LIBS quantitative analysis method based on multi-model calibration with acoustic feature labeling

Abstract

To improve the accuracy and long-term reproducibility of LIBS quantitative analysis, a method based on multi-model calibration with acoustic feature labeling (AFL-MMC) was proposed. The quantitative analytical capabilities of single-model calibration and AFL-MMC methods were comparatively investigated by analyzing trace-level Mn, Mo, V, and Cr in alloy steel specimens. For acoustic feature labeling, the first peak acoustic amplitude (AA), the acoustic energy (AE) before the first echo, and the acoustic wave (AW) band before the first echo from the plasma were selected. The results demonstrated a strong linear correlation between both the first peak AA and the pre-first echo AE with spectral intensity. When the pre-first echo AE was used as a feature label, the average relative errors were 14%, 17%, 15%, 17% and 19%, with the maximum average relative error decreasing from 29% to 16%. AE demonstrated the most pronounced improvement, followed by AA, whereas AW exhibited comparatively limited effectiveness. These results indicate that the integration of AFL-MMC significantly improves the accuracy and long-term reproducibility of LIBS quantitative detection.