论文部分内容阅读
建立了适用于激光诱导击穿光谱探测的多元线性回归、神经网络回归和支持向量机回归三种定量反演算法模型,以水体重金属Ni为例进行了回归实验测试和对比分析.多元线性回归、神经网络回归和支持向量机回归的平均相对标准偏差分别为7.60%,4.86%,2.35%;最大相对标准偏差分别为23.35%,15.20%,8.29%;平均相对误差分别为25.98%,10.58%,2.72%,最大相对误差分别为116.47%,47.38%,9.89%.研究为进一步实现水中痕量金属元素的快速定量分析提供了方法和数据参考.
Three quantitative inversion algorithms, multivariate linear regression, neural network regression and support vector machine regression, which are suitable for laser-induced breakdown spectroscopy detection, were established, and the regression experimental tests and comparative analyzes were carried out using heavy metal Ni as an example.Multivariate linear regression, The average relative standard deviations (RSDs) of neural network regression and support vector regression were 7.60%, 4.86% and 2.35%, respectively. The maximum relative standard deviations were 23.35%, 15.20% and 8.29% respectively. The average relative errors were 25.98% and 10.58% 2.72% and the maximum relative error were 116.47%, 47.38% and 9.89%, respectively.The research provided a method and data reference for further rapid quantitative analysis of trace metal elements in water.