Study on LIBS Standard Method via Key Parameter Monitoring and Backpropagation Neural Network

Abstract

This paper proposes a method based on key parameter monitoring and a backpropagation neural network to standardize LIBS spectra, named KPBP. By monitoring the laser output energy and the plasma flame morphology and using the backpropagation neural network algorithm to fit the spectral intensity, KPBP standardizes spectral segments containing characteristic lines. This study first conducted KPBP experiments on the spectra of pure aluminium, monocrystalline silicon, and pure zinc to optimize the KPBP model and then performed KPBP standardization on the characteristic spectral lines of a GSS-8 standard soil sample. The spectral intensity relative standard deviations (RSDs) of Al 257.51 nm, Si 298.76 nm, and Fe 406.33 nm dropped from 12.57%, 16.60%, and 14.10% to 3.40%, 3.20%, and 4.07%, respectively. Compared with the internal standard method and the standard normal variate method, KPBP obtained the smallest RSD. The study also used a GSS-23 standard soil sample and a Beijing farmland soil sample to conduct KPBP optimization experiments. The RSD of spectral intensity was still significantly reduced, proving that the KPBP method has stable effects and wide applicability to improve the repeatability of LIBS soil analysis.

Keywords

• laser-induced breakdown spectroscopy (LIBS)
• spectral repeatability
• key parameter monitoring
• backpropagation neural network
• quantitative analysis of soil elements