Hungary's radar stations need to be integrated according to the already completed and by our country also accepted Single European Sky project. At the same time, similarly to the neighbouring countries, we have to switch over to the remote operating of the stations.
The neural networks gather more and more ground in our dynamically developing world. Due to their increased generalising ability and being very robust they can be deployed to solve problems more efficiently in certain situations.
During the practical work I studied two possible applications of the neural networks to support the remote operation of the radars.
The first opportunity is supporting the direct error-correcting. This does the identifying of the error signals gathered during observing the technical units of the station. Using such a system the remote operation can be made safer and more efficient in case sufficient data are available.
The other opportunity is estimating the failures of the station units on the basis of Bayes network models. In case we can use real and accurate data, the network supports the estimation of the independent (marginal) failure-probability of the technical units.
I generated the data matrix pairs to use to train the two networks on the basis of different codes, since much more samples were necessary than available. The used codes can also be found in my thesis.