Design On Detection Method Of Hail Based On Fuzzy Logic With Adjustable Parameters

Hail brings huge economic losses to agriculture and transportation every year.It is particularly important to predict the dangerous weather information of hail in time.The dualpolarization weather radar can alternately emit horizontal and vertical polarized waves and obtain a variety of polarization parameters,which provides strong support for hail detection research.Therefore,it is very important to detect hail based on dual-polarization weather radar.Hail detection methods mostly focus on the analysis of hail features,and the precipitation particle classification algorithm only uses fuzzy logic algorithm to detect hail.Fuzzy logic uses expert experience values for effective classification,and can handle fuzzy information,concepts,and so on.Since its development,fuzzy logic algorithms have been widely used in the research of hail detection.The current fuzzy logic algorithms still have shortcomings: the performance of fuzzy logic algorithms depends on the membership function parameters,and the membership function parameters mainly rely on expert experience values,and expert experience values are not universal and may be affected by human factors.Therefore,it is very important to find a suitable method to adjust the membership function parameters.The main research contents of this thesis are summarized as follows:First,the polarization parameters of the dual-polarization weather radar and the characteristics of the polarization parameters of the precipitation particles are studied.Second,because traditional fuzzy logic algorithms mainly rely on expert experience values to determine the membership function parameters,and expert experience values are not universal,a fuzzy logic hail detection method based on two-dimensional statistical analysis is proposed.Two-dimensional statistical analysis is mainly used to obtain the bell-shaped membership function parameters and weight coefficients.Finally,the adjusted parameters are combined with the melting layer information and outlier information to classify the precipitation particles of the echo data,and the hail detection results are obtained.This method can adjust the membership function parameters more accurately without using expert experience values to realize hail detection.Third,although the fuzzy logic method of two-dimensional statistical analysis does not rely on expert experience,it may be affected by human factors.Therefore,an improved fuzzy neural network dual-polarization weather radar hail detection method is proposed,which can adaptively adjust the parameters of the bell-shaped membership function.It is mainly to determine the initial value of the bell-shaped membership function parameters before training the data,use the K-Means++ clustering algorithm to cluster the data,and perform statistical analysis on the clustered data to obtain the initial value of the bell-shaped membership function;then use fuzzy neural The network adjusts the parameters of the bell-shaped membership function,and finally uses the echo data to classify the precipitation particles,so as to achieve the purpose of hail detection.Experiments show that after setting the initial parameters,the fuzzy neural network algorithm can basically realize hail detection.