Using various machine learning algorithms for quantitative analysis in Laser induced breakdown spectroscopy

TL;DR

This paper evaluates various classical and nonlinear machine learning algorithms, including PCA integration, for quantitative analysis of aluminum samples using laser induced breakdown spectroscopy, identifying the most effective model.

Contribution

It introduces a comprehensive comparison of multiple machine learning approaches, including PCA integration, for LIBS-based quantitative analysis of aluminum alloys.

Findings

PCA combined with KSVR achieved the best prediction accuracy.

Nonlinear algorithms outperformed linear models in element concentration prediction.

Dimension reduction via PCA improved the performance of machine learning models.

Abstract

Laser induced breakdown spectroscopy technique is employed for quantitative analysis of aluminum samples by different classical machine learning approaches. A Q-switch Nd:YAG laser at fundamental harmonic of 1064 nm is utilized for creation of LIBS plasma for prediction of constituent concentrations of the aluminum standard alloys. In current research, concentration prediction is performed by linear approaches of support vector regression, multiple linear regression, principal component analysis integrated with MLR and SVR, and as well as nonlinear algorithms of artificial neural network, kernelized support vector regression , and the integration of traditional principal component analysis with KSVR, and ANN. Furthermore, dimension reduction is applied on various methodologies by PCA algorithm for improving the quantitative analysis. The results presented that the combination of PCA with KSVR algorithm model had the best efficiency in predictions of the most of elements among other classical machine learning algorithms.