基于混合核函数和PLS的胎面尺寸预测模型

Tread dimension prediction model based on mixed kernel function and PLS

为实现对生产过程中胎面尺寸的准确预测,分析胎面生产工艺,结合混合核函数和偏最小二乘法(PLS)的特点构造了一种基于混合核函数和PLS的软测量模型。该预测模型把辅助变量通过输入非线性函数映射到高维特征空间,在高维特征空间中采用线性建模方法建立模型后经过非线性函数映射回原数据空间形成非线性模型。应用效果表明,该模型具有较强的学习和泛化能力,并且具有较高的预测精度,能较好地满足胎面实际生产要求。

In order to accurately predict the tread dimension in the course of production,a prediction model based on mixed kernel function and partial least squares（PLS）was constructed by analyzing the tread processing technique.Firstly,auxiliary variables were mapped into the high dimension space;then a prediction model was constructed using linear modeling method in this high dimension space;finally the non-linear prediction model was obtained from it through non-linear function mapping.The results showed that the prediction model had good learning and generalization ability,and excellent accuracy,which could meet the requirement of production.