Application of Computational Intelligence Methods for the Automated Identification of Paper-Ink Samples Based on LIBS

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Computational Intelligence

Background:

• Laser-Induced Breakdown Spectroscopy (LIBS) is a valuable analytical technique across various scientific and industrial fields.
• A major limitation of LIBS is the time-consuming spectral interpretation and line identification, often requiring manual comparison with databases.
• This bottleneck hinders the efficient application of LIBS in real-world scenarios.

Purpose of the Study:

• To develop a fast and reliable classification system for quasi-destructively acquired LIBS spectra using computational intelligence.
• To specifically address the classification of paper-ink samples into 30 distinct classes.

Main Methods:

• Utilized Laser-Induced Breakdown Spectroscopy (LIBS) to collect spectra from 30 classes of paper-ink samples.
• Employed four preprocessing variants and seven distinct machine learning classifiers, including Random Forest, Support Vector Machine, and Neural Networks.
• Implemented 5-fold stratified cross-validation and tested on an independent dataset for robust evaluation.

Main Results:

• Achieved a high classification accuracy of 99.08% for paper-ink samples.
• The Random Forest classifier demonstrated superior performance in accurately categorizing the LIBS spectra.
• Successfully classified samples into 30 predefined classes, encompassing various ink and paper types.

Conclusions:

• Machine learning methods offer a reliable and accelerated solution for identifying paper-ink samples using LIBS.
• Computational intelligence effectively overcomes the limitations of traditional spectral analysis in LIBS.
• The developed system shows significant potential for enhancing the speed and efficiency of LIBS applications.