Machine learning-based calibration LIBS analysis of aluminium-based alloys

Poor accuracy in the compositional results using LIBS analysis is still a challenging task; different statistical methods have been employed to improve the accuracy of the quantitative investigations. In the present work, multi-spectral line calibration Artificial Neural network (MSLC-ANN), a new technique, has been utilized to improve the accuracy of the quantitative analysis. Eight samples with different known concentrations of Cu, Zn, Sn, Pb, Si were prepared where aluminium was used as a base metal. The observed line intensities were normalized with the base aluminium metal to reduce uncertainties in the line intensities measurements. The intensity ratios of the target and matrix elements were used to train ANN. The concentrations of the target samples were predicted using a trained Neural Network. To compare the results, a calibration curve (CC-LIBS) analysis was also performed. Compositions obtained using MSLC-ANN showed an excellent agreement with the certified compositions as compared to the compositions obtained using CC-LIBS. Thus, the present study demonstrated that optimized ANN is an excellent tool for the compositional analysis of alloys, soil, and other multi-elemental samples and it can also be used for quality control in alloys and mining industries.