Automatic Scaling Of F Layer For Ionogram Based On BP Neural Network

Ionospheric physics is a science based on experimental observation; the development and use of experimental observation equipment play an important role in ionospheric observation and study. The ionospheric digisonde is commonly used on ionospheric sounding which works by transmitting radio under sweep frequency mode. The ionospheric digisonde has worked f_or a long time, uninterrupted with ionospheric sounding, and it has accumulated a large amount of ionospheric data-ionogram. The ionospheric characteristic parameters and electron density profiles which used in ionospheric monitoring and prediction can get from the scaling and inversion of ionogram. The traditional manual mode of ionogram scaling is a heavy workload and it's a time-consuming eff_ort, so it's imperative to develop automatic scaling method of ionogram based on computer software. A new method f_or automatic scaling the F layer parameters of ionograms based on BP?Back Propagation? neural network is presented in this paper. This method is applied to the automatic scaling of ionograms collected with the newly developed portable digital ionosonde?PDI? by our lab and some preliminary results are obtained.Firstly, this thesis systematically introduced the basic knowledge of ionosphere and ionospheric sounding, the principle of radio detecting, the equipment of ionospheric observation and observational mode of ionosphere, mainly introduced ionospheric digisonde which is a conventional equipment of ionospheric sounding, explained contents and meaning of ionogram generated by ionospheric digisonde, and illustrated the significance of automatic scaling to the study of ionosphere. Besides, this paper summarizes several automatic scaling methods existed in China and abroad, including ARTIST, trend extrapolation, geometrical fuzzy, profile fitting, EOF and so on. Every method has its advantage, applications and limitations, so we propose a new automatic scaling method based on BP neural network.Secondly, introduced BP neural network and adaptive curve fitting technology in detail, and applied this technology to the new method f_or ionogram F layer scaling. This method firstly use local threshold which is equal to noise average power to remove noisy signal at different frequencies, and remove multiple reflection and leakage according to the wave features and amplitude of ionogram. Use A45 method to turn ionogram's sounding data into point cells and complete the pre-process of complicated BP neural network fitting. Complete the normalization of point cells' coordinate and use BP neural network to fit normalized value, get the F layer traces of O and X wave through the trained BP neural network. The cut-off points between F₁ and F₂ layer are f_ound out according to the variation of the slope of traces and read h'F₁, f_oF₁ and h'F₂ parameters. The curve fitting is used to fill the missing parts of F₂ cusp ultimately to complete reading the f_oF₂ parameters, and ultimately achieve the automatic scaling of ionogram F layer.Finally, applied this new method to the automatic scaling of ionograms collected with the newly developed portable digital ionosonde by our lab and some preliminary results are obtained. We counted acceptable percentage of F₂ layer parameters from automatic scaling of ionograms and analyze the reasons of few failed automatic scaling. The accuracies of the scaled f_oF₂ and h'F₂ are 81 percent and 78 percent respectively and this method is practical. We also counted the change of acceptable percentage of F₂ layer parameters from automatic scaling of ionograms by the time, and analyze the reasons of changes. Besides, we counted the parameter f_oF₂ of automatic scaling of ionograms generated in quite and magnetic storm periods respectively, and compare with f_oF₂ provided by IRI?International Reference Ionosphere? and analyzed result is given.