Research on LIBS spectral screening method based on perturbation variables

The prediction accuracy of LIBS in the quantitative analysis of complex alloy systems is often severely limited by the serious interference of spectral line overlap. To solve this problem, a new quantitative analysis method combining dynamic response strength (DRS) screening and two-dimensional correlation spectroscopy (2DCOS) feature extraction is proposed. In this method, the laser energy gradient is introduced as an external disturbance, and the overlapping signal is decoupled by using the dynamic difference of the energy response of different elements and energy level transitions. First, DRS analysis can quickly screen out highly active spectral lines sensitive to energy disturbance and effectively eliminate background noise. Then, the 2DCOS selectivity coefficient is introduced to quantify the dynamic coupling degree between spectral lines, and further eliminate the spectral lines with serious overlapping interference. Finally, the autocorrelation peak intensity (Iauto) in the synchronous two-dimensional correlation spectrum was extracted as the quantitative feature after feature extraction, and the prediction model was established by combining support vector regression (SVR). The results show that compared with the traditional full-spectrum partial least squares (PLS) method, the proposed strategy significantly improved the quantitative accuracy. The predictive determination coefficients (R2) of Cr and Ni elements reached 0.9985 and 0.9967, respectively. The root mean square error (RMSE) decreased to 0.1223 and 0.2099, respectively. The results show that the feature extraction method based on dynamic information can effectively eliminate spectral interference, which provides a new way for high-precision LIBS online analysis.