A novel multi-strategy collaborative correction method for LIBS spectral intensity under laser energy fluctuations

Laser-induced breakdown spectroscopy (LIBS) enables real-time, multi-element quantitative analysis of metals in atmospheric particulate matter. However, laser energy fluctuations alter plasma parameters (e.g., plasma temperature). Moreover, due to differences in transition probabilities and excitation characteristics among spectral lines, such variations also lead to inconsistent intensity variation trends across different lines. This increases nonlinearity in spectral intensity relationships and degrades quantitative model performance.

This paper proposes a novel Multi-Strategy Collaborative Spectral Correction (MSCSC) method. This method took 50 mJ as the reference energy, with the laser energy varying in the range of 30-70 mJ. First, a sliding window-based peak extraction strategy was adopted to identify the internal standard line and characteristic lines. Second, for the internal standard line, a plasma temperature-based spectral correction strategy was applied to correct it to the reference energy level. Finally, the characteristic spectral lines were classified into two categories by a Pearson correlation coefficient threshold: those exhibiting an intensity variation trend consistent with that of the internal standard line, and those exhibiting an inconsistent trend. The internal standard regression method and the transfer learning method were then respectively used to achieve spectral intensity correction. Experimental results show that the relative RSD of the internal standard line decreases by around 64.4%, and characteristic line RSDs decrease by over 50%.

The PLSR model based on corrected spectra achieves R2 P above 0.93 for Cr, Cu, Ni, Na, and K. RMSE_P is reduced by more than 60% and MRE_P is controlled within 0.17, thereby improving the quantitative analysis performance for sample elements under varying laser energies.