不同颗粒度铁屑LIBS光谱的基体效应及定量分析研究

激光诱导击穿光谱(LIBS)对固体进行检测时,受固体的表面物理形态和化学特性影响较大,因此,基体效应分析对LIBS在线检测研究有重要的意义。为了提高LIBS对表面凹凸不平样品成分在线检测的准确度,进行了LIBS对不同颗粒度铁屑样品的定量分析。实验所用的9种铁屑样品性状为松散的粉末、颗粒或长条状,为防止激光与样品相互作用时发生飞溅,将样品粘到双面胶上进行固定。采用的激发波长为1 064 nm、脉冲能量为35 mJ,探测器延时和积分门宽分别设置为1和10μs。为评估样品颗粒度不同导致的基体效应对LIBS光谱的影响,首先,利用主成分分析(PCA)对系列样品进行分类,结果显示,粉末状的四个样品被分出,即颗粒度不同导致的基体效应是样品光谱信号差异的主要原因。其次,以C3、 C5两个样品研磨前后的基体元素特征谱线FeⅠ330.635 nm为研究对象,通过对比谱线的强度和相对标准偏差(RSD)发现,颗粒度越小,谱线强度越大,稳定性越好。为校正LIBS光谱基体效应的干扰,采用了样品研磨预处理和光谱数据预处理两种方法。将细长条状的C3和C5两个样品进行研磨,研磨后谱线的强度和稳定性有较大提升;分别研究了强度归一化、多元散射校正(MSC)以及两者结合对光谱进行处理的效果,三种光谱预处理均使谱线的稳定性得到显著提高。通过支持向量机(SVM)对Cu元素的定量结果进行了评估和对比,结果发现,采用研磨样品并结合强度归一化与MSC预处理得到的校正效果最优,最终使S1和S2两个待测样品的Cu元素预测相对误差(RE)分别降为1.745%和1.857%,预测均方根误差(RMSEP)降为0.020。该研究可为表面凹凸不平样品的LIBS检测提供一定的方法依据和参考。

Matrix Effect and Quantitative Analysis of Iron Filings with Different Particle Size Based on LIBS

Laser Induced Breakdown Spectroscopy(LIBS) analysis with solid target is greatly affected by the physical morphology and chemical properties of the sample surface. Therefore, the analysis of matrix effect is greatly significant for the study of LIBS online detection. In view of the fact that the current research objects of LIBS are mostly flat samples, in this work, LIBS was used for the quantitative analysis of iron filings with different granularity. The nine kinds of iron filings used in the experiment were loose powder, granules or strips. In order to avoid splashing during the laser ablation of iron filings, the samples were stuck to the double-sided tape for fixing before the experiment. The 1 064 nm laser with the energy of 35 mJ was used as the ablation source. And the delay and gate width of detector were 1 and 10 μs. In order to evaluate the influence of the matrix effect caused by the different particle size of the samples on the LIBS spectrum, firstly, the series of samples were classified by principal component analysis(PCA). The results showed that the four samples in powder form were separated, that is, the matrix effect caused by the difference in particle size was the main reason for the difference in the spectral signals of the sample. Secondly, the elemental characteristic line of Fe Ⅰ 330.635 nm in the C5 sample before and after grinding was taken as the research object. It was found by comparing the intensity and relative standard deviation(RSD) that the smaller the particle size, the stronger the line intensity and the better the stability. In order to correct the interference of the matrix effect on LIBS spectrum, two methods of sample grinding pretreatment and spectral data pretreatment were used. For the sample grinding pretreatment, the slender strip samples of C3 and C5 were grinded in the experiment, and the intensity and stability of the lines were greatly improved after grinding. For the spectral data pretreatment, the intensity normalization, multiple scattering correction(MSC) and the combination of them were studied respectively. The three spectral pretreatments all significantly improved the stability of the line. The quantitative results of Cu element were evaluated and compared by support vector machine(SVM) method. It was found that the calibration results obtained by the grinding samples combined with intensity normalization and MSC pretreatment were the best. Finally, the predicted relative error(RE) of the Cu element in S1 and S2 samples was reduced to 1.745% and 1.857%, respectively, and the root mean square error of prediction(RMSEP) was 0.020. This study can provide a certain method basis and reference for LIBS detection of the samples with irregular surface.