论文部分内容阅读
基于单脉冲激光诱导击穿光谱(LIBS)检测技术,以Cu在324.74nm处和Ca在317.95nm处的两条特征谱线强度作为多元非线性定标的自变量,饲料中Cu含量作为因变量,对猪饲料中Cu元素的含量进行了定量分析。比较了单变量分析方法、交叉降维近似多元非线性模型、多元二次非线性模型和平方降维近似多元非线性模型的分析结果,并对验证样品进行预测分析。结果表明,交叉降维近似多元非线性模型与其他三种分析方法相比预测效果更好,建模集预测浓度与实际浓度的相关拟合系数为0.9799,预测集的相关拟合系数为0.8597,平均相对误差为8.12%。
Based on single-pulse laser-induced breakdown spectroscopy (LIBS), two characteristic line intensities of Cu at 324.74 nm and Ca at 317.95 nm were used as independent variables for multivariate nonlinear calibration. Cu content in feed was used as dependent variable , Quantitative analysis of the content of Cu in pig feedstuffs was carried out. The univariate analysis method, the cross-dimensional reduction approximate multivariate nonlinear model, the multivariate quadratic nonlinear model and the square dimensionality reduction approximate multivariate nonlinear model were compared, and the verification samples were predicted and analyzed. The results show that the cross-dimensional reduction approximate multivariate nonlinear model is better than the other three analysis methods, the correlation coefficient of the predicted concentration and the actual concentration of the model set is 0.9799, the correlation coefficient of the prediction set is 0.8597, The average relative error was 8.12%.