Laser induced breakdown spectroscopy (LIBS) combined with machine learning for the classification of high voltage composite insulators
Loading...
Date
2024-10-15
Authors
Καντεμίρης, Γεώργιος
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In the present master thesis, Laser Induced Breakdown Spectroscopy (LIBS) assisted with machine learning algorithms was employed for classifications on composite insulators’ housing material. The classifications were specifically based on the type, pollution level, and degradation level due to fire. The available insulator samples originated from Crete’s and Rhodes’ power transmission networks and were 33 in total.
LIBS is an analytical spectroscopic technique that involves the interaction of a focused, high-power laser beam with a sample under study, providing multi-elemental analysis based on the radiation emitted by the induced plasma. The sample under study can be in any state of matter (i.e., solid, liquid, and gas), while LIBS apparatus can provide rapid, remote, and in-situ measurements.
The three algorithms utilized were Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Random Forest (RF). The former was used to reduce the dimensionality of the spectra, while the latter two were used to create predictive models that classify LIBS spectra.
The models showed promising results for classifying insulator’s type and pollution levels, achieving accuracies up to 100%. However, models designed to classify fire degradation levels did not perform adequately.
Description
Keywords
Laser induced breakdown spectroscopy, Machine learning, High voltage insulators, Spectroscopy, Laser