Research On Fault Diagnosis Of Antenna Arrays Based On Bayesian Learning

With the rapid development of science and technology,antenna has become an indispensable part in communication,military and other fields.The increasingly large-scale array antenna often weakens its performance in all aspects due to the existence of failure elements.Therefore,it is urgent to detect the failure elements in the array quickly and accurately,so as to avoid the deterioration of array performance.Array antenna fault diagnosis is a method to reversely restore the array excitation weight through the far-field detection data,so as to further judge the information of failure elements in the array.Because many classical algorithms have the problem of high sampling,the ill conditioned problem caused by reducing the sampling rate has become our key research content.Firstly,we will solve the ill conditioned problem of reverse reconstruction through the theory of Compressed Sensing(CS)in this paper;Secondly,Sparse Bayesian Learning(SBL)architecture is introduced to break through the limitations of traditional compressed sensing;Then,under the SBL framework,the prior information of the array antenna with failure elements is further mined,and the corresponding prior model is constructed.Finally,the diagnosis algorithm with better diagnosis quality and faster calculation speed is obtained.The specific research contents of this paper are as follows:1.An array antenna diagnosis method based on CS theory is proposed.Firstly,the physical characteristics of far-field radiation of array antenna are studied.Using the sparsity of failure elements in the array,a linear model with sparse unknowns is constructed by differential processing;Based on the linear model,the sparse signal is restored by using the traditional CS method,specifically:orthogonal matching pursuit(OMP)algorithm and Bayesian compressed sensing(BCS);Finally,a series of simulation diagnosis cases are designed to verify the applicability of the proposed method and explore its advantages and disadvantages.2.An array antenna diagnosis method based on Bernoulli Gaussian(BG)prior and generalized approximate message passing(GAMP)Bayesian learning is proposed.Firstly,the BG prior is constructed by using the binary characteristics of the excitation when the failure elements is completely damaged in the array antenna;Then,the expectation maximization(EM)algorithm is used for Bayesian inference,and GAMP is embedded into EM algorithm to decouple the likelihood function,so as to further reduce the computational complexity;Finally,a series of array diagnosis simulation experiments are carried out by using the proposed method,and compared with the classical CS algorithm to verify the superior performance of the proposed method,such as applicability,high diagnosis quality and high robustness.3.An array antenna diagnosis method based on Pattern-Coupled(PC)hierarchical Gaussian priori and GAMP Bayesian learning is proposed.Firstly,a soft coupling mechanism is designed by using the spatial statistical correlation and block-sparse of failure elements in the array,and a hierarchical Gaussian prior model based on the soft coupling mechanism is constructed;Then,GAMP algorithm is used to obtain the approximate likelihood probability,and EM algorithm is used for Bayesian inference;Finally,the proposed method is simulated to verify its feasibility,and compared with other classical algorithms.The results show that the algorithm greatly improves the diagnosis accuracy and efficiency.4.An array antenna diagnosis method based on cluster structured sparsity(Cluss)and variational Bayesian learning(VBL)is proposed.Firstly,in order to solve the high computational complexity caused by enhancing the sparsity of the vector to be measured,the vector to be measured is divided into two items,and Gaussian prior and cluster sparse prior are constructed respectively;Then Bayesian inference is carried out by using variable Bayesian learning method;Finally,the proposed method is simulated in different diagnosis scenarios to verify its feasibility,universality and robustness.