LIBS结合图像筛选方法提高钢铁中Cu、Cr、Mn元素检测稳定性研究

优质特种钢材和低端粗钢之间的性能差异主要受其构成元素种类及其成分含量的影响，因此，如何快速准确地对物质成分进行定性及定量分析对钢铁产品的质量评估至关重要。针对传统方法难以实现对钢铁合金成分的快速准确检测的难题，采用激光诱导击穿光谱(LIBS)结合等离子体图像信息的方法，通过快速地对不同元素的特征光谱强度与激发生成的等离子体图像进行采集，分析两者之间的相关性，并通过提取的图像特征信息的异常值剔除了部分无效光谱数据，进而实现了对钢铁成分的高精度分析。通过分析延迟时间和激光能量等不同实验条件对元素特征光谱强度及其对应等离子体图像的影响规律，不仅证明了等离子体图像与光谱之间存在相关性，还利用等离子体图像特征信息的局部最优值确定了最优延迟时间、激光能量分别为1 000 ns与50 mJ,并根据图像特征的平均阈值来筛选无效光谱数据。结果表明，图像筛选优化数据后，各元素谱线校准模型的决定系数(R²)分别从原始数据的0.978、0.986、0.957、0.935提升至0.995、0.997、0.968、0.957,且其定标曲线对未知样品元素的预测浓度相对标准偏差(RSD)下降为原...

Enhancement of Detection Stability for Cu, Cr, and Mn in Steel Using LIBS Coupled with Image Screening Methods

The performance differences between high-quality special steel and low-end crude steel are mainly influenced by the types of constituent elements and their composition levels. Therefore, how to rapidly and accurately analyze the material composition is crucial for assessing the quality of steel products. Addressing the challenge of traditional methods in achieving rapid and accurate detection of steel alloy components, this paper adopted the Laser-Induced Breakdown Spectroscopy (LIBS) combined with plasma image information. It collected the characteristic spectral intensities of different elements and the plasma images generated through rapid acquisition, analyzed the correlation between the two, and removed some invalid spectral data by extracting abnormal values from the image feature information. This allowed for a high-precision analysis of the steel composition. This paper, by analyzing the influence of different experimental conditions such as delay time and laser energy on the characteristic spectral intensity of elements and their corresponding plasma images, not only demonstrated the correlation between plasma images and spectra but also determined the optimal delay time and laser energy as 1000ns and 50mJ respectively based on local optimal values of image features. It further filtered out invalid spectral data based on the average threshold of image features. The results showed that after optimizing data through image filtering, the determination coefficients (R2) of the calibration models for each element''s spectral lines improved from 0.978, 0.986, 0.957 and 0.935 of the original data to 0.995, 0.997, 0.968 and 0.957 respectively. Additionally, the relative standard deviation (RSD) of the predicted concentrations of unknown sample elements by the calibration curves decreased to about 50% of the RSD of the original data''s predicted concentrations. Therefore, it can be concluded that the use of LIBS combined with image filtering methods can reduce errors in quantitative analysis and improve the accuracy of prediction results.