基于果蝇算法优化支持向量回归机的纺丝性能预测  

Prediction of Spinning Performance Based on Fruit Fly Algorithm Optimized Support Vector Regression

作　　者：郭凡[1] 丁永生[1,2] 郝矿荣[1,2] 任立红[1,2] 肖纯材[1] 

机构地区：[1]东华大学信息科学与技术学院,上海201620 [2]数字化纺织服装技术教育部工程研究中心,上海201620

出　　处：《系统仿真学报》2014年第10期2360-2364,共5页Journal of System Simulation

基　　金：国家自然科学基金重点项目(61134009);国家自然科学基金(61473077);国家自然科学基金海外及港澳学者合作研究基金(61428302);上海领军人才专项资金;上海市科学技术委员会重点基础研究项目(13JC1407500;11JC1400200);上海市教育委员会科研创新项目(14ZZ067)

摘　　要：提出了基于支持向量回归机(SVR)的纤维伸长率预测方法。该方法根据有限的学习样本,建立了纺丝速度、纺丝温度、吹风速度和吹风温度4个参数作为影响纤维质量的关键参数组成的预测参数集合与纤维伸长率之间的一种非线性映射,对纤维伸长率进行预测。以工业过程中实际的样本数据得到纺丝伸长率的SVR预测模型。采用果蝇优化算法优化SVR预测模型的参数,并与BP神经网络及传统SVR的预测模型进行结果对比分析。实验结果表明,果蝇算法优化SVR预测方法优于BP及传统SVR预测方法,有着良好的泛化能力,能够对实际纤维生产过程起到一定的指导作用。The prediction of fiber elongation rate was studied based on Support Vector Regression（SVR） method. Based on the limit samples, taking spinning velocity, spinning temperature, quenching velocity, and quenching temperature as parameters set, and fiber elongation rate as predict objective, the SVR method established a nonlinear mapping between predict parameters set and fiber elongation rate to predict performance of the fiber spinning process. The predictive model for fiber elongation rate was obtained based on existing industry data, and the Fruit Fly Optimization Algorithm（FOA） was used to optimize the parameters in SVR. The prediction results of the proposed method were compared with those of BP neural networks and traditional SVR. Simulation results demonstrate that the proposed method can obtain better predict results, and has a good ability of generalization. The method is of guiding value to actual production process.

关 键 词：支持向量回归机 果蝇优化算法 BP神经网络 纤维拉伸率 性能预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]