摘要
提出采用主成分-BP算法建立纸浆卡伯值近红外光谱法在线测量模型。结果表明,这种算法由于既考虑到了近红外光谱响应的非线性因素,又可防止BP算法在建模时出现“过拟合”的现象,利用该算法建立的纸浆卡伯值测量模型与一元回归、多元回归和主成分回归等线性方法相比,具有更高的预测精度。
The principal component-back propagation network(PCR-BP)algorithm is proposed to use in the determination of pulp kappa number by near-infrared spectroscopy.The algorithmcould deal with non-lin-earity of near-infrared spectral response and avoid the over-fitting in the modeling of BP neural network.Com-pared with the linear algorithms such as principal component regression(PCR),multivariable linear regres-sion(MLR)and unary linear regression(ULR),the proposed algorithmgave better precision of kappa number prediction.
出处
《分析测试学报》
CAS
CSCD
北大核心
2005年第4期10-12,16,共4页
Journal of Instrumental Analysis
基金
国家自然科学基金资助项目(29974011
30170756)
霍英东教育基金资助项目(71066)
