首页|激光诱导击穿光谱的自适应基线校正方法

激光诱导击穿光谱的自适应基线校正方法

Adaptive Baseline Correction Method for Laser-induced Breakdown Spectroscopy

扫码查看
原文链接
  • 维普
  • 万方数据
为了降低远程激光诱导击穿光谱原位分析中连续背景辐射、随机噪声和样品基体效应对特征光谱的影响,提出一种通过适应度函数将粒子群优化和非对称重加权正则最小二乘相融合的自适应基线校正方法,旨在提高远程激光诱导击穿光谱的分析能力.以三种掺杂相近微量元素的铝基合金标样为研究对象,相同实验条件下利用所提出的基线校正方法研究并对比自适应迭代重加权正则最小二乘算法、非对称重加权正则最小二乘算法处理后的光谱信噪比与拟合基线的变化趋势.为了验证所提方法的有效性,构建了三次核函数支持向量机的铝基合金标样精细分类模型,其独立测试集的混淆矩阵分类准确率为100%.结果表明,自适应基线校正方法所拟合的基线变化趋势与激光诱导击穿光谱的实际基线轨迹相吻合,且特征光谱能够清晰地分离.该方法可有效降低随机噪声对光谱数据分析的影响,对于提高远程激光诱导击穿光谱原位分析的准确性和可靠性具有意义.
An adaptive baseline correction method was proposed by combining Particle Swarm Optimization(PSO)and asymmetrically reweighted Penalized Least Squares(arPLS)through fitness function to reduce or eliminate the influence of continuous background radiation,random noise,and sample matrix effect on the characteristic spectrum in the in-situ analysis of long-range Laser-Induced Breakdown Spectroscopy(LIBS).The PSO-arPLS approach is intended to increase the remote LIBS's analytical capacity.The method adds the regular function to the loss function,turning the limited problem into an unconstrained problem,and uses the"asymmetric weighting"approach to accomplish the adaptive baseline correction goal.PSO and arPLS were combined by fitness function and applied to an aluminum-based alloy with trace metal elements as the research sample.Particle swarm automatically discovered the optimal parameters of arPLS fitting baseline to achieve the balance of weight vector and smoothing parameters in baseline signal.The spectral Signal-to-Noise Ratio(SNR)and noise reduction effect of the short-wave and long-wave spectral segments were examined,using the gathered 6061 series aluminum base alloy LIBS as an example.PSO-arPLS was then compared with the conventional airPLS and asPLS calibration methods.Finally,the kernel Support Vector Machine(SVM)model is trained using the original LIBS data set of aluminum base alloy and the data set after baseline correction using the aforementioned three methods,and the confusion matrix of the model is analyzed to confirm the validity of the suggested baseline correction method.The results demonstrate that the suggested PSO-arPLS approach can not only reduce the spectral baseline fluctuation but also increase the spectral SNR and boost the spectral dynamic range when compared to conventional airPLS and asPLS calibration methods.PSO-arPLS can effectively preserve the LIBS signal with spectral characteristics after correction.In contrast,the fitting baselines proposed by airPLS and asPLS methods lead to serious loss of LIBS characteristic peak region and low fitness for eliminating LIBS noise.Therefore,when correcting LIBS spectra with low SNR,the baseline trend can basically overlap with the uncharacterized peak region,and the PSO-arPLS method also shows a good effect in noise elimination.Continuous background radiation is successfully monitored,and the fitted baseline lies within the cross-range of the characteristic spectrum and noise signal.In conclusion,the PSO-arPLS algorithm performs well in noise reduction and corrects the low SNR LIBS spectral baseline region,which nearly overlaps with the non-characteristic peak region.Three different types of aluminum-based alloys doped with comparable trace elements were employed as research objects in this paper.Under the same experimental conditions,the fitting baseline was utilized to analyze and evaluate the variation trend of the spectral SNR addressed by the airPLS and asPLS algorithms.A cubic kernel SVM fine classification model for aluminum-base alloys was created to confirm the efficacy of the suggested approach.The independent test set's confusion matrix classification had 100%accuracy.The kernel SVM model was trained using the original LIBS data set and the suggested baseline corrected data set,and the model's confusion matrix was examined.The outcomes demonstrate that the PSO-arPLS technique could more effectively classify and identify the LIBS data.The 11.8%increase in classification accuracy further demonstrates the PSO-arPLS method's beneficial effects on data analysis.PSO-arPLS approach has strong noise robustness and can overcome the effects of continuous background radiation and LIBS noise at a great distance.Additionally,the proposed method for LIBS adaptive baseline correction can be used in real-world industrial contexts and significantly enhances remote LIBS's capacity for qualitative analysis.

Long-range laser-induced breakdown spectroscopyBaseline correction methodIn situ analysisKernel support vector machineAluminum base alloy standard sample

闫红宇、赵宇、陈媛媛、刘昊、王杰、张瑞、王志斌

展开 >

中北大学 仪器与电子学院,太原 030051

中北大学 信息与通信工程学院,太原 030051

远程激光诱导击穿光谱 基线校正方法 原位分析 核函数支持向量机 铝基合金标样

国家自然科学基金山西省自然科学研究面上项目

62105302202203021221103

2024

10.3788/gzxb20245306.0630001

光子学报
中国光学学会 中国科学院西安光学精密机械研究所

光子学报

CSTPCD北大核心
影响因子：0.948
ISSN：1004-4213
年,卷(期)：2024.53(6)