【摘要】 应用偏最小二乘法(partial least square,PLS)建立烤烟中氯的近红外分析模型。通过光谱预处理和模型的逐步优化最终确定,相关系数为0.989 6,均方估计残差(RMSEC)0.019 0,预测标准差(RMSEP)为0.085 6,交互验证均方差(RMSECV)为0.038 5,模型优化指数(PI)为80,主因子(Factor)为16。选取有代表性的57个验证样品对模型进行检验,检验结果平均相对误差为10.41%。通过对模型的t-检验,行业标准方法和近红外光谱法的测定值无显著性差异。更多还原

【Abstract】 Chemometrics was applied to data treatment in the rapid NIRS determination of chlorine in flue-cured tobacco.The method of partial least square was used in the establisment of analytical mathematical model through spectral pretreatment and step-by-step optimization,and the following parameters were obtained: correlation coefficient of the analytical model 0.989 6,RMSEC 0.019 0,RMSEP 0.085 6,RMSECV 0.038 5,PI 80 and main factor 16.The analytical model was tested by analysis of 57 typical samples for their chlorine contents by the present method.Average relative error of the results obtained were 10.41%.By applying the t-test to test the analytical results,no significant difference was found between the values found by the proposed method(i.e.,the NIRS method) and the method of trade standard.更多还原