基于小波相关向量机的产品质量模型  

Product quality model based on wavelet relevance vector machine

作　　者：何飞[1] 黎敏[1] 阳建宏[1] 徐金梧[1] 

机构地区：[1]北京科技大学机械工程学院,北京100083

出　　处：《北京科技大学学报》2009年第7期934-938,共5页Journal of University of Science and Technology Beijing

基　　金：国家自然科学基金资助项目(No.50705069);国家高技术研究发展计划资助项目(No.2007AA04Z169);教育部博士点基金资助项目(No.20070008050);湖北省重点基金资助项目(No.2007A19)

摘　　要：针对常用的质量建模方法精度不高且难以给出预测区间,提出了基于小波相关向量机的产品质量模型.应用仿真数据和带钢热镀锌锌层质量的实际生产数据分别建立了小波相关向量机模型.结果表明,小波相关向量机方法与支持向量机及传统的相关向量机相比,具有更好的预测精度,而且给出了预测区间.多组带钢热镀锌锌层质量实际数据的相对预测误差的平均值为4·52%,为保证产品质量提供必要的决策支持和分析手段.According to the fact that a common method for product quality modeling has not very high modeling accuracy and its prediction intervals can not be given, a model of product quality based on wavelet relevance vector machine was proposed. The simulation data and the real field data of zinc coating mass from strip hot-dip galvanizing were used for validation. The results show that the model based on wavelet relevance vector machine has a higher prediction precision than those based support vector machine and relevance vector machine, and its prediction intervals can he given. The zinc coating mass forecasting model based on wavelet relevance vector machine for multi-group data has an average of the relative prediction error of 4.52 % ; thus for the quality control, it provides the necessary decision supports and analysis tools.

关 键 词：小波核函数 相关向量机 预测区间 锌层质量 质量模型 

分 类 号：TG335.22[金属学及工艺—金属压力加工] O212.8[理学—概率论与数理统计]