Enhanced Accuracy for the Determination of Chromium in Steel by Laser-Induced Breakdown Spectroscopy (LIBS) and Extreme Random Tree (ERT)

Abstract

The content of chromium (Cr) affects the mechanical and wear resistance of steel. Hence, the determination of Cr in steel is of great significance to determine the quality. In this study, a new method based on laser-induced breakdown spectroscopy (LIBS) combined with extreme random tree (ERT) is reported to determine Cr in steel. First, the Cr I 425.43 nm line was selected for univariate quantitative analysis. Next, multivariate analysis of Cr was performed using partial least squares regression (PLSR), random forest (RF), and ERT. The abilities of the models for the determination were validated by the fitting coefficient (R 2), root mean square error of calibration (RMSEC), root mean square error of prediction (RMSEP), and average relative error (ARE). The R 2, RMSEC, RMSEP, and ARE values of ERT were 0.9984, 0.0274 wt.%, 0.0299 wt.%, and 2.43%, respectively. Compared with the univariate model, PLSR, and RF models, the R 2 of ERT increased by 0.1317, 0.0306, and 0.013, respectively; RMSEC decreased by 0.0987, 0.0955, and 0.0174 wt.%; RMSEP decreased by 0.1463, 0.0876, and 0.002 wt.%; and ARE decreased by 17.88%, 16.98%, and 5.95%, respectively. The results show that combining LIBS with the ERT model for multivariate analysis improves the elemental analysis of steel.