Multivariate Calibration of the Composition of Low-Alloy Steels Using Pre-Processed Low-Resolution Emission Spectra with Spectral Variables Selection

Abstract

Calibration of concentrations of C, Mn, Si, Cr, Ni, and Cu by the low-resolution laser induced breakdown spectroscopy is made in reference to low-alloy steels etalons. Data preprocessing in the form of spectrum normalization at Fe II 252.0609 nm emission line wavelength and baseline correction, as well as spectral variables selection with an original method of searching combination moving window for the partial least squares method made it possible to build multivariate calibration models for all considered elements with the following characteristics: for C (in the concentration range up to 0.7%) the value of root-meansquare error and residual predictive deviation is 0.04 % and 4.7, Mn (up to 1.9%) – 0.02 % and 24.8, Si (up to 0.9%) – 0.01 % and 12.9, Cr (up to 1%) – 0.01 % and 21.8, Ni (up to 0.7%)– 0.007 % and 23.3, Cu (up to 0.5%) – 0.006 % and 23.2, respectively. Models are quantitative (residual predictive deviation > 3) for all six elements considered, including carbon.