Publication: Evaluation of artificial neural network method applied to probe diagnostics in plasma
Loading...
Identifiers
Publication date
2018-07
Defense date
2018-07-12
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Plasma diagnostics consist on predicting the plasma parameter by measuring their properties.
Models to predict these parameters are based on theoretical equations, which sometimes are
complicated and tedious to solve. This thesis aims to evaluate a new method to obtain these
plasma parameters using a branch of artificial intelligence, the Artificial Neural Network.
To develop this thesis, the software MATLAB was used to create the neural network. Two theories
were tested: Planar Langmuir Probe and Orbital Motion Limited. And two problems were solved:
forward problem to obtain the properties from the parameters and the inverse problem to obtain
the parameters from the properties. The later approach is used to evaluate the parameters obtained
by network with experimental data from laboratory measurements.
Artificial Neural Network was implemented successfully in both forward and inverse problems
with incredible accuracy. The results obtained have created a theoretical framework for plasma
diagnostic using neural networks and it has been concluded that they can work with any theory
presented, as long as as proper database is obtained. Regarding the simulation with experimental
data, the results were not good respect the measurements taken. Although the network performed
well, the database used for training and simulation had discrepancies from realistic data since it was
obtained from equations. Sources of error of this last problem are found and different approaches
and work to be developed in the future are proposed.
Description
Keywords
Plasma diagnostics, Artificial Intelligence, Artificial neural networks, Matlab, Planar Langmuir Probe, Orbital Motion Limited