Publication detail

Spectra transfer between distinct LIBS systems using shared standards and machine learning

VRÁBEL, J. KÉPEŠ, E. NEDĚLNÍK, P. POŘÍZKA, P. KAISER, J.

Original Title

Spectra transfer between distinct LIBS systems using shared standards and machine learning

Type

abstract

Language

English

Original Abstract

Mutual incompatibility of distinct spectroscopic systems is among the most limiting factors in Laser-Induced Breakdown Spectroscopy (LIBS). This problem is ruling out the possibility of shared libraries of standards and trustworthy inter-laboratory comparison. However, the general solution to this problem is almost impossible due to the change of physical conditions during experiments and widely varying analytical performances of spectrometers. We demonstrate the possibility of spectra transfer for a special case, where both systems measure simultaneously from the same plasma. Extensive datasets measured as hyperspectral maps of heterogeneous specimens are used for the training of machine learning (ML) models that are able to transfer spectra between systems. We use a latent representation (obtained from an autoencoder) of the data measured on the master system, where data from the subordinate system are mapped by a fully-connected artificial neural network (ANN) to corresponding locations (see Fig. 1).

Keywords

transfer library; LIBS; machine learning; transfer learning; spectroscopic data

Authors

VRÁBEL, J.; KÉPEŠ, E.; NEDĚLNÍK, P.; POŘÍZKA, P.; KAISER, J.

Released

5. 9. 2022

URL

BibTex

@misc{BUT180062,
  author="Jakub {Vrábel} and Erik {Képeš} and Pavel {Nedělník} and Pavel {Pořízka} and Jozef {Kaiser}",
  title="Spectra transfer between distinct LIBS systems using shared standards and machine learning",
  year="2022",
  url="https://libs2022.com/wp-content/uploads/2022/09/BookAbstracts1-9-22_pagenumber.pdf",
  note="abstract"
}