论文部分内容阅读
为监测奶粉中的镁(Mg)元素含量,本研究利用激光诱导击穿光谱(LIBS)技术对奶粉中Mg元素进行定量检测.对于每个样品,采用压片机在20 MPa压力下进行压片处理,然后利用高精度光谱仪在200～750 nm波段范围内获取压片样品的LIBS光谱.根据LIBS光谱特征,将光谱划分为4个波段,并进行初步的波段优选和光谱预处理分析.在此基础上,采用竞争性自适应重加权算法(CARS)对波长变量进行优选,再应用偏最小二乘法(PLS)建立奶粉中Mg元素含量的预测模型,并对预测集样本进行预测.研究结果表明,LIBS技术结合CARS变量选择方法可以用于奶粉中Mg元素含量的定量检测,最优CARS-PLS预测模型的校正集和预测集的决定系数及平均相对误差分别为0.9999,0.20％和0.9742,3.29％,优于原始光谱所建立的PLS模型,且所用波长变量仅为PLS模型的7.7％.由此表明,CARS方法能有效选择有用的波长变量,可简化预测模型及提高预测模型的稳定性.本研究为奶粉中镁元素含量的快速定量分析提供参考.“,”In order to monitor the content of magnesium ( Mg) in milk powder, laser induced breakdown spectroscopy ( LIBS) technique was used to determine Mg content in milk powder quantitatively in this research. Tablet machine was used to press each milk powder sample under pressure of 20 MPa, then the LIBS spectra of pressed samples were acquired by a high precision spectrometer in the wavelength range of 200～750 nm. The spectra were divided into 4 wavelength bands according to the characteristics of LIBS spectra, then preliminary optimization of wavelength band selection and spectral pretreatment were conducted. After that, competitive adaptive reweighted sampling ( CARS ) method was used to select informative wavelength variables, and partial least square ( PLS) method was used to develop calibration model of magnesium in milk powder. At last, the calibration model was used to predict the magnesium content in prediction samples. The results indicated that LIBS technique combined with CARS method is feasible for determining magnesium content in milk powder quantitatively. The coefficients of determination and average relative errors of the best CARS-PLS model in calibration and prediction sets were 0. 9999, 0. 20％ and 0. 9742, 3. 29％, respectively. The result of the best CARS-PLS model was better than that of PLS model based on original spectra. Furthermore, the number of used variables in the best CARS-PLS model was only 7. 7％ of that used in PLS model. So it can be known that CARS method can select the informative wavelength variables effectively, simplify calibration model and improve stability of calibration model. This study can provide a reference for rapid quantitative analysis of magnesium in milk powder.