基于集成多支持向量回归融合的上浆率在线软测量方法  

Online soft measurement of sizing percentage based on intergrated multiple SVR fusion by Bagging

作　　者：田慧欣[1,2] 贾玉凤 

机构地区：[1]天津工业大学电气工程与自动化学院,天津300387 [2]天津工业大学电工电能新技术天津市重点实验室,天津300387 [3]天津三星通信技术研究有限公司,天津300381

出　　处：《纺织学报》2014年第1期62-66,共5页Journal of Textile Research

基　　金：国家自然科学基金项目(61203302);天津市应用基础及前沿技术研究计划项目(11JCYBJC07000);中国纺织工业联合会科技指导性项目(2012060)

摘　　要：现有浆纱过程上浆率的检测无法实现实时在线测量,直接影响纱线产品质量的保障。为此,提出一种新的基于Bagging多SVR融合的建模方法,建立上浆率在线软测量模型。首先对浆纱过程进行分析,确定影响上浆率的主要因素。将这些主要因素作为模型的输入,用不同的核函数、损失函数和参数建立基本SVR模型。使用Bagging将多个SVR模型进行融合,使其优势互补,得到最终的上浆率在线软测量模型。使用实际生产数据对模型进行检验,并将其与传统软测量方法进行比较。结果表明基于Bagging多SVR融合的上浆率在线软测量模型的性能优于传统软测量模型,并具有较高的测量精度,完全能够满足实际生产的需要。Nowadays, real-time online detecton of the sizing percentage cannot be realized yet, which directly influence the quality assurance of yarn products. Therefore, a new modeling method is proposed for real-time online soft measurement of sizing percentage based on multiple SVR fusion by Bagging. Firstly, the main factors that influence sizing percentage are obtained by analyzing the sizing process. Then the basic SVR soft sensor models are built by different kernel functions, different loss functions and different parameters. Finally, the different basic SVR models are integrated by Bagging so that complementing mutual advantages is realized and finally the percentage is obtained. The model is verified by practical traditional soft senor method, demonstrating that the former online soft measurement model of sizing production data and compared with the outperforms the latter, which has higher testing precision, and can fully satisfy the needs of practical production.

关 键 词：上浆率 浆纱过程 软测量 BAGGING 支持向量回归机 

分 类 号：TS105.12[轻工技术与工程—纺织工程]