一种改进BP网络结构用于化学中的非线性校正

针对化学领域中的非线性关系特点,在常规BP网络基础上,提出了一种“杂交”型BP网络,包含两个隐层,并有输入层到输出层的直连接。它可很好地解释数据中同时存在的线性及非线性关系,效果优于多元回归法及普通BP算法。     

Modeling composite emulsion polymerization kinetics

The conversion profiles of a number of factorial designed experiments used to study composite emulsion polymerization were modeled using a deterministic mathematical construct as well as an empirical neural network approach. In the deterministic modeling approach, existing mechanistic models for emulsion polymerization were employed for which estimates of rate constants were obtained from established literature sources as well as experiments. Fitting of the kinetic data was done using nonlinear fitting algorithms to adjust the estimated rate constants to provide the best fit of the conversion profiles. In the case of the empirical modeling using neural networks, the neural net inputs were in the form of the factor levels of the various experimental designs. Several nonrelated experimental designs could be combined in this way to serve as the input, whereas the conversion profiles were targeted as outputs. Following the successful implementation of both modeling strategies, a hybrid modeling approach was tested by combining the neural network predictive power to estimate values for rate constants while retaining the aforementioned mechanistic models to fit the data. © 2012 Wiley Periodicals, Inc. Int J Chem Kinet 45: 101–117, 2013     

On the inverse miniemulsion copolymerization and terpolymerization of acrylamide, <Emphasis Type="Italic">N,N′</Emphasis>-methylenebis(acrylamide) and methacrylic acid

The kinetics of free-radical copolymerization and terpolymerization of acrylamide (AAm), N, N′-methylenebis(acrylamide) (MBA) and methacrylic acid (MA) in the inverse water/monomer/cyclohexane/Tween 85 miniemulsion was investigated. Polymerizable sterically-stable miniemulsions were formulated in cyclohexane as a continuous medium. Polymerizations are very fast and reach the final conversion within several minutes. The dependence of the polymerization rate vs. conversion is described by a curve with two nonstationary rate intervals. The maximum rate of polymerization slightly increases with increasing concentration of crosslinking monomer (MBA) and strongly decreases by the addition of MA. The rate of polymerization is inversely proportional to the 0.9 ^th and 1.8 ^th power of the particle concentration without and with MA, respectively. The number of polymer particles is inversely proportional to the 0.18 ^th and 0.13 ^th power of MBA concentration. The kinetic and colloidal parameters of the miniemulsion polymerization are discussed in terms of microemulsion polymerization model.     

人工神经网络在溶解度预测方面的应用

溶解度作为一项重要的物化指标,一直是化学学科的研究重点。然而,通过实验测量获得数据耗时费力,因此,科研人员建立了多种理论方法来进行估算,其中,人工神经网络因其能够关联复杂的多变量情况而受到广泛关注。本文综述了人工神经网络在物质溶解度预测方面的应用,介绍了应用最广泛的3种神经网络(BP神经网络、小波神经网络、径向基神经网络)的模型结构、预测方法和预测优势,探讨了神经网络的不足以及改进方法。文章最后对神经网络在物质溶解度预测方面的发展前景进行了展望。与其他方法相比,人工神经网络技术在物质溶解度预测方面具有预测结果精确度高、操作简单等特点,具有广阔的应用前景,但输入变量选择、隐含层节点数确定、避免局部最优等问题还需逐步建立系统的理论指导。     

Optimization methodology applied to feed‐forward artificial neural network parameters

This article recommends a methodology for developing a neural network with great chances to be an optimal one. The method is based on trial and error in determining the internal parameters of the network considered as having a significant influence over its performance: the number of hidden layers, activation function, number of neurons in the hidden layers, training epochs, learning rate, and momentum term. This optimization methodology is presented in two separate sections: first of them contains a series of practical considerations recommended for neural network modeling, and the second is represented by the proposed optimization algorithm, formulated in six steps and based on the practical statements. Two case studies are chosen to exemplify the use of the algorithm for finding the near optimal neural network: the dependence of the reduced and intrinsic viscosities of the siloxane‐organic copolymers of the solution concentration, temperature, and copolymer type, differing by the siloxane sequence length. The two siloxane‐organic polyazomethines resulted by the reaction of a fully aromatic bisazomethine diol with α,ω‐bis(chloromethyl)oligodimethylsiloxanes. © 2009 Wiley Periodicals, Inc. Int J Quantum Chem, 2011     

人工神经网络预测煤炭成浆浓度的研究

考虑煤炭的多种理化特性建立了成浆浓度的神经网络预测模型,对其数据预处理方法、学习率和中间层节点数等进行了深入讨论。水分、挥发分、分析基碳、灰分和氧等五个因子对于煤炭成浆性的预测起到主导作用。五因子、七因子和八因子神经网络模型对煤炭成浆浓度的预测误差分别为：0.53%、0.50%和0.74%,而现有回归分析方程的误差为1.15%,故神经网络模型比回归分析方程有更好的预测能力,尤以七因子模型最佳。     

Neural network models for infrared spectrum interpretation

A neural network model having a layer of hidden units is described which can identify functional groups in organic compounds, based on their infrared spectra. This network shows substantially better performance than the simple linear model reported earlier. The effect of the training set size and composition, the number of hidden units used, and the training time were studied.     

神经网络方法在血管紧张素转换酶抑制剂定量构效关系建模中的应用

对20个ACEI化合物用量子化学方法进行结构优化并计算出10个参数,用9种不同隐含层节点数的BP神经网络研究了ACEI的定量构效关系,建立了节点为10/6/1的三层BP神经网络模型。结果表明:以量化理论计算所得参数可以构建合理的ACEI定量构效关系模型,神经网络模型M6的r2=0.995,S=0.050,6个验证集化合物的残差平方和为0.002,预测能力明显强于多元线形回归模型,亦优于同类文献报道,可作为ACEI研发领域中预测先导化合物活性的理论工具。     

Microdroplet size prediction in microfluidic systems via artificial neural network modeling for water-in-oil emulsion formulation

In this paper, an experimental study and modeling by artificial neural networks were carried out to predict the generated microdroplet dimensionless size in a microfluidic system in order to formulate a water-in-oil emulsion. The various parameters that affect the size of microdroplets (flow rates, viscosities, surface tensions of both the two phases and the diameter of the microchannel) are studied and further grouped into dimensionless numbers; we used these numbers as input to the neural network and the dimensionless length as output. The better neural network architecture has 10 neurons in the hidden layer with a mean square error of 1.4 10^?6 and a determination’s coefficient near 1 value. The relative importance of inputs on the size of the microdroplets has been determined using the Garson algorithm and the results are in good agreement with other works.     

Development of an inorganic cations retention model in ion chromatography by means of artificial neural networks with different two-phase training algorithms

This paper describes development of artificial neural network (ANN) retention model, which can be used for method development in variety of ion chromatographic applications. By using developed retention model it is possible both to improve performance characteristic of developed method and to speed up new method development by reducing unnecessary experimentation. Multilayered feed forward neural network has been used to model retention behaviour of void peak, lithium, sodium, ammonium, potassium, magnesium, calcium, strontium and barium in relation with the eluent flow rate and concentration of methasulphonic acid (MSA) in eluent. The probability of finding the global minimum and fast convergence at the same time were enhanced by applying a two-phase training procedure. The developed two-phase training procedure consists of both first and second order training. Several training algorithms were applied and compared, namely: back propagation (BP), delta-bar-delta, quick propagation, conjugate gradient, quasi Newton and Levenberg-Marquardt. It is shown that the optimized two-phase training procedure enables fast convergence and avoids problems arisen from the fact that every new weight initialization can be regarded as a new starting position and yield irreproducible neural network if only second order training is applied. Activation function, number of hidden layer neurons and number of experimental data points used for training set were optimized in order to insure good predictive ability with respect to speeding up retention modelling procedure by reducing unnecessary experimental work. The predictive ability of optimized neural networks retention model was tested by using several statistical tests. This study shows that developed artificial neural network are very accurate and fast retention modelling tool applied to model varied inherent non-linear relationship of retention behaviour with respect to mobile phase parameters.     

QSAR modelling using combined simple competitive learning networks and RBF neural networks

The aim of this study was to propose a QSAR modelling approach based on the combination of simple competitive learning (SCL) networks with radial basis function (RBF) neural networks for predicting the biological activity of chemical compounds. The proposed QSAR method consisted of two phases. In the first phase, an SCL network was applied to determine the centres of an RBF neural network. In the second phase, the RBF neural network was used to predict the biological activity of various phenols and Rho kinase (ROCK) inhibitors. The predictive ability of the proposed QSAR models was evaluated and compared with other QSAR models using external validation. The results of this study showed that the proposed QSAR modelling approach leads to better performances than other models in predicting the biological activity of chemical compounds. This indicated the efficiency of simple competitive learning networks in determining the centres of RBF neural networks.     

On the Robustness of Forward and Inverse Artificial Neural Networks for the Simulation of Ethylene/1‐Butene Copolymerization

Forward and inverse artificial neural network (ANN) models are used to describe ethylene/1‐butene copolymerization with a model catalyst having two site types. The forward ANN predicts number and weight average molecular weights, average comonomer content, and polymer yield as a function of a set of polymerization conditions, while the inverse model estimates polymerization conditions needed to produce copolymers with desired microstructures. The forward model is found to be robust and resilient to random noise introduced into the datasets. The inverse model, however, leads to multiple solutions (several polymerization conditions can produce polymers with similar microstructures) and is sensitive to random noise in the data. Although the polymerization conditions estimated from inverse ANN are different from the model data, the estimated polymerization conditions are found to provide similar microstructures even with the random noise.     

Usage of Artificial Neural Network (Back Propagation) in Optimising Salicylic Acid Determination with Ferric(III) Nitrate

Abstract

The use of artificial neural networks (ANN) in optimizing salicylic acid (SA) determination is presented in this paper. A simple and rapid spectrophotometric method for salicylic acid (SA) determination was carried out based on the complexation of salicylic acid–ferric(III) nitrate, SAFe(III). The SA forms a stable purple complex with ferric(III) nitrate at pH 2.45. The useful dynamic linear range is 0.01–0.35 g/L. It has a maximum absorption at 524 nm and the stability is more than 50 hours. The results were used for artificial neural networks (ANNs) training to optimize data. For training and validation purposes, a back‐propagation (BP) artificial neural network (ANN) was used. The results showed that ANN technique was very effective and useful in broadening the limited dynamic linear response range mentioned to an extensive calibration response (0.01–0.70 g/L). It was found that a network with 22 hidden neurons was highly accurate in predicting the determination of SA. This network scores a summation of squared error (SSE) skill and low average predicted error of 0.0078 and 0.00427 g/L, respectively.     

Rapid in situ determination of total oil concentration in water using ultraviolet fluorescence and light scattering coupled with artificial neural networks

Real-time measurement of total oil concentration in complex samples is required in wastewater discharge streams from ships and processing industries. A novel technology has been developed for the accurate quantification of a variety of single oils and their mixtures. Four major types of oils (lube oils 2190 and 9250, diesel fuel marine (DFM), and jet fuel (JP5)), each of which consisted of a dozen subtypes of oil samples, were examined to obtain both fluorescence and light scattering spectra as a function of concentration of single oils and mixtures. Tremendous variations in both fluorescence and scattering were observed among oil types, subtypes, and mixtures. The spectral response of an oil mixture was not the simple summation of respective single oils. To account for all these variations, a multivariate, nonlinear calibration method is applied to associate instrumental responses with oil concentrations using artificial neural networks (ANNs). The neural network architecture has been established by optimizing network parameters such as epochs, the number of neurons in the hidden layer, and learning rates in order to achieve the maximum accuracy of oil concentration measurements. It is demonstrated that the simultaneous, combined use of fluorescence and light scattering significantly improves the accuracy of measurement for oil samples. The newly developed technique permits the reliable, real-time determination of the total concentration of various oils and mixtures in water.     

Phenomenological and neural-network modeling of cephalosporin C production bioprocess

Cephalosporin C production process withCephalosporium acremonium ATCC 48272 in synthetic medium was investigated and the experimental results allowed the development of a mathematical model describing the process behavior. The model was able to explain fairly well the diauxic phenomenon, higher growth rate during the glucose-consumption phase, and the production occurring only in the sucrose-consumption phase. Moreover, the process was simulated utilizing the neural-networks technique. Two feed-forward neural-networks with one hidden layer were employed. Both models, phenomenological and neural-networks based, satisfactorily describe the bioprocess. The difficulties in determining kinetic parameters are avoided when neural networks are utilized.     

Modeling and simulation of free radical polymerization of styrene under semibatch reactor conditions

The first part of this approach is concerned with the elaboration of a radical polymerization model of styrenne, based on a kinetic diagram that includes chemical and thermal initiation, propagation, termination by recombination and chain transfer to the monomer. Furthermore, volume contraction during polymerization is considered, as well as the gel and glass effects. The mathematical formalism that describes the model in terms of moments is explored in detail. The model was then used to predict the changes in monomer conversion and molecular weight after intermediate addition of initiator and monomer. The results of this operation are dependent on the conditions of the reaction mass, quantity, and moment of substance addition. Therefore, the simulations were performed at different times with respect to the gel effect; before, during and after this phenomenon, and also with respect to different temperatures and initiators. Increasing the initiator concentration before the gel effect leads to an earlier appearance of the phenomenon and to a decrease in molecular weight. The ratio reveals a polydispersity index smaller for the intermediate addition of initiator. No significant changes take place during or after the gel effect. If along with the initiator, unreacted monomver (used to dissolve the initiator) enters the reactor, a small dip in conversion is observed. The general conclusion of this paper reveals the intermediate addition of initiator as a method to control polymer properties and to prevent the “dead-end” polymerization of styrene.     

Poly(<Emphasis Type="SmallCaps">l</Emphasis>-lactide) networks with tailored water sorption

A poly(l-lactide) diol was obtained through ring opening polymerization of l-lactide, using 1,6 hexanediol and tin(II) 2 ethylhexanoate as a catalyst. In the second step, the poly(l-lactide) macromer (mLA) was obtained by the reaction of poly(l-lactide) diol with methacrylic anhydride. The effective incorporation of the polymerizable end groups was assessed by Fourier transform infrared spectroscopy and nuclear magnetic resonance (¹H NMR). Besides, poly(l-lactide) networks (pmLA) were prepared by photopolymerization of mLA. Further, the macromer was copolymerized with 2-hydroxyethyl acrylate seeking to tailor the hydrophilicity of the system. A set of hydrophilic copolymer networks were obtained. The phase microstructure of the new system and the network architecture was investigated by differential scanning calorimetry, infrared spectroscopy, dynamic mechanical spectroscopy, thermogravimetry, and water sorption studies.     

A NEURAL NETWORK STUDY ON GLASS TRANSITION TEMPERATURE OF POLYMERS

In this paper, an atificial neural network model is adopted to study the glass transition temperature of polymers. Inour artificial neural networks, the input nodes are the characteristic ratio C_∞, the average molecular weigh M_e betweenentanglement points and the molecular weigh M_(mon) of repeating unit. The output node is the glass transition temperature T_g,and the number of the hidden layer is 6. We found that the artificial neural network simulations are accurate in predicting theoutcome for polymers for which it is not trained. The maximum relative error for predicting of the glass transitiontemperature is 3.47%, and the overall average error is only 2.27%. Artificial neural networks may provide some new ideas toinvestigate other properties of the polymers.