无监督数据挖掘辅助激光诱导击穿光谱用于铜冶炼光谱结构解析

在有色冶炼领域,元素成分检测是保证冶炼质量的重要一环。目前国内有色冶炼企业多采用X射线荧光光谱法进行检测,该方法需要样品制备,造成冶炼状态无法实时反馈,严重影响冶炼过程优化。研究了无监督数据挖掘算法辅助激光诱导击穿光谱技术用于铜冶炼光谱结构解析。实验中,首先选择4种铜冶炼物料作为实验样品,然后利用激光诱导击穿光谱技术(LIBS)激发样品获得18750个光谱数据,通过盲源分离技术对所有光谱进行分析,最终提取得到3个特征光谱。进一步研究发现,3个特征光谱与Cu、Fe、Ca元素光谱有一一对应关系。在此基础上,提出了LIBS光谱的定量化评价指标,量化结果表明分解模型对18750个光谱都能达到很高的评分,说明铜冶炼光谱能够良好地被3个特征光谱重构,即铜冶炼光谱存在显著的光谱结构。以上结论在实际应用中具有重要研究价值,可用于光谱快速评价、异常光谱剔除、光谱信号提纯、元素谱线选取、样品定性/半定量分析等,为LIBS技术应用于在线铜冶炼成分分析奠定基础。

Unsupervised data mining assisted laser-induced breakdown spectroscopy was used to analyze the spectral structure of copper smelting

In the field of nonferrous smelting, element component detection is an important part to ensure the smelting quality. At present, domestic non-ferrous smelting enterprises mostly use X-ray fluorescence method for detection, which requires sample preparation, resulting in no real-time feedback of smelting state, which seriously affects the optimization of smelting process. Unsupervised data mining algorithm assisted laser-induced breakdown spectroscopy was used to analyze the spectral structure of copper smelting. In the experiment, four kinds of copper smelting materials were first selected as experimental samples, and then 18750 spectral data were obtained by LIBS instrument excitation samples. All spectra were analyzed by blind source separation technique, and finally three characteristic spectra were extracted. Further study found that the three characteristic spectra had a one-to-one relationship with the spectra of Cu, Fe and Ca. On this basis, the quantitative evaluation index of LIBS was proposed, and the quantitative results showed that the decomposition model could achieve a high score for all 18750 spectra, indicating that the copper smelting spectrum could be well reconstructed by the three characteristic spectra, that was, the copper smelting spectrum had a significant spectral structure. The above conclusions had important research value in practical applications, and could be used for rapid spectral evaluation, abnormal spectral elimination, spectral signal purification, element spectral line selection, qualitative/semi-quantitative analysis of samples, etc., laying a foundation for the application of LIBS technology in the analysis of copper smelting components on line.