Different manufacturers do not produce the same quality of children’s Ca–Fe–Zn oral liquid due to different production materials and processes. To improve the phenomenon of counterfeit and imitation oral liquid on the market and effectively monitor its quality, laser-induced breakdown spectroscopy (LIBS) fingerprinting with sample preparation methods can provide a tool for real-time and rapid detection of oral liquids. The sample preparation methods include filter paper adsorption (FPA), filter paper adsorption with elemental Cu (FPA with Cu), adding dropwise to glass slides (ADS), adding dropwise to glass slides with elemental Cu (ADS with Cu), and gel preparation (GP). This work collected LIBS spectrum of oral liquids from eight manufacturers. The model for eXtreme Gradient Boosting (XGBoost) was constructed for classifying oral liquids based on five sample preparation methods. The accuracy was 91.25, 94.17, 55.42, 91.25, and 91.29%, respectively. The results show that the FPA method is more straightforward, efficient, and less affected by the specificity of the color of the sample. Both ADS and GP are susceptible to the color characteristics of the sample and are not well suited to the direct detection of transparent liquids. This work demonstrated that oral liquids could be discriminated by analyzing LIBS spectrum combined with the XGBoost model. Additionally, sample preparation, like the simple FPA method, can improve the accuracy of LIBS classification.

不同厂家由于生产材料和工艺不同，生产的儿童钙铁锌口服液质量也不尽相同。为了改善市场上假冒仿制口服液的现象，有效监控口服液的质量，激光诱导击穿光谱（LIBS）指纹图谱结合样品制备方法可以为口服液的实时快速检测提供工具。样品制备方法包括滤纸吸附（FPA）、滤纸吸附单质铜（FPA with Cu）、滴加到载玻片上（ADS）、滴加到载玻片上单质铜（ADS with Cu）、凝胶制备（全科医生）。这项工作收集了八家制造商的口服液的 LIBS 光谱。构建了极限梯度提升 (XGBoost) 模型，用于基于五种样品制备方法对口服液进行分类。准确率分别为 91.25%、94.17%、55.42%、91.25% 和 91.29%。结果表明，FPA方法更加简单、高效，且受样品颜色特异性的影响较小。ADS和GP都容易受到样品颜色特性的影响，不太适合透明液体的直接检测。这项工作证明，通过分析 LIBS 光谱并结合 XGBoost 模型可以区分口服液。此外，样品制备与简单的 FPA 方法一样，可以提高 LIBS 分类的准确性。