Accurate prediction analysis of steel alloy elements by femtosecond laser-ablation spark-induced breakdown spectroscopy and out-of-bag random forest regression

Abstract

Femtosecond laser-ablation spark-induced breakdown spectroscopy (fs-LA-SIBS) and out-of-bag random forest regression (OOB-RFR) were developed for accurate quantitative analysis of the elements manganese (Mn), chromium (Cr), and nickel (Ni) in steel alloys. A femtosecond laser operating at 1 kHZ was used as the laser ablation source and a spark discharge unit was utilized to enhance the peak intensity of atomic emission. The characteristic lines of Mn, Cr, and Ni elements in steel alloys were determined using the NIST database. The two hyperparameters n_tree and m_try of RFR were optimized through OOB error, and the OOB-RFR model was obtained. In the experiment, the predictive performance of OOB-RFR, support vector machine (SVM) regression, and partial least squares regression (PLS) on steel alloy spectral data was compared and analyzed. The results indicate that the OOB-RFR model exhibits lower computational iterations and higher performance than other models in regression prediction of Mn, Cr, and Ni elements. In the OOB-RFR model, the root mean square error (RMSE), the average relative error (MRE) and the coefficient of determination (R²) of Mn, Cr, and Ni elements in the training samples were 0.0051, 0.0018 and 0.9976, respectively, and the RMSE, MRE and R² of the three elements in testing samples were 0.0018, 0.0009 and 0.9369, respectively. Therefore, the combination of the fs-LA-SIBS technique and OOB-RFR has great potential in the rapid determination of elements in steel and even in the metallurgical field.