Spectral preprocessing to improve accuracy of quantitative detection of elemental Cr in austenitic stainless steel by laser-induced breakdown spectroscopy

Abstract

The production and consumption of austenitic stainless steel account for about 70% of stainless steel worldwide. The content of chromium (Cr) must be accurately detected and controlled to form a stable austenite structure and obtain strong properties in production. Laser-induced breakdown spectroscopy (LIBS) can be used to detect the Cr content of austenitic stainless steel in a complex production process. However, LIBS signals may be weak and unstable because the experimental signals are seriously affected by noise, self-absorption, the matrix effect, and the instability of the shot-to-shot signal, rendering the quantitative detection results inaccurate and unstable. The spectral-preprocessing methods of baseline correction and denoising can improve the accuracy of quantitative detection of LIBS. An improved segmented Hermite cubic-interpolation method is proposed herein to correct the baseline offset and produce baseline signals that are smooth and convergent (to overcome the Runge phenomenon). Empirical mode decomposition (EMD) based on the wavelet method is proposed to remove LIBS noise; this is done by exploiting the adaptivity of EMD to refine the wavelet-scaling coefficients. Compared with other denoising methods, the proposed method has good denoising evaluation indices and stability and, thus, effectively removes the noise. To verify detection accuracy, the internal standard quantitative method is used to detect the Cr content, and a cyclic-inversion prediction method is designed to verify detection stability. The results show that the correlation coefficient of the calibration curve is improved, the root-mean-square error is reduced, and the average relative error of the predicted Cr content decreases from 10.46% to 3.858%.