In recent years, LIBS quantitative analysis based on multivariate regression has been received great attention, and variable selection is critical for improve accuracy of multivariate regression analysis of LIBS. In present study, sequential backward selection combined with random forest was proposed to improve the detection accuracy of sulfur and phosphorus element in steel. Firstly, the LIBS spectra line of S and P were identified by the NIST database. Secondly, input variables for RF calibration model were selected and optimized by SBS, and RF model parameters (ntree and mtry) were optimized by out-of-bag (OOB) estimation. Finally, optimized input variables and model parameters were employed to build SBS-RF calibration model for quantitative analysis of P and S in steel. Results showed that the SBS-RF model has a good predictions for S (R2 = 0.9991) and P (R2 = 0.9994) compared with univariate method, PLS model and traditional RF model. LIBS coupled with SBS-RF is an effective method for quality supervision and control of steel products.

中文翻译：

---

LIBS技术和顺序向后选择随机森林（SBS-RF）定量检测合金钢中的有害元素

近年来，基于多元回归的LIBS定量分析受到了广泛的关注，变量选择对​​于提高LIBS多元回归分析的准确性至关重要。为了提高钢中硫，磷元素的检测准确度，提出了顺序后向选择与随机森林相结合的方法。首先，通过NIST数据库识别S和P的LIBS光谱线。其次，通过SBS选择和优化RF校准模型的输入变量，并通过袋外（OOB）估计对RF模型参数（ntree和mtry）进行优化。最后，采用优化的输入变量和模型参数建立了SBS-RF校准模型，用于钢中P和S的定量分析。结果表明，SBS-RF模型对S的预测很好（R2 = 0。9991）和P（R2 = 0.9994）与单变量方法，PLS模型和传统RF模型进行比较。LIBS与SBS-RF结合使用是对钢材进行质量监督和控制的有效方法。