Transcript OUTLINE

PREDICTING INTERPLANETARY
SHOCKS USING NEURAL NETWORKS
A.İbrahim PALA
METU Aerospace Engineering Department
Senior Student
21.12.2005
OUTLINE
WHAT IS AN INTERPLANETARY SHOCK?
WHAT IS A NEURAL NETWORK?
HOW THE NEURAL NETWORKS ARE USED
TO PREDICT THE INTERPLANETARY
SHOCKS?
CONCLUSION
INTERPLANETARY SHOCKS
 Charged particles are
constantly streaming
outward from the Sun.
 When events such as
Coronal Mass Ejections
(CMEs) occur on the Sun,
an Interplanetary (IP) shock
may be generated.
INTERPLANETARY SHOCKS
The abrupt boundary
formed at the front of
a plasma cloud
moving much faster
than the rest of the
solar wind, as it
pushes its way
through
interplanetary space.
NEURAL NETWORKS
 NEURONS
SYNAPSE
(INPUT->WEIGHTS)
COLLECTOR
ACTIVATION
FUNCTION->OUTPUT
NEURAL NETWORKS
 CONSTRUCTING A NEURAL NETWORK BY BRINGING
LOTS OF NEURONS TOGETHER AND
INTERCONNECTING AND TRAINING THEM
 SYSTEM PARALLELISM AND STRUCTURAL
DISTRIBUTION OF OVERALL FUNCTION
->compensating the errors due to a nonfunctional
neuron
 GENERALIZATION AND LEARNING ABILITY
->giving meaningful outputs for also the nontrained inputs
PREDICTING THE ARRIVAL OF
IP SHOCKS BY NEURAL NETWORKS
IP Shocks causes the incoming particle
flux observed at or near Earth to increase
by orders of magnitude.
 Sharp increases in particle intensity are
called Energetic Storm Particle (ESP)
events
PREDICTING THE ARRIVAL OF
IP SHOCKS BY NEURAL NETWORKS
INPUTS: Observations from
two ACE instruments:
 The Electron, Proton, and
Alpha Monitor (EPAM): particle
intensity at five different
energy levels. This data is used
to detect the onset of an
event, several days in advance,
and to predict the ESP
event / IP shock arrival.
PREDICTING THE ARRIVAL OF
IP SHOCKS BY NEURAL NETWORKS
INPUTS
 The Solar Wind Electron,
Proton, and Alpha
Monitor (SWEPAM): solar
wind plasma
temperature, density, and
speed. This data is used
to detect the actual IP
shock arrival.
PREDICTING THE ARRIVAL OF
IP SHOCKS BY NEURAL NETWORKS
(Real-time Interplanetary Shock Prediction
(RISP) System )
TRAINING THE ARTIFICIAL NEURAL NETWORK
TO PREDICT THE TIME REMAINING UNTIL THE
SHOCK ARRIVES
TRAINED BY 37 OF THE 56 SELECTED SHOCKS
PERFORMANCE ASSESSED BY FORECASTING
ARRIVAL TIME COUNTDOWNS FOR 19 SHOCKS
NOT USED DURING TRAINING
PREDICTING THE ARRIVAL OF
IP SHOCKS BY NEURAL NETWORKS
Particle intensities for test sequence 9.
Predicted time until shock arrival.
REFERENCES
 www.physics.usyd.edu.au/~cairns/teaching/lecture11/node5.html+i
nterplanetary+shocks&hl=tr
 http://www-spof.gsfc.nasa.gov/Education/wgloss.html
 http://sd-www.jhuapl.edu/ACE/EPAM/RISP/
 Vandegriff , J., Wagstaff, K., Ho , G., Plauger, J. Forecasting space
weather: Predicting interplanetary shocks using neural networks.
Adv.Space Res. 36, 2323-2327, 2005.
 www.srl.caltech.edu/ACE/
 Efe, M.Ö., Kaynak, O., Yapay sinir ağları ve uygulamaları, Bogazici
University,2005