选择性集成LTDGPR模型的自适应软测量建模方法

随着时间的增加,传统时间差(TD)模型会出现性能显著下降的问题。为了提高TD模型的可靠性和预测精度,同时考虑过程的时滞特征,基于一种选择性集成策略,提出一种局部时间差高斯过程回归(LTDGPR)模型的自适应软测量建模方法。首先,提取出数据库中的时滞动态信息,对建模数据进行重构;然后,采取局部化策略对差分后的重构样本进行统计划分,得到LTDGPR模型集。对于新来的输入样本,选择部分泛化能力强的LTDGPR模型进行集成,估计出含一定时间差的主导变量动态偏移值;最后,基于TD模型思想对当前时刻主导变量值进行在线预测。通过脱丁烷塔过程的数据建模仿真研究,验证了所提方法的有效性和精度。     

基于深度集成支持向量机的工业过程软测量方法

基于支持向量机（SVM）的软测量建模方法已经在工业过程控制领域得到广泛应用,然而传统支持向量机直接针对原始测量变量建立模型,未能充分挖掘数据的内在特征信息以提高预测精度。针对该问题,本文提出一种基于深度集成支持向量机（DESVM）的软测量建模方法。该方法首先利用深度置信网络（DBN）来对数据进行深层次的信息挖掘,提取出数据的内在特征,然后引入基于Bagging算法的集成学习策略,构建基于深度数据特征的集成支持向量机模型,以提升软测量预测模型的泛化能力。最后通过数值系统和真实工业数据对方法进行应用分析,结果表明本文提出的方法能够有效提升支持向量机软测量模型的预测精度,能够更好地预测过程质量指标的变化。     

基于ART-SVR的过程建模及在干点软测量中的应用

针对石油化工生产过程通常呈高度非线性,且生产过程数据呈非连续、具有一定类别特性等特征,提出基于自适应谐振神经网络（adaptive resonance theory,ART）和支持向量回归（support vector regression, SVR）相结合的建模方法（ART-SVR）。首先,基于建模样本,通过ART将样本模式空间分割成若干模式特性相近的子空间;然后,对各子空间分别采用SVR建立各自模型,实现基于样本模式空间分割的“分段”建模。仿真试验和在石脑油干点软测量建模的实际应用表明：ART-SVR模型的拟合精度和预测精度均优于全局SVR模型。     

基于半监督集成即时学习的橡胶混炼过程门尼黏度软测量研究

针对工业橡胶混炼过程中门尼黏度标记数据有限,导致模型预测性能受限的问题,提出了一种半监督(SS)集成即时学习(EJIT)高斯过程回归(GPR)软测量方法,称为SSEJITGPR。当查询样本到来时,该方法通过在线迭代学习的方式获取高置信度伪标记样本,其中使用集成后的即时学习高斯过程回归(JITGPR)模型对非标记样本进行预测,并以集成预测方差作为置信度评价准则。随后,基于伪标记样本扩充后的建模数据库构建多样性的半监督JITGPR基模型。最后,采用有限混合机制实现基模型的自适应集成。与传统门尼黏度软测量方法相比,SSEJITGPR在处理局部过程特征、克服标记样本不足、预测可靠度不高等问题上表现出显著优势,其有效性和优越性通过工业案例进行了验证。     

局部最小二乘支持向量机回归在线建模方法及其在间歇过程的应用

当间歇生产切换于不同的工艺条件时,由于新工况下的样本一般很少,且批次间存在着不确定性（由于原材料波动或过程动态特性波动等）,基于全局学习的建模方法（如最小二乘支持向量机回归,LSSVR）建立的模型泛化性能不强。将局部学习融入LSSVR中,提出一种局部LSSVR（local LSSVR, LLSSVR）的间歇过程在线建模方法。结合前一批次离线优化后的LSSVR参数,针对待预测新样本在线选择与之相关的近邻样本集并基于此进行建模。以建立青霉素发酵过程的菌体浓度为例,验证了LLSSVR算法能够从过程的第2个生产批次开始在线建立较准确的预报模型,较LSSVR有着更好的推广能力、适应性和鲁棒性。     

基于深度集成支持向量机的工业过程软测量方法

基于支持向量机(SVM)的软测量建模方法已经在工业过程控制领域得到广泛应用,然而传统支持向量机直接针对原始测量变量建立模型,未能充分挖掘数据的内在特征信息以提高预测精度。针对该问题,本文提出一种基于深度集成支持向量机(DESVM)的软测量建模方法。该方法首先利用深度置信网络(DBN)来对数据进行深层次的信息挖掘,提取出数据的内在特征,然后引入基于Bagging算法的集成学习策略,构建基于深度数据特征的集成支持向量机模型,以提升软测量预测模型的泛化能力。最后通过数值系统和真实工业数据对方法进行应用分析,结果表明本文提出的方法能够有效提升支持向量机软测量模型的预测精度,能够更好地预测过程质量指标的变化。     

基于FEEMD-AE与反馈极限学习机组合模型预测研究与应用

针对现有工业过程非平稳时间序列中的特征提取及预测问题,提出了基于快速集合经验模态分解（fast ensemble empirical mode decomposition,FEEMD）、近似熵（approximate entropy,AE）和反馈极限学习机（feedback extreme learning machine,FELM）的组合模型。首先,针对复杂非平稳时序数据,采用FEEMD方法将其分解为从高频到低频的相对平稳的本征模态函数分量和余项;其次,为解决经过FEEMD分解出来的分量复杂度问题,运用近似熵（AE）计算分量复杂度并进行特征重构,以降低分量复杂性;然后,基于传统ELM结构,通过引入反馈机制,在输出层与隐含层之间增加反馈层用来记忆隐含层输出数据,并计算数据趋势变化率动态更新反馈层输出,形成反馈极限学习机（FELM）,对非线性动态系统的下一时刻输出进行预测;最后,将所提出的组合预测模型通过UCI标准数据集与精对苯二甲酸（PTA）溶剂系统进行建模仿真,仿真结果表明,提出的组合模型预测方法能够得到较高的预测精度,为实际生产操作优化提供了一定的指导。     

基于最近邻与神经网络融合模型的软测量建模方法

软测量建模能够有效地解决生产过程中在线分析仪表测量滞后大、价格昂贵、维护保养复杂等问题。目前,基于数据驱动的神经网络是软测量建模的主要工具之一。而在建模数据的采集过程中,主导变量的采集相对辅助变量要困难得多,由此产生了大量缺失标签的数据。但传统的软测量建模方法却忽视了这些无标签数据,只利用少量的有标签数据建模,从而影响了模型的预测精度。为了解决标签缺失的问题,采用最近邻算法对无标签数据进行伪标记,同时设计了由卷积操作与门限循环单元神经网络(GRU)结合的网络结构来进一步利用无标签数据,提取不同时刻数据中的动态特征,提高神经网络的预测精度。最后将该方法应用于丙烯精馏塔塔顶丙烷浓度的预测,实验结果表明该模型能有效处理非线性动态系统的标签缺失问题,具有更高的预测精度。     

基于Gath-Geva算法和核极限学习机的多阶段间歇过程软测量

间歇过程具有较强的非线性,多阶段、慢时变及批次间存在变化,采用单一预测模型不能反映间歇过程的多阶段特性及阶段间过渡特性。提出一种基于Gath-Geva聚类和核极限学习机（kernel extreme learning machine,KELM）的多模型软测量方法。首先采用主成分分析（principal component analysis,PCA）对输入做特征提取,然后利用Gath-Geva算法对间歇过程进行多阶段工况划分,根据生产工况特性划分为不同的操作阶段后,分别建立局部KELM模型。对任一待预测样本,分别计算其对应各局部模型的预测值,最后采用贝叶斯集成,将其隶属于各局部模型的模糊隶属度作为权重和预测值融合得到最终预测值。以青霉素发酵数据进行实验测试,结果表明所提多模型算法相较于单一模型,具有更高的预测精度。     

用于发酵过程在线建模的自适应局部最小二乘支持向量机回归方法

提出一种基于自适应局部学习的最小二乘支持向量机回归（LSSVR）在线建模方法。考虑样本间的距离和角度信息以获得更全面合理的相似样本集,推导了采用快速留一法在线优化模型参数的准则,并给出了发酵过程在线自适应模型选择的策略。以链激酶流加发酵过程为例,验证了所提出算法能够从过程的第2批次开始,同时对活性菌体浓度和链激酶浓度进行较准确的在线预报,较普通的局部LSSVR等建模方法具有更高的预报精度和自适应性。     

基于时间差分和局部加权偏最小二乘算法的过程自适应软测量建模

工业过程软测量模型常常因为过程的变量漂移、非线性和时变等问题而使得预测性能下降。因此,时间差分已被应用于解决过程变量漂移问题。但是,时间差分框架下的全局模型往往不能很好地描述过程非线性和时变等特性。为此,提出了一种融合时间差分模型和局部加权偏最小二乘算法的自适应软测量建模方法。时间差分模型可以大大减少过程变量漂移的影响,而局部加权偏最小二乘算法作为一种即时学习方法,可以有效解决过程非线性和时变问题。该方法的有效性在数值例子和工业过程实例中得到了有效验证。     

Application of online support vector regression for soft sensors

Soft sensors have been widely used in chemical plants to estimate process variables that are difficult to measure online. One of the crucial difficulties of soft sensors is that predictive accuracy drops due to changes in state of chemical plants. Characteristics of adaptive soft sensor models such as moving window models, just‐in‐time models and time difference models were previously discussed. The predictive accuracy of any traditional models decreases when sudden changes in processes occur. Therefore, a new soft sensor method based on online support vector regression (SVR) and the time variable was developed for constructing soft sensor models adaptive to rapid changes of relationships among process variables. A nonlinear SVR model with the time variable is updated with the most recent data. The proposed method was applied to simulation data and real industrial data, and achieved higher predictive accuracy than traditional ones even when time‐varying changes in process characteristics happen. © 2013 American Institute of Chemical Engineers AIChE J 60: 600–612, 2014     

Dynamic soft sensor development based on Gaussian mixture regression for fermentation processes

The dynamic soft sensor based on a single Gaussian process regression (GPR) model has been developed in fermentation processes.However,limitations of single regression models,for multiphase/multimode fermentation processes,may result in large prediction errors and complexity of the soft sensor.Therefore,a dynamic soft sensor based on Gaussian mixture regression (GMR) was proposed to overcome the problems.Two structure parameters,the number of Gaussian components and the order of the model,are crucial to the soft sensor model.To achieve a simple and effective soft sensor,an iterative strategy was proposed to optimize the two structure parameters synchronously.For the aim of comparisons,the proposed dynamic GMR soft sensor and the existing dynamic GPR soft sensor were both investigated to estimate biomass concentration in a Penicillin simulation process and an industrial Erythromycin fermentation process.Results show that the proposed dynamic GMR soft sensor has higher prediction accuracy and is more suitable for dynamic multiphase/multimode fermentation processes.     

基于集成学习的多产品化工过程软测量建模方法

化工过程通常具有非线性、时变以及多产品等特性。针对上述特点,在集成学习框架下建立自适应软测量模型。首先,面向具有多个产品的化工对象,借助k近邻法,以统计假设检验理论为依据,提出一种自适应局部化方法,获得多样性程度高的局部模型集合。然后,根据未知样本量化局部模型的泛化能力,通过选择性集成方法获得主导变量的估计值。此外,为了对主导变量估计值的精度进行评估,基于局部模型泛化误差,给出一种通用性高的模型性能评价方法。在仿真的盘尼西林生产过程上的运行结果验证了所提方法的有效性。     

自适应软测量方法在动液面预测中的研究与应用

针对传统人工检测方法在测量动液面时存在精度低、实时性差等问题,采用软测量技术来完成对动液面的测量工作.根据对现场数据特性的分析,提出采用经验模态分解和基于黑洞的最小二乘支持向量机预测相结合的算法来实现动液面软测量建模;通过构建模型性能评价模块,动态更新模型,解决在油田生产过程中,静态模型不能完全反映生产工况导致模型失效的问题,提高算法的自适应能力及预测量精度.最后通过对油田生产现场监测数据进行实验验证,结果表明,该方法对油田动液面测量精度高,对生产波动的自适应能力强,满足油田现场测试使用要求,提高油田生产自动化程度.     

Soft sensor model derived from Wiener model structure: modeling and identification

The processes of building dynamic and static relationships between secondary and primary variables are usually integrated in most of nonlinear dynamic soft sensor models. However, such integration limits the estimation accuracy of soft sensor models. Wiener model effectively describes dynamic and static characteristics of a system with the structure of dynamic and static submodels in cascade. We propose a soft sensor model derived from Wiener model structure, which is an extension of Wiener model. Dynamic and static relationships between secondary and primary variables are built respectively to describe the dynamic and static characteristics of system. The feasibility of this model is verified. Then the expression of discrete model is derived for soft sensor system. Conjugate gradi-ent algorithm is applied to identify the dynamic and static model parameters alternately. Corresponding update method for soft sensor system is also given. Case studies confirm the effectiveness of the proposed model, alternate identification algorithm, and update method.     

Adaptive soft sensor for online prediction and process monitoring based on a mixture of Gaussian process models

Linear models can be inappropriate when dealing with nonlinear and multimode processes, leading to a soft sensor with poor performance. Due to time-varying process behaviour it is necessary to derive and implement some kind of adaptation mechanism in order to keep the soft sensor performance at a desired level. Therefore, an adaptation mechanism for a soft sensor based on a mixture of Gaussian process regression models is proposed in this paper. A procedure for input variable selection based on mutual information is also presented. This procedure selects the most important input variables for output variable prediction, thus simplifying model development and adaptation. Apart from online prediction of the difficult-to-measure variable, this soft sensor can be used for adaptive process monitoring. The efficiency of the proposed method is benchmarked with the commonly applied recursive PLS and recursive PCA method on the Tennessee Eastman process and two real industrial examples.     

Soft sensor based on multi-phase stacking ensemble model with self-selected primary learner for batch processes

In batch processes, existing soft sensing methodologies encounter substantial challenges when confronted with nonlinearity and multi-phase issues. In response to these challenges, an innovative soft sensing framework known as the multi-phase stacking ensemble model with self-selected primary learner is proposed. The main innovation of this framework lies in the solution to the primary learner selection issue within the stacking model. To commence, the batch process is divided into multiple phases employing a Gaussian mixture model, thereby establishing local ensemble models for each phase. Subsequently, the iterative self-selection of primary learners strategy is proposed, which iteratively selects suitable primary learners for these models, optimizing their combination of primary learners for each local model. This primary learner selection strategy effectively enhances the accuracy of predictions in the stacking ensemble model. To further enhance the performance, Bayesian optimization is utilized to tune the hyperparameters of each local ensemble model. This step guarantees optimal performance of the model across diverse phases. Extensive simulation experiments are conducted on an industrial penicillin fermentation process to validate the effectiveness of the proposed framework. According to the findings, the model demonstrated superior performance compared to existing single-learner soft sensing methods and commonly utilized ensemble-based soft sensing methods in terms of both R² score (0.97584) and RMSE (0.0513). Overall, this framework offers a novel approach for selecting primary learners in stacking ensemble models and enhancing the predictive performance in batch processes for soft sensing.     

基于动态多核相关向量机的软测量建模研究

针对污水处理过程中存在的多变量耦合、强非线性以及参数时变等问题，提出基于多核学习相关向量机的软测量建模方法，并采用粒子群算法对多核权重以及核参数进行优化。同时，引入时间差分(time difference)方法改进多核相关向量机的动态特性。为了验证所提模型的有效性，通过一仿真案例与单核相关向量机、多层前馈神经网络和基于遗传算法的支持向量机进行对比研究。结果表明，所提模型具有更好的预测效果。最后，对模型的鲁棒性在数据漂移和异常的场景下进行了讨论。