Predictive modeling of mercury speciation in combustion flue gases using GMDH-based abductive networks

Modeling mercury speciation is an important requirement for estimating harmful emissions from coal-fired power plants and developing strategies to reduce them. First-principle models based on chemical, kinetic, and thermodynamic aspects exist, but these are complex and difficult to develop. The use of modern data-based machine learning techniques has been recently introduced, including neural networks. Here we propose an alternative approach using abductive networks based on the group method of data handling (GMDH) algorithm, with the advantages of simplified and more automated model synthesis, automatic selection of significant inputs, and more transparent input–output model relationships. Models were developed for predicting three types of mercury speciation (elemental, oxidized, and particulate) using a small dataset containing six inputs parameters on the composition of the coal used and boiler operating conditions. Prediction performance compares favourably with neural network models developed using the same dataset, with correlation coefficients as high as 0.97 for training data. Network committees (ensembles) are proposed as a means of improving prediction accuracy, and suggestions are made for future work to further improve performance.     

NEURAL NETWORK-BASED PREDICTIVE CONTROL FOR MULTIVARIABLE PROCESSES

In this paper, the systematic derivations of setting up a nonlinear model predictive control based on the neural network are presented. This extends our previous work (Chen, 1998) into a multivariable system to explore the characteristics of the design. There are two stages for the development of nonlinear neural network predictive control: a neural network model and a control design. In the neural network model design, a parallel multiple-input, single-output neural network autoregressive with a model of exogenous inputs (NNARX) is proposed for multistep ahead predictions. In control design, the controller with extended control horizon is developed. The Levenberg-Marquardt algorithm is applied to training the NNARX model. Also, the sequential quadratic programming is used to search for the optimal manipulated inputs. The gradient of the objective function and constraints that require computation of Jacobian matrices are completely derived for optimization calculation. To demonstrate the control ability of MIMO cases, the proposed method is applied through two nonlinear simulation problems.     

Neural network-based predictive control for multivariable processes

In this paper, the systematic derivations of setting up a nonlinear model predictive control based on the neural network are presented. This extends our previous work (Chen, 1998) into a multivariable system to explore the characteristics of the design. There are two stages for the development of nonlinear neural network predictive control: a neural network model and a control design. In the neural network model design, a parallel multiple-input, single-output neural network autoregressive with a model of exogenous inputs (NNARX) is proposed for multistep ahead predictions. In control design, the controller with extended control horizon is developed. The Levenberg-Marquardt algorithm is applied to training the NNARX model. Also, the sequential quadratic programming is used to search for the optimal manipulated inputs. The gradient of the objective function and constraints that require computation of Jacobian matrices are completely derived for optimization calculation. To demonstrate the control ability of MIMO cases, the proposed method is applied through two nonlinear simulation problems.     

A neural network-based production process modeling and variable importance analysis approach in corn to sugar factory

Corn to sugar process has long faced the risks of high energy consumption and thin profits. However, it’s hard to upgrade or optimize the process based on mechanism unit operation models due to the high complexity of the related processes. Big data technology provides a promising solution as its ability to turn huge amounts of data into insights for operational decisions. In this paper, a neural network-based production process modeling and variable importance analysis approach is proposed for corn to sugar processes, which contains data preprocessing, dimensionality reduction, multilayer perceptron/convolutional neural network/recurrent neural network based modeling and extended weights connection method. In the established model, dextrose equivalent value is selected as the output, and 654 sites from the DCS system are selected as the inputs. LASSO analysis is first applied to reduce the data dimension to 155, then the inputs are dimensionalized to 50 by means of genetic algorithm optimization. Ultimately, variable importance analysis is carried out by the extended weight connection method, and 20 of the most important sites are selected for each neural network. The results indicate that the multilayer perceptron and recurrent neural network models have a relative error of less than 0.1%, which have a better prediction result than other models, and the 20 most important sites selected have better explicable performance. The major contributions derived from this work are of significant aid in process simulation model with high accuracy and process optimization based on the selected most important sites to maintain high quality and stable production for corn to sugar processes.     

Temperature control during microwave heating process by sliding mode neural network

In microwave heating applications, Lambert’s law is a common way to calculate power distribution. However, because of the complex application environment, Lambert’s law is not precise for the unknown power distribution on material surfaces. During the microwave heating process, the system process parameters can only be partly known by experience. Therefore, for such situations, to make the entire heating process safe, a sliding mode combined with a neural network algorithm is proposed. The algorithm is designed to calculate the suitable input power at each control period to make the material temperature follow the reference trajectory, which is determined by experience. The simulation and actual application results demonstrate that the proposed algorithm can commendably control the heating process. The difference between the reference trajectory and the material sampling temperature may exceed 1°C initially. However, as time progresses, the difference gradually decreases. Nonetheless, due to the low conduction coefficient, a single microwave heating process may take a long time. Therefore, many actual applications combine convective heat transfer with microwave. This article also discusses the control method of multiple inputs including microwave power and convective heat transfer with unknown model parameters. Another neural network is constructed to identify the unknown parameters. The algorithm is designed to obtain the suitable input power and input convective heat transfer at each control period. The simulation results show that the control algorithm can work well under multiple inputs. The material temperature on both the surfaces and the interior can follow the reference trajectory with a satisfactory difference, and suitable inputs can be obtained with few fluctuations during the learning process.     

The control of MSF desalination plants based on inverse model control by neural network

In this paper, a nonlinear inverse model control strategy based on neural network is proposed for MSF desalination plant. Artificial neural networks (ANNs) can handle complex and nonlinear process relationships, and are robust to noisy data. The designed neural networks consist of three layers identified from input–output data and trained with a descent gradient algorithm. The set point tracking performance of the proposed method was studied when the disturbance is present in the MSF system. Three controllers are designed for controlling the top brine temperature, the level of last stage and salinity. These results show that a neural network inverse model control strategy (NNINVMC) is robust and highly promising to be implemented in such nonlinear systems. Also the comparison between the top brine temperature of the proposed model and NN predicted data from the literature supports the accuracy of the model.     

基于模糊递归神经网络的污泥容积指数预测模型

污泥容积指数（SVI）,一个关键的污泥沉降性能评价指标。针对污水处理过程中污泥膨胀关键水质参数污泥容积指数难以准确在线测量,且实验室取样测量方法时间久、精度低,提出了一种改进型的模糊递归神经网络（HRFNN）用来预测污泥容积指数的变化,通过在网络第三层加入含有内部变量的反馈连接来实现输出信息的反馈。实验结果表明,与其他模糊神经网络相比,该网络的规模小、精度高,处理动态信息的能力明显加强。     

Oil fingerprint identification technology using a simplified set of biomarkers selected based on principal component difference

Gas chromatography–mass spectrometry (GC–MS), which can separate and quantify thousands of individual petroleum biomarker compounds, is generally acknowledged as the most powerful technique for oil fingerprinting nowadays. Traditional oil fingerprint studies employ the whole suite of biomarkers measured in chromatographic analysis, which is prone to introducing ambiguous variables in the whole set and being time and labour intensive. To extract the most representative and meaningful indicators for the oil fingerprinting and identification, this paper proposes a method based on principal component difference to select a simplified set of biomarkers, providing the possibility of faster elution and analysis procedures. For the purpose of further verifying the reliability and accuracy of our method, identification simulation experiments including principal component analysis (PCA) spatial clustering, hierarchical clustering, and generalized regression neural network are carried out with the whole set and the simplified set of biomarkers, respectively. All the results and analyses demonstrate that the simplified set of biomarkers selected by our proposed method can achieve almost the same or even better identification results than those of the whole set of biomarkers.     

人工神经网络用于氯代芳烃定量结构-活性关系研究

采用误差反传前向人工神经网络(Artificial neural network,ANN)建立了37种氯代芳烃的结构与其对孔雀鱼的急性毒性之间的定量关系模型(ANN模型)。以37种氯代芳烃的量子化学参数作为输入,急性毒性作为输出,采用内外双重验证的办法分析和检验所得模型的稳定性,所构建网络模型的相关系数为0.996 5、交叉检验相关系数为0.991 1、标准偏差为0.04、残差绝对值≤0.18,应用于外部预测集,外部预测集相关系数为0.988 4;而多元线性回归(Multiple linear regression,MLR)法模型的相关系数为0.949 6、交叉检验相关系数为0.928 8、标准偏差为0.14、残差绝对值≤0.32,外部预测集相关系数为0.950 5。结果表明,ANN模型获得了比MLR模型更好的拟合效果。     

Modelling of a woodchip refiner using artificial neural network

Neural networks can be an attractive alternative to mathematical modelling of complex and poorly understood processes if input/output data can easily be obtained. Woodchip refining falls into this category. The mechanism of the refining process is still being studied and no thorough models have yet been developed. A feed-forward neural network is proposed for modelling of woodchip refiners. The outputs predicted by the neural network are compared with industrial refiner data. It is also shown that a modified neural network structure can be used to optimize refiner operation and product quality. The advantages and disadvantages of neural network model application in simulation and optimization of industrial processes are discussed.     

ANALYSIS AND DETECTION OF OUTLIERS AND SYSTEMATIC ERRORS IN INDUSTRIAL PLANT DATA

This article describes the analysis of industrial process data to detect outliers and systematic errors. Data reconciliation is an important step in adjusting mathematical models to plant data. The quality of the data directly affects the quality of adjustment of the model for modeling, simulation, and optimization purposes. To detect these errors in a multivariable system is not an easy task. If the origin of the abnormal values is known, these values can be immediately discarded. On the other hand, if an error or an extreme observation is not clearly justified, the decision whether or not to discard these values must be based on statistical analysis. In this work, in addition to process knowledge, the methodology employed involves an approach based on statistical analysis, first-principle equations, neural network models, and a composite of these. The neural network based approach was used to represent the process in order to classify similar inputs and outputs, i.e., to identify clusters. The elimination of gross errors was performed by the similarity principle or by hypothesis testing for means. The system studied is the Isoprene Production Unit of BRASKEM, the largest Brazilian petrochemical plant. The analysis of the process was undertaken by using a one-year database. The frequency of the data collection of the monitoring variables was 15 minutes.     

ANALYSIS AND DETECTION OF OUTLIERS AND SYSTEMATIC ERRORS IN INDUSTRIAL PLANT DATA

This article describes the analysis of industrial process data to detect outliers and systematic errors. Data reconciliation is an important step in adjusting mathematical models to plant data. The quality of the data directly affects the quality of adjustment of the model for modeling, simulation, and optimization purposes. To detect these errors in a multivariable system is not an easy task. If the origin of the abnormal values is known, these values can be immediately discarded. On the other hand, if an error or an extreme observation is not clearly justified, the decision whether or not to discard these values must be based on statistical analysis. In this work, in addition to process knowledge, the methodology employed involves an approach based on statistical analysis, first-principle equations, neural network models, and a composite of these. The neural network based approach was used to represent the process in order to classify similar inputs and outputs, i.e., to identify clusters. The elimination of gross errors was performed by the similarity principle or by hypothesis testing for means. The system studied is the Isoprene Production Unit of BRASKEM, the largest Brazilian petrochemical plant. The analysis of the process was undertaken by using a one-year database. The frequency of the data collection of the monitoring variables was 15 minutes.     

人工神经网络研究取代芳烃的定量构效关系

采用误差反传前向人工神经网络（artificial neural network,ANN）建立了24种取代芳烃的结构与其对发光菌的急性毒性之间的定量关系模型（ANN模型）。以24种取代芳烃的量子化学参数作为输入,急性毒性作为输出,采用内外双重验证的办法分析和检验所得模型的稳定性和外部预测能力。所构建网络模型的相关系数为0.9834、交叉检验相关系数为0.9780、标准偏差为0.11、残差绝对值≤0.33,应用于外部预测集,外部预测集相关系数为0.9955;而多元线性回归（multiple linear regression,MLR）法模型的相关系数为0.9786、标准偏差为0.12、残差绝对值≤0.36,外部预测集相关系数为0.9904。结果表明,ANN模型获得了比MLR模型更好的拟合效果。     

Modelling of A Fluidized Bed Drier Using Artificial Neural Network

ABSTRACT

Proper modelling of a fluidized bed drier (FBD) is important to design model based control strategies. A FBD is a non-linear multivariable system with non-minimum phase characteristics. Due to the complexities in FBD conventional modelling techniques are cumbersome. Artificial neural network (ANN) with its inherent ability to “learn” and “absorb” non-linearities, presents itself as a convenient tool for modelling such systems.

In this work, an ANN model for continuous drying FBD is presented. A three layer fully connected feedfordward network with three inputs and two outputs is used. Backpropagation learning algorithm is employed to train the network. The training data is obtained from computer simulation of a FBD model from published literature. The trained network is evaluated using randomly generated data as input and observed to predict the behaviour of FBD adequately.     

人工神经网络用于氟化酚的定量构效关系研究

采用误差反传前向人工神经网络（ANN）建立了16种氟化酚的结构与其对梨形四膜虫的毒性之间的定量结构-活性关系（QSAR）模型。以16种氟化酚的量子化学和理化参数作为输入,对梨形四膜虫的急性毒性作为输出,采用内外双重验证的办法分析和检验所得模型的稳定性和外部预测能力,所构建网络模型的相关系数为0.999 8、交叉检验相关系数为0.981 8、标准偏差为0.01、残差绝对值≤0.04,应用于外部预测集,外部预测集相关系数为0.993 6;而多元线性回归（MLR）法模型的相关系数为0.980 2、标准偏差为0.119、残差绝对值≤0.28,外部预测集相关系数为0.980 3。结果表明,ANN模型获得了比MLR模型更好的拟合效果。     

Studies on the Applicability of Artificial Neural Network (ANN) in Continuous Stirred Ultrafiltration

An artificial neural network model of a continuous stirred ultrafiltration process, is proposed in the present study, which is able to predict permeate volumetric flux and permeate concentration at different bulk concentration, stirrer speed, pressure and time. Because of the complexity in generalization of the phenomenon of ultrafiltration by any mathematical model, the neural network proves to be a very promising method for the purpose of process simulation. The network uses the Back‐propagation Algorithm for evaluating the connection strengths, representing the correlations between inputs (bulk concentration, stirrer speed, pressure and time) and output (permeate concentration and flux). The network employed in the present study uses four input nodes corresponding to the operating variables, and two output nodes corresponding to the measurement of the performance of the network (flux and permeate concentration). Experiments were performed to constitute the learning databases for the continuous stirred ultrafiltration process using PEG‐6000 solute, and cellulose acetate membrane of 5000 MWCO. The network employed in the present study uses two hidden layers, with the optimum number of nodes being thirty and twenty. A leaning rate of 0.3, and momentum factor of 0.4 was used. The results predicted by the model were in good agreement with the experimental data, and the average deviations for all the cases are found to be well within ±10 %.     

基于粗糙集与人工神经网络的烟气机故障诊断

提出了一种粗糙集理论与神经网络集成的烟气机故障诊断方法。首先应用SOM网络对故障诊断数据中的连续属性进行离散化,然后根据粗糙集理论,借助遗传算法进行故障诊断决策系统约简,获得最优决策系统。最后在最优决策系统的基础上,设计RBF神经网络对烟气机故障进行诊断。试验结果显示,该方法可以有效提高烟气机故障诊断的精度和效率。     

Molecular reconstruction of complex hydrocarbon mixtures: An application of principal component analysis

Three methods for reconstruction of the detailed molecular composition of complex hydrocarbon mixtures, based on their global properties, are compared: a method based on the Shannon entropy criterion, an artificial neural network and a multiple linear regression model. In spite of the broad range of naphthas included in the training set, the application range of the last two methods proved to be limited. Principal component analysis allowed to identify their three‐dimensional ellipsoidal application range. In this subspace, the artificial neural network is more accurate than the multiple linear regression model and the Shannon entropy method. However, outside its application range, the performance of the neural network, as well as the regression model, decreases drastically. In contrast, the performance of the Shannon entropy method is not influenced by the characteristics of the considered naphtha, but rather depends on the number of available commercial indices. The Shannon entropy method yields comparable results to the artificial neural network, provided that a sufficient amount of distillation data is available to supply information on the carbon number distribution. Combining the reconstruction methods with a fundamental simulation model illustrates the necessity of having accurate feedstock reconstruction methods since they allow to capture the full power of fundamental simulation models for the simulation of industrial processes. © 2010 American Institute of Chemical Engineers AIChE J, 2010     

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.