神经网络建模方法在维生素C发酵过程中的应用

传统的标准“黑箱”型人工神经网络已较为广泛地应用于生化过程中的状态预估等多个方面,然而结合过程先验知识或部分机理模型的混合神经网络建模方法能给出更令人满意的结果.本文将其应用于2-酮基-l-古龙酸（2-KLG）发酵过程的状态估计,并将其结果与传统神经网络模型进行了比较,混合模型明显优于单一神经网络方法.     

发酵过程生物量软测量技术的研究进展

生物量是发酵过程中的关键过程参数之一,它直接影响着发酵过程的优化和控制。综述了近年来发酵过程生物量软测量技术的研究现状,讨论了基于过程机理分析、回归分析、状态估计和神经网络等的软测量建模方法,对基于神经网络和改进的神经网络建模方法进行了分析。指出基于多尺度建立软测量混合模型,是实现发酵过程生物量在线测量的有效方法,并给出了建立混合模型需要解决的关键问题。     

基于混合建模技术的聚氯乙烯粒径分布预测

针对聚氯乙烯粒径分布在线软测量问题,提出了一种基于机理分析和神经网络的混合建模方法,并将该建模方法应用于聚氯乙烯粒径分布建模研究中。混合模型由机理模型和误差补偿模型所组成。通过机理分析建立氯乙烯悬浮聚合过程的单体液滴群体平衡(Population Balance Equation,简称PBE)模型,由于聚氯乙烯成粒过程的复杂性和强非线性,单纯的机理模型预测与实际分析值相比仍存在一定偏差,因此利用人工神经网络建模方法建立了基于BP神经网络的单体液滴群体平衡模型修正模型,对单体液滴群体平衡模型的输出进行修正,由此建立起聚氯乙烯粒径分布混合模型。由于混合模型既能按照液滴分散与聚并机理对聚氯乙烯颗粒的成长过程进行描述,同时又充分利用了生产现场数据对模型误差进行修正,应用到聚氯乙烯生产过程的测试结果表明,与单纯机理模型相比,聚氯乙烯粒径分布混合模型具有更佳的预测效果。     

基于卷积神经网络的撞击流反应器浓度场混合特性

基于PLIF测试技术结合卷积神经网络技术提出混合性能预测方法，分析水平对置撞击流反应器浓度场混合特性，能准确预测其内部浓度场的混合均匀度及混合时间。基于卷积神经网络构建了混合性能预测模型，利用水平对置撞击流反应器浓度场实验数据对构建的模型进行有监督地训练并进行预测，预测结果显示对混合均匀度的预测准确率达95%，计算效率提高了99.99%。为更好地理解混合性能预测模型对混合均匀度的预测机理，本文对其卷积层输出进行可视化处理，通过功率谱分析卷积核的响应给出了撞击流反应器浓度场特征提取的物理解释。最后利用预测模型搭建混合均匀度快速获取系统并应用于撞击流混合特性研究。所提出的基于卷积神经网络的预测模型可以有效分析水平对置撞击流反应器的混合特性，预测模型可靠、适用范围广，为深度学习算法应用于撞击流领域提供了方案经验。     

基于混合粒子群优化的异类多模型软测量方法研究与应用

基于异类组合预测模型可提高模型的预测精度及鲁棒性的思想,提出一种基于混合粒子群优化的异类多模型非线性组合软测量建模的新方法。即先分别用混合粒子群优化的径向基函数神经网络、最小二乘支持向量机及部分最小二乘算法对训练集训练得出子模型,然后将具有性能互补性的三个子模型的输出作为反向传播网络的输入得到最后结果。用混合粒子群优化的方法来选取径向基函数神经网络和最小二乘支持向量机的模型参数,该方法克服了常用的交叉验证法耗时与盲目性问题。三层反向传播网络具有无限逼近特性,使得整个组合预测模型具有更好的泛化能力和预报精度。将其应用于汽油调合系统中研究法辛烷值的预测,仿真结果表明,该方法是可行且有效的。     

聚合反应过程质量指标的推理估计混合模型

针对聚合反应过程的非线性、时变性和不确定性，提出了一种多类型混联混合推理估计模型。该模型以过程机理知识为基础框架，以各种神经网络和回归辩识模型的计算结果作为混合模型中各子模型或机理模型的过程参数。为了体现过程的多模式集成特点，该混合模型充分利用各种类型模型的不同特性，既保证按照动力学规律描述聚合反应过程特性，又充分利用现场运行和分析的数据，辩识模型结构参数，使所建模型不必完全依赖对过程特性的认识。将该混合模型用于聚丙烯腈生产过程质量指标的推理估计，现场应用效果证明了这种模型的优良性能。     

神经网络遗传算法及其在乙烯裂解过程预测中的应用研究

针对复杂神经网络模型的结构特性和ＢＰ网络梯度下降法的缺点,将遗传算法与神经网络相结合,提出了基于改进连续化遗传算法的神经网络学方法,并将其应用于大型乙烯裂解产品分布预测,获得了较好的过于模拟结果。     

Prediction Model for Aromatization Reacting System of LPG Based on Back-Propagation Neural Network

将BP神经网络应用于液化石油气芳构化反应体系,建立了该体系在指定反应条件下的神经网络模型,网络的训练结果达到了预设要求,且网络输出结果与试验值能够较好的吻合,所建模型能较准确地预测指定反应条件下该体系的产物分布状况.分析结果可为LPG芳构化技术的进一步开发、生产过程中的产物分布预测和工业过程的模拟提供一些指导.     

基于SNNs-RR的聚丙烯熔融指数软测量

提出了一种组合神经网络-岭回归（SNNs-RR）建模方法,并将该方法应用于聚丙烯熔融指数软测量研究中.通过多个单一神经网络的合理组合可显著改善神经网络模型的泛化能力,而选择合适的组合权重对组合神经网络模型是否具有良好预测性能是至关重要的,因此提出了采用岭回归方法来选择合适的组合权重.通过与单一神经网络模型的预测结果进行比较,表明基于SNNs-RR的聚丙烯熔融指数软测量模型具有更佳的预测精度和鲁棒性.     

利用结构逼近式混合神经网络实现间歇反应器的建模

提出了一种新的混合神经网络建模方法——结构逼近式混合神经网络。基于此结构建立的混合神经网络可以充分利用已知非线性系统的结构信息,使神经网络“灰盒”化,更好地解释和描述系统各变量间的因果关系,从而提高网络的建模精度和模型的可靠性。本文介绍了这类神经网络的基本特性、拓扑结构和训练方法。报告了一个典型放热液相二级平行间歇反应的建模过程;并针对间歇反应过程测量滞后的情况,与两种不同的混合神经网络模型作了比较,仿真和比较结果证明了方法的有效性。     

Hybrid knowledge-based tool for mixing equipment selection – the fuzzy neural network

The hybrid knowledge-based system proposed in this paper consists of a “stiff” segment, viz. the expert system based on the object-oriented approach, and a flexible part, viz. the neural network. Some of the input parameters of the problem and output parameters of the “stiff” system are presented as the fuzzy numbers. Detailed information is also presented about the development of the neural network. The most evident advantages of the proposed introduction of a hybrid architecture of the knowledge-based system are a faster evaluation and generation of design alternatives and support of systematic searches and storage of experience. In addition, the resulting ability to extrapolate results would be unattainable with separately acting stiff and flexible systems. A system for the estimation of the parameters of a mixing system for wastewater treatment is presented as an example to illustrate the principles of the hybrid system.     

Artificial and Hybrid Fuzzy Linear Neural Network‐Based Estimation of Seed Oil Content of Safflower

Inexpensive and rapid methods for measurement of seed oil content by near infrared reflectance spectroscopy (NIRS) are useful for developing new oil seed cultivars. Adopting default multiple linear regression (MLR), the predictions of safflower oil content were made by 20–140 samples using a Perten Inframatic 8620 NIR spectrometer. Although the obtained interpolation results of MLR had desired accuracy, the extrapolation was extremely poor. The extrapolation determination coefficient (R²) and standard error (SE) of cross validation for MLR models were 0.63–0.78 and 3.71–4.44, respectively. In order to overcome the accuracy limitation of linear MLR models, a common suggestion is to use a nonlinear artificial neural network (ANN); however, it needs a large number of data to yield significant accurate results. We developed a novel robust hybrid fuzzy linear neural (HFLN) network to capture simultaneously linear and nonlinear patterns of data with a limited number of safflower samples. Empirical extrapolation results showed that the HFLN had higher R² (=0.85) and lower SE (=1.83) compared to those obtained by MLR and ANN models. It is concluded that hybrid methodologies could be used to construct efficient and appropriate models for estimation of seed oil content set up on NIR system.     

Dynamic On‐Line Reoptimization Control of a Batch MMA Polymerization Reactor Using Hybrid Neural Network Models

A hybrid neural network model based on‐line reoptimization control strategy is developed for a batch polymerization reactor. To address the difficulties in batch polymerization reactor modeling, the hybrid neural network model contains a simplified mechanistic model covering material balance assuming perfect temperature control, and recurrent neural networks modeling the residuals of the simplified mechanistic model due to imperfect temperature control. This hybrid neural network model is used to calculate the optimal control policy. A difficulty in the optimal control of batch polymerization reactors is that the optimization effort can be seriously hampered by unknown disturbances such as reactive impurities and reactor fouling. With the presence of an unknown amount of reactive impurities, the off‐line calculated optimal control profile will be no longer optimal. To address this issue, a strategy combining on‐line reactive impurity estimation and on‐line reoptimization is proposed in this paper. The amount of reactive impurities is estimated on‐line during the early stage of a batch by using a neural network based inverse model. Based on the estimated amount of reactive impurities, on‐line reoptimization is then applied to calculate the optimal reactor temperature profile for the remaining time period of the batch reactor operation. This approach is illustrated on the optimization control of a simulated batch methyl methacrylate polymerization process.     

自增长混合神经网络及其在燃料电池建模中的应用

为了提高非线性辨识的精度, 提出了一种基于混合算子的自增长混合神经网络。该神经网络通过自增长的混合隐含层结构, 包括加算子和乘算子, 形成神经元个数少、结果精确、增长快速的网络。论文在级联神经网络的结构基础上, 提出GQPSOI算法来引导神经网络的结构自增长以及权值更新。通过对燃料电池的建模与比较分析, 证明了方法的有效性和良好的应用前景。     

用基团贡献+神经网络+颗粒群优化相结合的方法来预测有机化合物的闪点温度

The flash points of organic compounds were estimated using a hybrid method that includes a simple group contribution method (GCM) implemented in an artificial neural network (ANN) with particle swarm optimization (PSO). Different topologies of a multilayer neural network were studied and the optimum architecture was determined. Property data of 350 compounds were used for training the network. To discriminate different substances the molecular structures defined by the concept of the classical group con-tribution method were given as input variables. The capabilities of the network were tested with 155 substances not considered in the training step. The study shows that the proposed GCM+ANN+PSO method represent an excellent alternative for the estimation of flash points of organic compounds with acceptable accuracy (AARD 1.8%; AAE 6.2 K).     

Hybrid modelling of yeast production processes – combination of a priori knowledge on different levels of sophistication

Process models are used to formulate knowledge about process behaviour. They are applied, e.g., to predict the process' future behaviour and for state estimation when reliable on-line measuring techniques to monitor the key variables of the process are not available. There are different sources of information available for modelling, which provide process knowledge in different representations. Some elements or aspects may be described by physically based mathematical models and others by heuristically obtained rules of thumb, while some information may still be hidden in the process data recorded during previous runs of the process. Heuristic rules are conveniently processed with fuzzy expert systems, while artificial neural networks present themselves as a powerful tool for uncovering the information within the process data without the need to transform the information into one of the other representations. Artificial neural networks and fuzzy technology are increasingly being employed for modelling biotechnological processes, thus extending the traditional way of process modelling by mathematical equations. However, a sufficiently comprehensive combination of all these techniques has not yet been put forward. Here, we present a simple way of combining all the available knowledge relating to a given process. In a case study, we demonstrate the development of a hybrid model for state estimation and prediction on the example of a yeast production process. The model was validated during a cultivation performed in a standard pilot-scale fermenter.     

Control of a Batch Polymerization System Using Hybrid Neural Network ‐ First Principle Model

In this work, the utilization of neural network in hybrid with first principle models for modelling and control of a batch polymerization process was investigated. Following the steps of the methodology, hybrid neural network (HNN) forward models and HNN inverse model of the process were first developed and then the performance of the model in direct inverse control strategy and internal model control (IMC) strategy was investigated. For comparison purposes, the performance of conventional neural network and PID controller in control was compared with the proposed HNN. The results show that HNN is able to control perfectly for both set points tracking and disturbance rejection studies.