基于旋转曲面变换PSO算法的神经网络用于胺类有机物毒性分类

神经网络模型能有效地模拟非线性的输入输出关系。本研究应用三层前馈网络对51种胺类有机物进行了结构-毒性关系的分类研究。常规的神经网络权值训练算法,例如误差反传算法,存在着收敛速度慢,容易陷入局部极值点等问题。因此提出旋转曲面变换粒子群优化算法,将被优化函数的局部极小点变换为全局最大点,同时不改变比局部极小点的值更小的区域的函数形状。此方法和粒子群优化相结合,能使待优化函数跳出局部极值点,提高训练神经网络权值的效率。实验结果显示,基于旋转曲面变换粒子群优化算法的神经网络,权值训练过程收敛速度较快,且自检误差和预报误差都较小,是一种有效的胺类有机物毒性分类方法。     

基于混沌自适应粒子群人工神经网络的气体在聚合物中的溶解模型

为提高溶解预测模型的效率和关联度, 建立基于混沌理论、自适应粒子群优化(PSO)算法和反向传播(BP)算法的混沌自适应PSO-BP神经网络模型, 并对二氧化碳(CO₂)在聚苯乙烯(PS)和聚丙烯(PP)中、氮气(N₂)在PS中的溶解度进行预测试验. 模型选用压力和温度作为输入参数, 使用试探法确定隐含层结点个数为8, 输出为预测的溶解度. 模型融合混沌理论、自适应PSO和BP算法各自的优势, 提高了训练速度和预测精度. 结果表明, 混沌自适应PSO-BP神经网络有很好的预测能力, 预测值与实验值相当吻合, 通过与传统BP神经网络和PSO-BP神经网络的比较可知, 其预测精度和相关性均明显较优, 预测平均绝对误差(AAD), 标准偏差(SD)和平方相关系数(R²)分别为0.0058, 0.0198和0.9914.     

基于LS-SVM的稀土萃取组分含量软测量

为了解决稀土萃取分离过程元素组分含量在线检测的难题, 提出了稀土萃取过程组分含量的一种最小二乘支持向量机(LS-SVM)软测量方法. 利用量子粒子群算法来优化LS-SVM的参数及核函数参数. 仿真结果表明, 所提出的软测量方法是有效的, 比已有的神经网络软测量方法能更好的实现稀土萃取过程中元素组分含量的在线估计.     

粒子群算法优化中药化学成分的毛细管电泳分离条件

建立了一种基于粒子群优化算法的毛细管电泳条件辅助优化方法。以丹参为研究对象,将改良的色谱指数方程用于评价酚酸类成分的电泳分离性能,用粒子群优化算法对分离条件进行全局寻优,获得最佳的区带电泳分离条件(5.0 mmol/L硼砂,18.5 mmol/L磷酸二氢钠,6.1%乙腈,运行电压18.2 kV)。为进一步改善分离,在所获优化条件下添加50.0 mmol/L SDS,在胶束电动毛细管色谱分离模式下使酚酸类成分(原儿茶醛、丹参素、丹酚酸B等)得到更好分离。本方法准确可靠,可推广应用于其他复杂化学体系的毛细管电泳分离条件优化。     

基于粒子群优化的波段选择方法在多组分同时测定中的应用

多组分同时测定时,组分间发生相互作用,应作波段选择.提出了改进的粒子群优化算法进行随机的波段选择.所提出的方法用于邻、间、对硝基苯酚的浓度预测,在208～481 nm范围内,以0.15 mol/L NaOH为溶剂,配置27组混合液作训练集,27组作预测集.训练集的均方根误差(RMSE)分别为0.1257、0.2228 和 0.0846; 预测集RMSE分别为0.2070、0.1507 和0.394,得到了较好的预测结果.     

基于模糊最小二乘支持向量机的蒸发过程预测控制研究

针对氧化铝蒸发过程的多变量、非线性和大滞后特点及不同时间和空间样本数据不同的特征,提出了基于末位淘汰机制的混沌粒子群算法的综合加权模糊最小二乘支持向量机蒸发过程预测控制方法.用变异混沌粒子群算法对模型预测控制进行滚动优化,计算出最优控制序列.以某氧化铝厂蒸发过程生产数据进行实验验证分析,结果表明: 模型预测结果中相对误差小于8%的样本达到93.9%,出口浓度稳定在240 g/L附近,其控制性能得到显著改善,同时也起到了降低能耗的目的.     

基于近红外光谱的汽油辛烷值预测与模型优化

在波长12000~4000 cm-1范围内采集了60组汽油样品的近红外光谱数据,并基于此数据研究汽油辛烷值预测过程中的模型优化问题:采用五折交叉验证法,比较了最小均方二乘法(PLS)、神经网络(ANN)和支持向量机(SVM)的辛烷值预测模型,发现SVM算法较稳定可靠,更适合于小样本情况下的光谱分析;SVM模型下分别采用网格寻优(Grid)、遗传算法(GA)和粒子群算法(PSO)进行惩罚参数C和RBF核函数参数gamma两个参数优化,总体最佳M SE分别为0.00444,0.0038和0.03262,GA优化参数能力最强;基于GA参数优化下SVM模型,研究了主分量分析(PCA)和连续投影算法(SPA)的特征优化方法,发现PCA泛化能力优于SPA,采用4个主分量(PCs)已经能达到原始光谱相当的预测性能。优化得到组合模型:PCA-GA-SVM,基本上满足工业级辛烷值预测的需要,方法对石油组分精确解析具有积极意义。     

基于粒子群算法的最小二乘支持向量机在红花提取液近红外定量分析中的应用

提出一种基于粒子群算法的最小二乘支持向量机(PSO-LS-SVM)方法,用于建立红花提取过程关键质控指标的定量分析模型.近红外光谱数据经波段选择、预处理和主成分分析(降维)后,利用粒子群优化(PSO)算法对最小二乘支持向量机算法中的参数进行优化,然后使用最优参数建立固含量和羟基红花黄色素A(HSYA)浓度的定量校正模型.将校正结果与偏最小二乘法回归(PLSR)和BP神经网络(BP-ANN)比较,并将所建的3个模型用于红花提取过程未知样本的预测.结果表明,BP-ANN校正结果优于PSO-LS-SVM和PLSR,但是对验证集和未知样品集的预测能力较差,而PSO-LS-SVM和PLSR模型的校正、验证结果相近,相关系数均大于0.987,RMSEC和RMSEP值相近且小于0.074,RPD值均大于6.26,RSEP均小于5.70％.对于未知样品集,pSO-LS-SVM模型的RPD值大于8.06,RMSEP和RSEP值分别小于0.07％和5.84％,较BP-ANN和PLSR模型更低.本研究所建立的PSO-LS-SVM模型表现出较好的模型稳定性和预测精度,具有一定的实践意义和应用价值,可推广用于红花提取过程的近红外光谱定量分析.     

粒子群最小二乘支持向量机结合偏最小二乘法用于芝麻油质量的鉴别

结合粒子群最小二乘支持向量机(PSO-LSSVM)与偏最小二乘法(PLS)提出一种基于气相色谱技术的新方法,对芝麻油进行真伪鉴别,并对掺伪品中掺假比例进行定量分析。采用主成分分析法(PCA)对857个样本的脂肪酸色谱数据进行分析,优选主成分作为最小二乘支持向量机(LSSVM)的输入向量。利用粒子群算法(PSO)优化LSSVM,构建芝麻油掺伪鉴别的两级分类模型,同时运用PLS建立掺伪芝麻油中掺伪油脂的定量校正模型,两级分类模型的准确率分别达到了100%和98.7%,定量分析模型的平均预测标准偏差(RMSEP)为3.91%。结果表明,本方法的鉴别准确性和模型泛化能力均优于经典的BP神经网络和支持向量机(SVM),可用于食用油脂加工和流通环节的质量控制,为食用油质量的准确鉴定提供了一条有效途径。     

粒子群算法结合支持向量机回归法用于近红外光谱建模

研究了最小二乘法支持向量机(LSSVM)应用于烟丝样品和小麦样品的近红外光谱建模,采用粒子群优化算法(PSO)优化LSSVM的参数。通过对烟草样品和小麦样品的近红外光谱建模和预测,并与常规的偏最小二乘法(PLS)比较发现,PSO-LSSVM法具有更好的预测效果和稳健性。     

Neural-network modeling of hot-compression test curves for calendering gasket materials

In this paper, we report on the use of neural networks (NNs) to estimate hot-compression test (HCT) curves for calendering gasket materials on the basis of their formulas. The NNs were used to demonstrate their potential during optimizing formulas for new calendering gasket materials. In the past, and to a large extent even now, the optimization of a new calendering gasket material was based on a process of trial and error, which takes a long time and is expensive because of the need for repeated experimental tests. And even after the completion of all this testing the final formula of the gasket material need not necessarily be the optimum one. We have shown that it is possible, with the assistance of a NN that was trained with appropriate data from just a small number of HCT curves, to satisfactorily investigate the valid ranges of the input data. On the basis of this investigation some valuable information was obtained that will make it easier to develop new calendering gasket materials. Using NNs, the speed of convergence to the final formula of the calendering gasket material can be much faster, because there is no need carry out many experimental HCT tests.     

Application of SiO2–water nanofluid to enhance oil recovery

The present study aims to investigate effects of nanofluid flooding on EOR and also compares its performance with water flooding in field scale using the published experimental data provided from core-scale studies. The nanofluid is based on water including silica nanoparticles. The relative permeability curves of water, nanofluid and oil for a light crude oil core sample obtained in an experimental study are used in this numerical investigation. A 2D heterogeneous reservoir model is constructed using the permeability and porosity of the last layer of SPE-10 model. It has been shown that nanofluid flooding can substantially improve the oil recovery in comparison with the water flooding case. Afterward, the operational parameters of the 13 injection and production wells have been optimized in order to meet the maximum cumulative oil production. First, pattern search (PS) algorithm was implemented which has a good convergence speed, but with a high probability of trapping in local optimum points. Particle swarm optimization (PSO) approach has also been employed, which requires a large number of population (to approach the global optimum) with so many simulations. Accordingly, a hybrid PSO–PS algorithm with confined domain is proposed. The hybrid algorithm starts with PSO and depending on the distribution density of the values of each parameter, confines the searching domain and provides a proper initial guess to be used by PS. It is concluded that the hybrid PSO–PS method could obtain the optimal solution with a high convergence speed and reduced possibility of trapping in local optimums.

     

Optimized partition of minimum spanning tree for piecewise modeling by particle swarm algorithm. QSAR studies of antagonism of angiotensin II antagonists

In quantitative structure-activity relationship (QSAR) modeling, when compounds in a training set exhibit a significant structural distinction between each other, in particular when chemicals of biological interest interacting on the receptor involve a different mechanism, it might be difficult to construct a single linear model for the whole population of compounds of interest with desired residuals. Developing a piecewise linear local model can be effective to circumvent the aforementioned problem. In this paper, piecewise modeling by the particle swarm optimization (PMPSO) approach is applied to QSAR study. The minimum spanning tree is used for clustering all compounds in the training set to form a tree, and the modified discrete PSO is applied to divide the tree to find satisfactory piecewise linear models. A new objective function is formulated for searching the appropriate piecewise linear models. The proposed PMPSO algorithm was used to predict the antagonism of angiotensin II. The results demonstrated that PMPSO is useful for improvement of the performance of regression models.