CIESC Journal

    Next Articles

Building Predicting Model of Paper Tensile Strength Based on Partial Least-Squares Approach

LI Yuan-hua, TAO Jing-song, LI Ji-geng, LIU Huan-bin   

  1. State Key Lab of Pulp and Paper Engineering, South China University of Technology, Guangzhou 510640, China
  • Online:2014-03-24 Published:2014-09-05
  • Supported by:

    supported by the Major Science andTechnology Projects of Science and Technology in Guangdong (2010A080801002); the Fund of State Key Lab of Pulp and Paper Engineering, South China University of Technology (201233); the Fundamental Research Funds for the Central Universitie (2014ZZ0055)

基于偏最小二乘法的纸张抗张强度预测模型

李远华, 陶劲松, 李继庚, 刘焕彬   

  1. 华南理工大学制浆造纸工程国家重点实验室, 广东 广州 510640
  • 通讯作者: 陶劲松
  • 基金资助:

    2010广东省科技计划项目重大科技专项(2010A080801002);华南理工大学制浆造纸工程国家重点实验室开放基金(201233);2014中央高校基本科研业务费专项资金资助(2014ZZ0055)

Abstract: The problems existing in the mechanism models include hard to obtaining the parameters and most of them had little relationship with production which reduced its practicability. To solve it, partial least-squares method (PLS) was taken to establish model to predict paper strength based on a production line of a corrugated paper mill. Selecting parameters and its data aided with mechanism analysis, this model was built up and the key parameters were obtained. The results showed that this model had a good precision that its Person's value was 0.732, Root Mean Square Error (RMSE) was 276 N·m-1, and Mean Relative Error (MRE) was 5.17%. Besides that, the model had a good analytical ability that the key elements could be interpreted very well from mechanism angle.

Key words: chemical processes, paper sheet tensile strength, product engineering, partial least-squares, prediction

摘要: 为了解决抗张强度机理模型中参数难以测量,偏离工厂生产变量,实际指导关联性差的问题。以一瓦楞原纸生产线为研究对象,先通过经验分析、筛选影响纸张抗张强度的生产过程变量,再收集对应变量的生产数据后,应用偏最小二乘法建立抗张强度预测模型并对重要影响变量进行分析。结果表明该模型具有较好精度,其皮尔逊相关系数r=0.732,均方根误差RMSE值为276 N·m-1,平均相对误差MRE值为5.17%。同时得到了影响纸张抗张强度的6个重要生产过程变量,通过机理分析和现场验证,发现结果具有较好的现实吻合度。

关键词: 化学过程, 纸张抗张强度, 产品工程, 偏最小二乘法, 预测

CLC Number: