Artificial neural network-based modeling of pressure drop coefficient for cyclone separators

In order to build the complex relationships between cyclone pressure drop coefficient (PDC) and geometrical dimensions, representative artificial neural networks (ANNs), including back propagation neural network (BPNN), radial basic functions neural network (RBFNN) and generalized regression neural network (GRNN), are developed and employed to model PDC for cyclone separators. The optimal parameters for ANNs are configured by a dynamically optimized search technique with cross-validation. According to predicted accuracy of PDC, performance of configured ANN models is compared and evaluated. It is found that, all ANN models can successfully produce the approximate results for training sample. Further, the RBFNN provides the higher generalization performance than the BPNN and GRNN as well as the conventional PDC models, with the mean squared error of 5.84 × 10^?4 and CPU time of 120.15 s. The result also demonstrates that ANN can offer an alternative technique to model cyclone pressure drop.     

AI models for correlation of physical properties in system of 1DMA2P-CO2-H2O

In this work, the density, viscosity, and specific heat capacity of pure 1-dimethylamino-2-propanol (1DMA2P) as well as aqueous unloaded and CO₂-loaded 1DMA2P solution (with a CO₂ loading of 0.04–0.70 mol CO₂/mol amine) were measured over the 1DMA2P concentration range of 0.5–3.0 mol/L and temperature range of 293–323 K. The observed experimental results of these thermophysical properties of the 1DMA2P-H₂O-CO₂ system were correlated using empirical models as well as artificial neural network (ANN) models (namely, back-propagation neural network [BPNN] and radial basis function neural network [RBFNN] models). It was found that the developed BPNN and RBFNN models could predict the experimental results of 1DMA2P-H₂O-CO₂ better than correlations using empirical models. The results could be treated as one of the accurate and potential methods to predict the physical properties for aqueous amine CO₂ absorption systems.     

Finite element-based machine learning approach for optimization of process parameters to produce silicon carbide ceramic complex parts

Design and fabrication of silicon carbide ceramic complex parts introduce considerable difficulties during injection molding. Due to the great importance in processing optimization, an accurate prediction on the stress and displacement is required to obtain the desired final product. In this paper, a conceptual framework on combination of finite element method (FEM) and machine learning (ML) method was developed to optimize the injection molding process, which can be used to manufacture large-aperture silicon carbide mirror. The distribution characteristics of temperature field and stress field were extracted from FEM simulation to understand the injection molding process and construct database for ML modeling. To select the most appropriate model, the predictive performance of three ML models were estimated, including generalized regression neural network (GRNN), back propagation neural network (BPNN) and extreme learning machine (ELM). The results show that the developed ELM model exhibits exceptional predictive performance and can be utilized to predict the stress and displacement of the green body. This work allows us to obtain reasonable technique parameters with particular attention to the loading speed and provides some fundamental guidance for the fabrication of lightweight SiC ceramic optical mirror.     

Prediction of BP neural network and preliminary application for suppression of low-temperature oxidation of coal stockpiles by pulverized coal covering

We have developed a new method to suppress spontaneous combustion of coal piles by covering the surface of coal piles with pulverized coal. Experimental studies of three type of coal samples from China (YJL, CYW, and SW) with particle size ratio of 10:1 were performed to investigate the low-temperature oxidation of coal pillars. In this work, we have also demonstrated that the distributions of oxygen concentration, the temperature field, as well as the spontaneous combustion of three typical Chinese coal samples can be predicted accurately using back-propagation neural network (BPNN) by MATLAB. Pearson correlation analysis showed that temperature and oxygen concentration highly depend on the ratio of pulverized coal thickness to coal piles thickness, activation energy, void ratio, wind speed, and low-temperature oxidation time. Three-layer BPNN models with five input factors were developed to predict the low-temperature oxidation process under pulverized coal. The prediction data of BPNN are fitting better with our experimental data, which confirms that BPNN modelling can accurately predict the low temperature oxidation process of coal.     

Generic AI models for mass transfer coefficient prediction in amine-based CO2 absorber,Part II: RBFNN and RF model

In this work, the radial basis function neural network (RBFNN) and random forest (RF) algorithms were employed to develop generic AI models predicting mass transfer coefficient in amine-based CO₂ absorber. The models with operating parameters as input gave quite different prediction performance in different CO₂ absorption systems. To secure better applicability, extra parameters related to amine type and packing characteristics were introduced to reasonably describe mass transfer behaviors, respectively. Moreover, the generic models were proposed by considering all influencing factors of mass transfer in CO₂ absorber column. Furthermore, the performance of BPNN, RBFNN, and RF models was completely compared and fully discussed in terms of AARE. All three generic models could predict mass transfer coefficient of CO₂ absorber very well. It was found that the BPNN models provide the best predication with AAREs of below 5%. The developed generic model could serve as a fast and efficient tool for preliminary selection and evaluation of potential amines for CO₂ absorption. The framework of generic ML model development was also clearly presented, which could provide theoretical basis and practical guidance for the implementation and application of ML models in the carbon capture field.     

基于BP神经网络的A/O脱氮系统外加碳源的仿真研究

对连续流缺氧/好氧（A/O）脱氮工艺处理低碳氮比（C/N）生活污水的外加碳源系统进行了仿真研究.由于处理系统的外加碳源量、总回流比和出水总氮（TN）之间存在的复杂非线性关系,很难用常规的参数型模型进行描述,给处理系统控制策略的实现带来较大的困难.针对该问题,引入了BP神经网络,通过神经网络对试验数据的学习建立系统的非参数型模型,通过该模型对系统进行仿真研究,可以达到优化碳源投加量的目的.研究结果表明,经过训练的BP神经网络模型可以很好地模拟处理系统,根据仿真分析结果可以实现碳源投加量的优化控制,这为污水处理系统在线最优控制的实现提供了一条可行的途径.     

基于虚拟技术的齿轮故障智能诊断研究

利用LabVIEW图形化编程语言开发了信号分析与处理、信号特征提取和故障诊断三大模块。信号特征提取由小波包分解来实现,故障诊断通过神经网络完成,小波包分解提取的齿轮振动信号各频段能量特征值作为神经网络的输入向量。以模拟故障实验台获取的齿轮典型故障振动信号训练神经网络,利用训练好的神经网络对齿轮进行故障诊断,实验结果表明：所开发的齿轮故障智能诊断系统能有效识别齿轮故障,较好地将虚拟技术应用于故障诊断领域。     

Color prediction of mushroom slices during drying using Bayesian extreme learning machine

Abstract

Color is an important appearance attribute of fruits and vegetables during drying processing, as it influences consumer’s preference and acceptability. Establishing color change kinetics model is an effective way for better understanding the quality changes and optimization of drying process. However, it is difficult to quickly and accurately predict color change kinetics during drying as it is highly nonlinear, complex, dynamic, and multivariable. To alleviate this problem, a new model based on extreme learning machine integrated Bayesian methods (BELM) has been developed for the prediction of color changes of mushroom slices during drying process. The effects of drying temperature (55, 60, 65, 70, and 75?°C) and air velocity (3, 6, 9, and 12?m/s) on color change kinetics of mushroom slices during hot air impingement drying were firstly explored and the experimental results indicated that both drying temperature and air velocity significantly affected the color attributes. Then, to validate the robustness and effectiveness of BELM, the basic extreme learning machine (ELM) and traditional back-propagation neural network (BPNN) models have also been employed to predict the color quality. In terms of prediction accuracy and execution time, BELM could achieve least similar or even better performance than ELM and BPNN. It overcame the overfitting problems of ELM. The test results of optimal BELM model by two new cases revealed that the lowest R² and highest RMSE of BELM model were 0.9725 and 0.0563, respectively. The absolute values of relative errors between the actual and predicted values were lower than 8.5%.     

Application of artificial neural networks to the analysis of two‐dimensional fluorescence spectra in recombinant E coli fermentation processes

A two‐dimensional (2D) spectrofluorometer was used to monitor various fermentation processes with recombinant E coli for the production of 5‐aminolevulinic acid (ALA). The whole fluorescence spectral data obtained during a process were analyzed using artificial neural networks, ie self‐organizing map (SOM) and feedforward backpropagation neural network (BPNN). The SOM‐based classification of the whole spectral data has made it possible to qualitatively associate some process parameters with the normalized weights and variances, and to select some useful combinations of excitation and emission wavelengths. Based on the classified fluorescence spectra a supervised BPNN algorithm was used to predict some of the process parameters. It was also shown that the BPNN models could elucidate some sections of the process's performance, eg forecasting the process's performance. Copyright © 2005 Society of Chemical Industry     

A Computational Framework Based on Ensemble Deep Neural Networks for Essential Genes Identification

Essential genes contain key information of genomes that could be the key to a comprehensive understanding of life and evolution. Because of their importance, studies of essential genes have been considered a crucial problem in computational biology. Computational methods for identifying essential genes have become increasingly popular to reduce the cost and time-consumption of traditional experiments. A few models have addressed this problem, but performance is still not satisfactory because of high dimensional features and the use of traditional machine learning algorithms. Thus, there is a need to create a novel model to improve the predictive performance of this problem from DNA sequence features. This study took advantage of a natural language processing (NLP) model in learning biological sequences by treating them as natural language words. To learn the NLP features, a supervised learning model was consequentially employed by an ensemble deep neural network. Our proposed method could identify essential genes with sensitivity, specificity, accuracy, Matthews correlation coefficient (MCC), and area under the receiver operating characteristic curve (AUC) values of 60.2%, 84.6%, 76.3%, 0.449, and 0.814, respectively. The overall performance outperformed the single models without ensemble, as well as the state-of-the-art predictors on the same benchmark dataset. This indicated the effectiveness of the proposed method in determining essential genes, in particular, and other sequencing problems, in general.     

PREDICTION OF QUALITY CHANGES DURING OSMO-CONVECTIVE DRYING OF BLUEBERRIES USING NEURAL NETWORK MODELS FOR PROCESS OPTIMIZATION

Artificial neural network (ANN) models were used for predicting quality changes during osmo-convective drying of blueberries for process optimization. Osmotic drying usually involves treatment of fruits in an osmotic solution of predetermined concentration, temperature and time, and generally affects several associated quality factors such as color, texture, rehydration ratio as well as the finish drying time in a subsequent drier (usually air drying). Multi-layer neural network models with 3 inputs (concentration, osmotic temperature and contact time) were developed to predict 5 outputs: air drying time, color, texture, and rehydration ratio as well as a defined comprehensive index. The optimal configuration of neural network model was obtained by varying the main parameters of ANN: transfer function, learning rule, number of neurons and layers, and learning runs. The predictability of ANN models was compared with that of multiple regression models, confirming that ANN models had much better performance than conventional mathematical models. The prediction matrices and corresponding response curves for main processing properties under various osmotic dehydration conditions were used for searching the optimal processing conditions. The results indicated that it is feasible to use ANN for prediction and optimization of osmo-convective drying for blueberries.     

PREDICTION OF QUALITY CHANGES DURING OSMO-CONVECTIVE DRYING OF BLUEBERRIES USING NEURAL NETWORK MODELS FOR PROCESS OPTIMIZATION

Artificial neural network (ANN) models were used for predicting quality changes during osmo-convective drying of blueberries for process optimization. Osmotic drying usually involves treatment of fruits in an osmotic solution of predetermined concentration, temperature and time, and generally affects several associated quality factors such as color, texture, rehydration ratio as well as the finish drying time in a subsequent drier (usually air drying). Multi-layer neural network models with 3 inputs (concentration, osmotic temperature and contact time) were developed to predict 5 outputs: air drying time, color, texture, and rehydration ratio as well as a defined comprehensive index. The optimal configuration of neural network model was obtained by varying the main parameters of ANN: transfer function, learning rule, number of neurons and layers, and learning runs. The predictability of ANN models was compared with that of multiple regression models, confirming that ANN models had much better performance than conventional mathematical models. The prediction matrices and corresponding response curves for main processing properties under various osmotic dehydration conditions were used for searching the optimal processing conditions. The results indicated that it is feasible to use ANN for prediction and optimization of osmo-convective drying for blueberries.     

Maximum solid concentrations of coal water slurries predicted by neural network models

The nonlinear back-propagation (BP) neural network models were developed to predict the maximum solid concentration of coal water slurry (CWS) which is a substitute for oil fuel, based on physicochemical properties of 37 typical Chinese coals. The Levenberg-Marquardt algorithm was used to train five BP neural network models with different input factors. The data pretreatment method, learning rate and hidden neuron number were optimized by training models. It is found that the Hardgrove grindability index (HGI), moisture and coalification degree of parent coal are 3 indispensable factors for the prediction of CWS maximum solid concentration. Each BP neural network model gives a more accurate prediction result than the traditional polynomial regression equation. The BP neural network model with 3 input factors of HGI, moisture and oxygen/carbon ratio gives the smallest mean absolute error of 0.40%, which is much lower than that of 1.15% given by the traditional polynomial regression equation.     

A uniform design‐based back propagation neural network model for amino acid composition and optimal pH in G/11 xylanase

A link between amino acid composition and optimal pH in G/11 xylanase was established. A back propagation neural network (BPNN) was used as the mathematical tool and a uniform design method was employed to optimise the architecture of the BPNN. Results showed that the calculated and predicted pHs fitted the optimal pHs of xylanase very well, with mean absolute percentage errors (MAPEs) of 3.02 and 4.06%, mean square errors (MSEs) of 0.19 and 0.19 pH unit and mean absolute errors (MAEs) of 0.11 and 0.19 pH unit respectively. The new model performed better in fitting and prediction compared with a previously reported model based on stepwise regression. Copyright © 2006 Society of Chemical Industry     

应用多层网络进行分子拓扑结构与物性的关联

应用具有极强非线性映射能力，且能以任意精度逼近函数的多层网络进行分子拓扑指数与性质关联。集中研究了烷烃的三个物性（沸点、临界温度和临界压力）与结构的关系，其中烷烃结构利用ａｄｈｏｃ指数描述；多层网络的学习算法则采用共轭梯度算法。所得结果良好，从而以非常简单的方式有效地实现了指数与性质的关联。     

Atomistic molecular modelling of crosslinked epoxy resin

In the present study, a new method was developed to construct atomistic molecular models of crosslinked polymers based on commercially important epoxy resin. This method employed molecular dynamics/molecular mechanics schemes and assumed close proximity. The generic Dreiding2.21 force-field and advanced compass force-field were used for the construction of models and prediction of properties, respectively. A polymer network with conversion up to 93.7% was successfully generated by this method. Density and elastic constants of the system were calculated from the equilibrated structure for the validation of the generated models. The simulated results compared reasonably with experimental data available. The developed method would hold great promise in further molecular simulations for structure and properties of epoxy resin or other cured systems.     

基于IGA优化的神经网络的小电流接地选线方法

为了克服基于BP网络的故障选线方法学习效率低、收敛速度慢和容易陷入局部极小的缺点,提出了基于免疫遗传算法( IGA)的神经网络来实现配电网故障选线.通过快速傅里叶变换和小波包变换从零序电流信号中提取相应的故障特征量作为神经网络的输入,利用免疫遗传算法对神经网络进行训练,完成训练的神经网络模型为故障选线模型.ATP仿真结...     

气相法二氧化硅/高温硅橡胶补强体系动态流变性能

采用橡胶加工分析仪,在一定的温度和频率下,分析了不同性质结构的气相法纳米二氧化硅在高温硅橡胶中的聚集体网络结构差异,以及聚集体网络对填充胶动态流变性能的影响. 研究发现,高分支结构聚集体填充胶具有较高的剪切模量及损耗模量,对应变的依赖性大,补强作用好.