Self-Learning Artificial Intelligence for Laser Induced Breakdown Spectroscopy: Data Analysis and System Control

Abstract

Laser-Induced Breakdown Spectroscopy (LIBS) is a technology that offers the possibility to determine fastly and without much preparation, the composition of a sample. LIBS extracts information from the sample by acquiring spectral lines emitted during the molecular breakdown process, allowing the detection and quantification of each present element. The analysis of the different emission lines contains information about chemical elements, isotopes and molecules, interpreting complex samples. LIBS is an emerging tool that is used in many different areas, from most industrial to the most technical ones, being also used in the healthcare sector. LIBS allows access with high precision to the samples' essence interpreting their composition in a minimally invasive way. Besides LIBS' practicability, the signal's analysis is not trivial due to its complexity and the physical phenomena present. The finite resolution of the devices does not allow the assumption of the lines' exclusivity and the quantum uncertainty exists, so the elements matching to the existing databases (e.g. NIST) is not direct. This dissertation is inserted in the research plan of INESC TEC - Center for Applied Photonics (CAP) and aims to contribute to the quantification of the elements on geological samples and CAP's developed techniques for signal processing and application of self-learning artificial intelligence. This dissertation's prime objectives comprise the contribution to the pre-processing of LIBS signal, development of artificial intelligence, and benchmarking the state-of-art techniques against methods used in cutting-edge laboratories. In a first stage, it will be addressed a resumed bibliography review on LIBS and its different quantification techniques and in a second stage, the exploration of the available laboratory devices, methods to be used and the relevant data.