基于仿射传播聚类的发酵过程建模

针对花生四烯酸(ARA)发酵过程复杂,机理模型表达不够准确以及单模型泛化能力弱的问题,提出采用基于仿射传播聚类的支持向量机(SVM)多模型建模算法进行该过程建模。该算法首先用仿射传播聚类(AP)算法对ARA样本数据进行聚类,再用SVM算法对各子类样本分别建立子模型。测试样本根据相似性的测度进行归类,并用所属子类的模型进行预测输出。ARA发酵过程的建模实验表明,与其他建模算法相比,基于仿射传播聚类的SVM多模型建模算法所建立的模型具有更高的回归精度和良好的泛化能力。     

基于PSO-SVM逆的赖氨酸发酵过程软测量

针对赖氨酸发酵过程非线性、大滞后、多变量动态耦合,关键生化参数难以实时在线测量等问题,提出一种改进的粒子群-支持向量机(PSO-SVM)逆发酵过程软测量建模方法。首先分析逆系统的存在性,并结合赖氨酸发酵过程,引入发酵特征信息和舍弃次要信息构造逆扩展模型;然后利用支持向量机离线辨识初始逆扩展模型,并根据系统输入与模型输出的偏差信号,采用粒子群算法对初始逆扩展模型进行在线校正;最后将校正后的逆扩展模型串联在原发酵过程之后构成复合伪线性系统,实现不直接可测关键生化参数的在线预测。以L-赖氨酸流加发酵过程为例,验证了所提算法能够对发酵过程关键生物量参数进行较准确的在线预测,较普通的SVM逆建模方法具有更高的预测精度。     

基于向量投影的代谢支持向量机乙烯精馏产品质量软测量建模

针对支持向量机(SVM)增量学习过程中易出现计算速度慢、稳定性差的缺陷,提出了一种基于向量投影的代谢支持向量机建模方法.该方法首先运用向量投影算法对训练样本进行预选取来减少样本数量,提高SVM建模速度.然后将新增样本"代谢"原则引入SVM增量学习过程中,以解决因新增样本不断加入而导致训练样本数量"爆炸"的问题.最后将该方法用于乙烯精馏产品质量软测量建模,实验结果表明,与传统SVM和最小二乘支持向量机(LSSVM)相比,向量投影的代谢SVM具有更好的预测结果.     

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。     

关键核网络及其在发酵过程在线建模中的应用

发酵过程通常采用流加补料操作,无稳态工作点、非线性强,且重要生物量往往无法在线测量。本文提出了一种适用于非线性多输入多输出的发酵过程在线建模方法：关键核网络（key kernel network,KKN）。结合过程的先验知识给出控制模型关键节点加入的准则,使其能自适应调整模型的复杂度,以提高建模的精度和速度,并给出了关键节点增加时KKN模型的在线递推形式。将KKN应用于青霉素发酵过程的在线建模,研究表明,KKN能同时快速、准确地预报菌体和产物浓度,且随着批次的增加,过程信息不断得到积累,模型精度逐渐提高。     

微生物发酵过程优化控制技术进展

微生物发酵过程优化控制技术是发酵工程的重要技术。综述了近年来微生物发酵过程优化控制技术的研究现状,讨论了机理分析建模、黑箱建模、混合建模等发酵过程建模方法,对基于模型的优化控制策略进行了分析。指出了基于混合模型和多目标优化策略建立动态优化控制器,是微生物发酵过程优化控制的有效方法,并给出了实现优化控制需要解决的关键问题。     

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。     

青霉素发酵过程中的混合建模

由于微生物发酵过程机理的复杂性和高度非线性,建立发酵过程的精确模型具有一定难度。传统的动力学模型预测虽然会与实际输出有一定偏差,但它在某种程度上反映了过程机理;神经网络建模方法属于"黑箱"方法,建模过程中没有用到任何先验知识,有一定的预测效果;因此单一的建模方法往往会不具备其他建模方法的优势。本文以传统的发酵动力学模型为基础,结合RBF神经网络进行混合建模的"灰箱"建模方法是一种比较好的建模思路,可以取得较满意的软测量效果。     

LSSVM过程建模中超参数选取的梯度优化算法

基于结构风险最小的最小二乘支持向量机（least squares support vector machine, LSSVM）为标准支持向量机（SVM）的约简;训练简易;性能良好。其模型精度受超参数影响;常规的网络搜索法很难搜得最佳超参数。在快速留一法的基础上;以全样本留一预测误差平方和最小化为目标;导出基于梯度的最优化算法;用以优选为LSSVM超参数;进而构建G-LSSVM模型。以柠檬酸发酵过程为算例对G-LSSVM进行检验;结果表明G-LSSVM的超参数选取耗时少;模型稳定性良好;且拟合和预报性能都优于标准SVM和神经网络。有望适用于机理不明、高度非线性、小样本的化工过程建模。     

与机理杂交的支持向量机为发酵过程建模

针对生物发酵过程机理复杂、高度非线性的特点,采用基于结构风险最小的支持向量机为发酵过程建模,其算法规范,建模复杂度低于神经网络方法,所建模型的预测效果更好.还将生化过程的动力学机理与支持向量机相结合,采用串联和串并联结构,提出与机理杂交的支持向量机建模方法,并为间歇式酒精发酵过程中酵母菌体浓度变化建立了预测模型.原理分析与试验结果表明与机理杂交的支持向量机建模方法,相比于单一近似的动力学模型、单一的支持向量机模型,以及机理杂交的神经网络模型,它的预测精度高,泛化能力强,性能更为优越.     

The Development of Intelligent Models for Liquid–Liquid Equilibria (LLE) Phase Behavior of Thiophene/Alkane/Ionic Liquid Ternary System

ABSTRACT

This paper develops three models based on artificial neural network (ANN), support vector machine (SVM) and least square support vector machine (LSSVM) algorithm for phase behavior of thiophene/alkane/ionic liquid ternary system. The shuffled complex evolution (SCE) was employed to acquire the optimal magnitudes of hyper parameters (σ² and γ) which are embedded parts of SVM and LSSVM models, and the trial and error was employed to obtain the optimal numbers of neuron and layers for ANN intelligent model. Gathering and using 618 LLE data, the comparison between the optimized version of applied intelligent models in giving the LLE was also made. The findings are indicative of a prefect agreement between the estimation from intelligent models and the experimental data. The finding also reveals that the performance of SVM in prediction of solubility is somewhat better than other intelligent models (i.e., ANN and SVM) as coefficient determination (R2) and root mean squared error (RMSE) are respectively 0.9961 and 0.0447 for test sets of data. This is likely due to the existence of structural risk minimization principle of SVM which is embodied in SVM algorithm and effectively minimizes upper bound of the generalization error, rather than minimizing the training error.     

Toward an intelligent approach for predicting surface tension of binary mixtures containing ionic liquids

Knowledge of the surface tension of ionic liquids (ILs) and their related mixtures is of central importance and enables engineers to efficiently design new processes dealing with these fluids on an industrial scale. It’s obvious that experimental determination of surface tension of every conceivable IL and its mixture with other compounds would be a herculean task. Besides, experimental measurements are intrinsically laborious and expensive; therefore, accurate prediction of the property using a reliable technique would be overwhelmingly favorable. To do so, a modeling method based on artificial neural network (ANN) trained by Bayesian regulation back propagation training algorithm (trainbr) has been proposed to predict surface tension of the binary ILs mixtures. A total set of 748 data points of binary surface tension of IL systems within temperature range of 283.1-348.15 K was used to train and test the applied network. The obtained results indicated that the predictive values and experimental data are quite matching, representing reliability of the used ANN model for such purpose. Also, compared with other methods, such as SVM, GA-SVM, GA-LSSVM, CSA-LSSVM, GMDH-PNN and ANN trained with trainlm algorithm the proposed model was better in terms of accuracy.     

Artificial neural network simulation of the condenser of seawater greenhouse in Oman

The prediction of freshwater production from the condenser of an agricultural seawater greenhouse is important for designing the greenhouse process. Two models, namely, Artificial Neural Network and multilinear regression (denoted as ANN and RA, respectively), were developed and tested to predict the freshwater production rate considering ambient solar intensity, condenser inlet moist-air temperature, humidity ratio and mass flowrate, and inlet coolant temperature. Statistical analysis indicated that all parameters significantly affected the prediction (p?<?0.05). The accuracy of the ANN and RA models was then compared to two models previously developed by Yetilmezsoy and Abdul-Wahab and Al-Ismaili et al. (denoted as Yetilmezsoy model and Al-Ismaili model, respectively). The ANN model showed the best prediction when seven statistical criteria were considered. The Pearson correlations for ANN, RA, Yetilmezsoy, and Al-Ismaili models were observed as 1.00, 0.98, 0.88, and 0.96, respectively, while mean absolute percentage errors (MAPE) were 17.84, 79.72, 63.24, and 80.50%, respectively. Hence it could be recommended to use ANN model for the prediction of freshwater production rate, however other three simple models could also be used with lower accuracy in the cases of unavailability of the ANN model.     

Prediction of Timber Kiln Drying Rates by Neural Networks

The purpose of this exploratory work was to apply artificial neural network (ANN) modeling to the prediction of timber kiln drying rates based on species and basic density information for the hem-fir mix that grows along the local coastal areas. The ANN models with three inputs (initial moisture content, basic density, and drying time) were developed to predict one output, namely, average final moisture content. The back-propagation algorithm, the most common neural network learning method, was implemented for testing, training, and validation. Optimal configuration of the network model was obtained by varying its main parameters, such as transfer function, learning rule, number of neurons and layers, and learning runs. Accurate prediction of the experimental drying rate data by the ANN model was achieved with a mean absolute relative error less than 2%, thus supporting the powerful predictive capacity of this modeling method.     

Prediction of Timber Kiln Drying Rates by Neural Networks

The purpose of this exploratory work was to apply artificial neural network (ANN) modeling to the prediction of timber kiln drying rates based on species and basic density information for the hem-fir mix that grows along the local coastal areas. The ANN models with three inputs (initial moisture content, basic density, and drying time) were developed to predict one output, namely, average final moisture content. The back-propagation algorithm, the most common neural network learning method, was implemented for testing, training, and validation. Optimal configuration of the network model was obtained by varying its main parameters, such as transfer function, learning rule, number of neurons and layers, and learning runs. Accurate prediction of the experimental drying rate data by the ANN model was achieved with a mean absolute relative error less than 2%, thus supporting the powerful predictive capacity of this modeling method.     

Dynamic crude oil fouling prediction in industrial preheaters using optimized ANN based moving window technique

The objective of this paper is to develop and validate a reliable, efficient and robust artificial neural network (ANN) model for online monitoring and prediction of crude oil fouling behavior for industrial shell and tube heat exchangers. To explore the complex dynamics of fouling, a new modeling strategy based on moving-window neural network approach is proposed. The essential character of this modeling approach is online updating of the ANN model whenever a new data block is available, so that it can effectively capture the slowly changing of process dynamics. The results of these models have been compared with appropriate sets of experimental data. The mean relative errors (MRE) of training and prediction subsets were about 6.61% and 8.06%, respectively. Since the data extraction in the refinery was performed every 2 h, the modeling approach led to an MRE of about 8% for fouling rate prediction of the next 50 h.     

Using Artificial Neural Network to Predict the Pressure Drop in a Rotating Packed Bed

Although rotating beds are good equipments for intensified separations and multiphase reactions, but the fundamentals of its hydrodynamics are still unknown. In the wide range of operating conditions, the pressure drop across an irrigated bed is significantly lower than dry bed. In this regard, an approach based on artificial intelligence, that is, artificial neural network (ANN) has been proposed for prediction of the pressure drop across the rotating packed beds (RPB). The experimental data sets used as input data (280 data points) were divided into training and testing subsets. The training data set has been used to develop the ANN model while the testing data set was used to validate the performance of the trained ANN model. The results of the predicted pressure drop values with the experimental values show a good agreement between the prediction and experimental results regarding to some statistical parameters, for example (AARD% = 4.70, MSE = 2.0 × 10^?5 and R² = 0.9994). The designed ANN model can estimate the pressure drop in the countercurrent flow rotating packed bed with unexpected phenomena for higher pressure drop in dry bed than in wet bed. Also, the designed ANN model has been able to predict the pressure drop in a wet bed with the good accuracy with experimental.     

Long-term prediction of nonlinear hydrodynamics in bubble columns by using artificial neural networks

An artificial neural network (ANN) model was proposed for the long-term prediction of nonlinear dynamics underlying holdup fluctuations in bubble columns with three different diameters of 200, 400 and 800 mm. Local holdup fluctuations were measured with an optical probe in the bubble columns. The superficial gas velocity was varied in the range of 33–90 mm/s. The time intervals between successive bubbles were extracted from the time series of holdup fluctuations to represent hydrodynamic behaviors in the system and used as training and validation data sets. The effect of data preprocessing as well as the numbers of nodes in input and hidden layers on the ANN training behavior was systematically investigated. The prediction capability of the ANN was evaluated in terms of time-averaged characteristics, power spectra and Lyapunov exponents. It was observed that: the ANN model, which was trained with experimental time series and gas velocity, can be used for the long-term prediction of dynamic characteristics in bubble columns by using random data as the initial input. The results indicate that the trained ANN models have the potential of modeling nonlinear hydrodynamic behaviors in bubble columns.