摘要
采用具有强非线性表达能力的径基函数(RBF)-偏最小二乘回归(PLSR)相结合的建模方法建立了对苯二甲酸中对羧基苯甲醛含量的软测量模型。该组合方法应用径基函数实现自变量样本数据矩阵的非线性变换,应用偏最小二乘回归消除复共线性对模型预报精度的影响,从而使模型预报性能良好,与直接采用偏最小二乘法建模计算相比,预报误差下降了18.4%。
A novel method integrating the radial basis function （RBF） with partial least squares regression （PLSR）, which can describe complex nonlinear system, was established. Firstly, the method applied RBF to carry out the nonlinear transformation for independent variables. Secondly, PLSR was applied to remove the correlation among the nonlinear transformed variables and obtained the model with high predicting correctness. Further, RBF-PLSR was applied to model the soft sensor to detect the content of 4-carbonxybenzaldehyde in terephthalic acid. Satisfactory results were obtained, in comparison with the calculation result from the partial least square model, the prediction error decreased by 18.4%.
出处
《石油炼制与化工》
CAS
CSCD
北大核心
2005年第12期50-53,共4页
Petroleum Processing and Petrochemicals
基金
国家自然科学基金(20506003)
上海启明星项目(04QMX1433)
国家973计划(2002CB312200)
国家863项目(2002AA412110)。
关键词
径基函数
偏最小二乘回归
对苯二甲酸
对羧基苯甲醛
软测量
模型
radial basis function
partial least square regression
terephthalic acid
4-carboxybenzaldehyde
soft sensor
model
