中国机械工程

• 先进材料加工工程 • 上一篇    下一篇

基于小波消噪和优化支持向量机的板形模式识别

吴忠强;康晓华;于丹琦   

  1. 燕山大学工业计算机控制河北省重点实验室,秦皇岛,066004
  • 出版日期:2018-01-10 发布日期:2018-01-04
  • 基金资助:
    国家自然科学基金资助项目 (U1260203);
    河北省"三三三人才工程"基金资助项目(A2016015002);
    河北省自然科学基金资助项目(F2016203006,F2017203304)
    National Natural Science Foundation of China (No. U1260203)
    Hebei Provincial Natural Science Foundation of China (No. F2016203006,F2017203304)

Flatness Pattern Recognition Based on Wavelet De-noising and Optimized SVM

WU Zhongqiang;KANG Xiaohua;YU Danqi   

  1. Key Lab of Industrial Computer Control Engineering of Hebei Province, Yanshan University, Qinhuangdao,Hebei,066004
  • Online:2018-01-10 Published:2018-01-04
  • Supported by:
    National Natural Science Foundation of China (No. U1260203)
    Hebei Provincial Natural Science Foundation of China (No. F2016203006,F2017203304)

摘要: 提高板形模式识别的精度利于得到高精度的控制效果。针对实测的板形信号中混有噪声信号的问题,利用双变量阈值小波去噪,克服了软硬阈值函数在处理小波系数方面存在的缺点,使得去噪的效果更好。将去噪后的板形信号离散化,作为支持向量机的学习样本,建立识别模型。引入布谷鸟算法优化支持向量机的参数。仿真结果表明,相比于粒子群和遗传算法,布谷鸟优化算法所需匹配的参数少,而获得的最优解更好。

关键词: 板形模式识别, 双变量阈值小波函数, 支持向量机, 布谷鸟优化算法

Abstract: Improving the precision of flatness pattern recognition was useful to get high precision control effectivenesses. Among the measured strip shape signals there were noise signals. Firstly, bivariate wavelet threshold method was used to eliminate noises which might overcome the disadvantages of the soft and hard threshold function methods in dealing with wavelet coefficients, and better de-noising effectivenesses might be got. The flatness signals were discretized after de-noising, and as learning samples of SVM, so a recognition model was established. The cuckoo algorithm was introduced to optimize the parameters of SVM. The simulation results show that compared with the particle swarm, genetics algorithm, the cuckoo optimization algorithm needs less matching parameters, and the better optimal solutions will be obtained.

Key words: flatness pattern recognition, bivariate wavelet threshold function, support vector machine(SVM), cuckoo algorithm

中图分类号: