障进行有效监测和诊断,与KPCA方法相比,改进的动态核主元分析方法对微小故障更为敏感。     

Deep neural network based recursive feature learning for nonlinear dynamic process monitoring

The data collected from modern industrial processes always have nonlinear and dynamic characteristics. The recently developed deep neural network method, stacked denoising auto-encoder (SDAE), can extract robust nonlinear latent variables from data against noise. However, it leaves the dynamic relationship unconsidered. To solve this problem, a novel algorithm named the recursive stacked denoising auto-encoder (RSDAE) is proposed. To learn the dynamic relationship, the RSDAE focuses on the predictability of the latent variables in the recurrence to contain the most dynamic variations. After the dynamic variations are extracted by the RSDAE, there is little autocorrelation left in the residuals. Then, the residuals can be monitored by principal component analysis (PCA). For the purpose of process monitoring, corresponding fault detection statistics are developed based on the RSDAE. Finally, a numerical case and the Tennessee Eastman process benchmark are used to demonstrate the effectiveness of the proposed algorithm.     

A novel robust dynamic method for NOx emissions prediction in a thermal power plant

NOx is a harmful by-product of coal-fired boilers, and accurate prediction of NOx emissions in the outlet of a boiler is essential for environmental protection. In recent years, data-driven models have been widely studied and applied in this area. However, dynamic characteristics are ignored by many existing models, leading to sub-optimal performance. Besides, outliers that occur in the operation data have adverse effects on the efficacy of these prediction models. To address these issues, this paper presents a novel method for predicting NOx concentration via integrating a robust dynamic probabilistic approach and the long short-term memory (LSTM). First, mutual information (MI) is applied to determine the input variables. Subsequently, a robust probabilistic method is proposed to extract dynamic latent features considering outliers. On this basis, the generated latent variables are further utilized to train the LSTM-based model, with which the intrinsic relation between inputs and NOx values are obtained. Based on the application to a 660 MW thermal power plant, the superiority of the proposed method is demonstrated in terms of high prediction accuracy.     

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.     

Data-driven prediction model of indoor air quality in an underground space

Several data-driven prediction methods based on multiple linear regression (MLR), neural network (NN), and recurrent neural network (RNN) for the indoor air quality in a subway station are developed and compared. The RNN model can predict the air pollutant concentrations at a platform of a subway station by adding the previous temporal information of the pollutants on yesterday to the model. To optimize the prediction model, the variable importance in the projection (VIP) of the partial least squares (PLS) is used to select key input variables as a preprocessing step. The prediction models are applied to a real indoor air quality dataset from telemonitoring systems data (TMS), which exhibits some nonlinear dynamic behaviors show that the selected key variables have strong influence on the prediction performances of the models. It demonstrates that the RNN model has the ability to model the nonlinear and dynamic system, and the predicted result of the RNN model gives better modeling performance and higher interpretability than other data-driven prediction models.     

基于核慢特征回归与互信息的常压塔软测量建模

针对工业过程的非线性及动态特性,提出了一种新的慢特征回归软测量方法。该方法首先通过添加时延数据构造动态数据集,利用互信息最大化准则筛选变量从而减少信息冗余的影响。同时该方法在慢特征分析的基础上引入核函数扩展,加强模型处理非线性数据的能力,并将获得的核慢特征用于回归建模。核慢特征分析通过分析样本的变化,提取具有缓慢变化特征的成分,可以有效地刻画工业过程的变化趋势,提升回归模型精度。最后该方法的有效性在常压塔常顶油干点与常一线初馏点的软测量模型中得到了验证。     

基于GSA-LSTM动态结构特征提取的间歇过程监测方法

间歇过程监测对于保证批次生产过程的稳定运行具有重要意义。传统过程监测方法难以提取间歇过程数据特有的非线性结构和动态时变特征。为此,提出了一种融合图采样聚合网络和长短期记忆网络(GSA-LSTM)的典型相关分析方法用于间歇过程在线监测。首先,利用K近邻方法将批次过程数据转化为图结构形式,利用图采样聚合网络(GraphSAGE)提取数据内部的结构特征,然后利用长短期记忆网络(LSTM)提取数据的非线性动态特征,通过权重系数将结构特征和动态特征融合得到更具有代表性的间歇过程数据特征。进一步地,利用典型相关分析方法对残差建立监测模型。最后将所提方法应用于数值例子和注塑过程监测,结果分析验证了所提方法的有效性。     

基于GSA-LSTM动态结构特征提取的间歇过程监测方法

间歇过程监测对于保证批次生产过程的稳定运行具有重要意义。传统过程监测方法难以提取间歇过程数据特有的非线性结构和动态时变特征。为此,提出了一种融合图采样聚合网络和长短期记忆网络(GSA-LSTM)的典型相关分析方法用于间歇过程在线监测。首先,利用K近邻方法将批次过程数据转化为图结构形式,利用图采样聚合网络(GraphSAGE)提取数据内部的结构特征,然后利用长短期记忆网络(LSTM)提取数据的非线性动态特征,通过权重系数将结构特征和动态特征融合得到更具有代表性的间歇过程数据特征。进一步地,利用典型相关分析方法对残差建立监测模型。最后将所提方法应用于数值例子和注塑过程监测,结果分析验证了所提方法的有效性。     

Fault monitoring based on mutual information feature engineering modeling in chemical process

A large amount of information is frequently encountered when characterizing the sample model in chemical process. A fault diagnosis method based on dynamic modeling of feature engineering is proposed to effectively remove the nonlinear correlation redundancy of chemical process in this paper. From the whole process point of view, the method makes use of the characteristic of mutual information to select the optimal variable subset. It extracts the correlation among variables in the whitening process without limiting to only linear correlations. Further, PCA (Principal Component Analysis) dimension reduction is used to extract feature subset before fault diagnosis. The application results of the TE (Tennessee Eastman) simulation process show that the dynamic modeling process of MIFE (Mutual Information Feature Engineering) can accurately extract the nonlinear correlation relationship among process variables and can effectively reduce the dimension of feature detection in process monitoring.