Pc5 Detection in Geosynchronous Particle Data Using Artificial Neural Network
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Geomagnetic pulsations are magnetohydrodynamic (MHD) waves in Earth‘s magnetosphere that can be seen as variations in the magnetic and electric fields, but also in particle data. Artificial Neural Network’s (ANN’s) ability to recognize patterns have been used to find Pc5 pulsations in geosynchronous electron flux data from LANL satellites. Building on previous results  presenting the Fourier transform to the ANN, additional inputs were used to separate false events from real pulsations. The ANN was taught to generate +1 for a pulsation and 0 for a non-pulsation. The procedure of letting the network search for pulsations that later would be used in training caused it to only detect pulsations within the frequency range of the first training set. To get around this, fake training data were produced to make the network detect pulsations with other frequencies. The ANN’s ability to separate false events form real pulsations was evaluated using one full year of data that it had not been introduced to before.
The three lowest energy channels 50-75 keV, 75-105 keV and 105-150 keV were scanned to make sure a pulsation pattern was detected simultaneously in all channels. A cut-off level was imposed to separate false events from real pulsations. There were several events that the ANN did not detect which is believed to be result of either insufficient training at these frequencies, or it might be caused by the extra inputs giving it the impression of a false event. The ANN were later used to scan all available data from all LANL-satellites from 1989 to 2007. The results indicates that Pc5 pulsations in the electron flux data mainly occurs in the afternoon sector around 14:00-15:00 hours local time. It also shows a dominant frequency around 3 mHz among these events as a possible fundamental mode oscillation. An increase of events around 4-4.5 mHz and 6-6.5 mHz as well.
Place, publisher, year, edition, pages
2011. , 54 p.
EES Examensarbete / Master Thesis, XR-EE-SPP 2011:004
Engineering and Technology
IdentifiersURN: urn:nbn:se:kth:diva-91473OAI: oai:DiVA.org:kth-91473DiVA: diva2:510396
Master of Science in Engineering - Electrical Engineering
UppsokPhysics, Chemistry, Mathematics
Blomberg, Lars, ProfessorCollier, Andrew
Blomberg, Lars, Professor