摘要
以聚丙烯酸酯类系列水基聚合物包膜控释肥料为样品,测定了包膜肥料养分的释放曲线并原位测定了肥料包膜的中红外光声光谱,分析了不同肥料的养分释放曲线以及不同包膜材料的红外光声光谱特征;采用广义回归神经网络模型(GRNN),以肥料包膜红外光声光谱的主成分作为GRNN模型的输入层,并以包膜肥料养分释放曲线为输出层,构建了预测养分释放曲线的GRNN模型。结果表明,GRNN模型能快速有效地预测包膜肥料养分释放曲线,其预测相关系数(R2)达0.93以上;包膜的探测深度明显影响释放曲线的预测误差,最小预测误差为7.14%,平均为10.28%,且基于包膜表层红外光声光谱的预测误差最小。因此,结合GRNN模型,红外光声光谱可为包膜肥料养分释放曲线的快速预测提供新手段。
The acrylate-like materials were used to develop the polymer coated controlled release fertilizer,the nutrients release profiles were determined,meanwhile the Fourier transform mid-infrared photoacoustic spectra of the coatings were recorded and characterized;GRNN model was used to predict the nutrients release profiles using the principal components of the mid-infrared photoacoustic spectra as input.Results showed that the GRNN model could fast and effectively predict the nutrient release profiles,and the predicted calibration coefficients were more than 0.93;on the whole,the prediction errors(RMSE) were influenced by the profiling depth of the spectra,the average prediction error was 10.28%,and the spectra from the surface depth resulted in a lowest prediction error with 7.14%.Therefore,coupled with GRNN modeling,Fourier transform mid-infrared photoacoustic spectroscopy can be used as an alternative new technique in the fast and accurate prediction of nutrient release from polymer coated fertilizer.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2012年第2期330-333,共4页
Spectroscopy and Spectral Analysis
基金
国家“十二五”科技支撑计划项目(2011BAD11B01)
中国科学院院地合作项目
美国蓝月基金项目资助
关键词
红外光声光谱
包膜肥料
GRNN模型
释放曲线
Fourier transform infrared photoacoustic spectroscopy
Coated fertilizer
GRNN model
Release profile