Laser-Induced Breakdown Spectroscopy Combined with Nonlinear Manifold Learning for Improvement Aluminum Alloy...

Area of Science:

• Analytical Chemistry
• Materials Science
• Spectroscopy

Background:

• Laser-Induced Breakdown Spectroscopy (LIBS) generates complex spectral data.
• High dimensionality in LIBS datasets presents challenges for accurate analysis and classification.
• Efficient dimensionality reduction is crucial for practical LIBS applications.

Purpose of the Study:

• To classify five aluminum alloy types noninvasively using LIBS.
• To evaluate manifold dimensionality reduction techniques for LIBS data.
• To integrate reduced-dimension data with a Support Vector Machine (SVM) classifier.

Main Methods:

• Utilized nonlinear manifold learning techniques: Isomap, Local Tangent Space Alignment (LTSA), Local Linear Embedding (LLE), and Laplacian Eigenmaps (LE).
• Compared nonlinear methods with linear techniques like Principal Component Analysis (PCA) and Multidimensional Scaling (MDS).
• Applied augmented partial residual plots to assess spectral data nonlinearity.
• Fed reduced dimensions into an SVM classifier for alloy classification.

Main Results:

• The Isomap-SVM model achieved the highest classification accuracy of 96.67%.
• Optimal classification was obtained with 11 dimensions and 18 nearest neighbors using Isomap.
• Nonlinear manifold learning techniques outperformed linear methods in this LIBS dataset.

Conclusions:

• Nonlinear manifold learning combined with multivariate analysis offers a robust approach for LIBS data classification.
• This method effectively reduces LIBS data dimensionality while preserving crucial information for alloy identification.
• The Isomap-SVM model demonstrates significant potential for rapid and accurate material analysis using LIBS.