Study of mineralization in geological samples by means of LIBS and neural networks

Varování

Publikace nespadá pod Ekonomicko-správní fakultu, ale pod Přírodovědeckou fakultu. Oficiální stránka publikace je na webu muni.cz.
Autoři

KLUS Jakub POŘÍZKA Pavel PROCHAZKA David MIKYSEK Petr NOVOTNÝ Jan NOVOTNÝ Karel KAISER Jozef

Rok publikování 2016
Druh Konferenční abstrakty
Fakulta / Pracoviště MU

Přírodovědecká fakulta

Citace
Popis This work aims on the description of possible element association within a sample of sandstone-hosted uranium ore by means of Laser-Induced Breakdown Spectroscopy (LIBS). As an element association in the interaction region and in terms of LIBS we refer to the simultaneous presence of spectral lines within a respective single spectrum. Presented results show element associations within a sandstone ore sample carrying high abundance of zirconium, uranium, niobium and hafnium. To manage this task a multivariate method was utilized, namely the self-organized maps (SOM). SOM is a type of artificial neural network, which provides dimensionality reduction based on the similarity of input data. Responses of SOM weights associated with certain elemental lines were easily discriminated as either simultaneous or isolated. Deduced association of U-Zr and isolation of Ti, Fe and Si responses is in good correlation with geological studies made on ores from the same place of origin. Finally a mineralization visualization performed in unconventional manners is shown. Instead of the creation of chemical map by rearranging the line intensities to the rectangular grid of points, a correlation of spectra with neuron responses was calculated. It is shown that map of correlation coefficients provides better insight into the problem of element associations or possible co-mineralization in general.
Související projekty:

Používáte starou verzi internetového prohlížeče. Doporučujeme aktualizovat Váš prohlížeč na nejnovější verzi.