阻尼最小二乘法用于激光诱导击穿光谱重叠特征谱线分离提取

摘要

In recent years,the technology of laser induced breakdown spectroscopy has been developed rapidly.As one kind of new material composition detection technology,laser induced breakdown spectroscopy can simultaneously detect multi elements fast and simply without any complex sample preparation and realize field,in-situ material composition detection of the sample to be tested.This kind of technology is very promising in many fields.It is very important to separate,fit and extract spectral fea-ture lines in laser induced breakdown spectroscopy,which is the cornerstone of spectral feature recognition and subsequent ele-ments concentrations inversion research.In order to realize effective separation,fitting and extraction of spectral feature lines in laser induced breakdown spectroscopy,the original parameters for spectral lines fitting before iteration were analyzed and deter-mined.The spectral feature line of chromium (CrⅠ:427. 480 nm)in fly ash gathered from a coal-fired power station,which was overlapped with another line(FeⅠ:427. 176 nm),was separated from the other one and extracted by using damped least squares method.Based on Gauss-Newton iteration,damped least squares method adds damping factor to step and adjust step length dynamically according to the feedback information after each iteration,in order to prevent the iteration from diverging and make sure that the iteration could converge fast.Damped least squares method helps to obtain better results of separating,fitting and extracting spectral feature lines and give more accurate intensity values of these spectral feature lines.The spectral feature lines of chromium in samples which contain different concentrations of chromium were separated and extracted.And then,the intensity values of corresponding spectral lines were given by using damped least squares method and least squares method sepa-rately.The calibration curves were plotted,which showed the relationship between spectral line intensity values and chromium concentrations in different samples.And then their respective linear correlations were compared.The experimental results showed that the linear correlation of the intensity values of spectral feature lines and the concentrations of chromium in different samples,which was obtained by damped least squares method,was better than that one obtained by least squares method.And therefore,damped least squares method was stable,reliable and suitable for separating,fitting and extracting spectral feature lines in laser induced breakdown spectroscopy.%近年来，激光诱导击穿广谱技术发展迅速。作为一种用于物质成分探测的新技术，它具有简单、快速、无需复杂样品制备、多种元素同时检测等优点，可实现待测样品物质成分现场、在线的检测，在很多领域都极具应用前景。激光诱导击穿光谱特征谱线的分离拟合提取是光谱特征识别与后续元素浓度定量反演研究的基础。为实现激光诱导击穿光谱重叠特征谱线的有效分离拟合提取，采用阻尼最小二乘法，分析并确定了迭代前的初始拟合参数值，实现了在重叠特征谱线情况下对某火力发电厂粉煤灰中的铬元素特征谱线的分离提取。阻尼最小二乘法基于高斯－牛顿迭代，在迭代步长中引入阻尼因子，在迭代的过程中根据每一步迭代后所反馈的信息动态的调整迭代步长，从而有效防止了迭代的发散，保证了迭代的快速收敛，最终使得元素特征谱线拟合提取的效果更佳、所得到的特征谱线强度值更准确。分别采用阻尼最小二乘法和最小二乘法对不同浓度的样品中铬元素特征谱线进行分离拟合提取并给出特征谱线的强度值，作出特征谱线强度值关于元素浓度的定标曲线并对比两种方法所得结果的线性相关性。结果表明，阻尼最小二乘法所得结果的线性相关性更高，该方法稳定、可靠，适用于激光诱导击穿光谱重叠特征谱线的分离拟合提取。