Application Research Of Deep Learning In RF Circuit Fingerprint Identification And Reliability Analysis Under Irradiation Effect

With the development of wireless communication and space technology,more and more electronic devices or systems are sent into space to perform military or civilian space missions,but the large amount of rays and high-energy particles in the space environment will be the electronic components in which they operate.Serious threats,which in turn affect the communication reliability of the entire electronic device or system,which puts new and higher requirements on space electronic reconnaissance and aerospace monitoring technology.Radio frequency signal recognition of electronic components based on conventional parameters in traditional non-spatial radiation environment The method is no longer applicable.Therefore,this paper proposes a method for fingerprint identification of irradiated electronic components(RF circuits)in a space radiation environment based on deep learning technology.According to the influence of irradiation effects on the fingerprint characteristics of electronic components,the same type And the batch electronic components are classified and identified under different irradiation doses and annealing characteristics,and the service life of the electronic equipment or system is inferred based on the identification results,and then the reliability of the service period is estimated.The main contents and results of the thesis are as follows:1.A fingerprint identification algorithm based on residual network and irradiation effect is proposed.In order to verify the relationship between the number of network layers and the recognition performance,a different layer residual network suitable for inputting one-dimensional signals is built.The experimental results show that the recognition effect of the 18-layer network is the best.The accuracy of the identification of RF circuits with 4 different types and batches of irradiation doses is 83.92%,and 4 different annealing time nodes for RF circuits receiving 100Krad irradiation dose.The recognition accuracy rate reached 82.52%.2.Based on the results 1,continue to deeply explore the impact of RF signals with different input dimensions on recognition performance.Firstly,time-frequency transform is performed on four radio frequency signals of the same type and batch with different irradiation doses,and the one-dimensional time domain signal is transformed into two-dimensional time-frequency image,and the model parameter quantity is reduced by dimensionality reduction to improve the training efficiency..The experimental results show that the two-dimensional image contains more different features,and the accuracy of model recognition is improved by 0.45%.However,it takes time to change the time-frequency transform.In practical applications,it is difficult to meet the real-time requirements of the system.3.A radio frequency circuit fingerprint recognition algorithm based on convolutional cyclic network and irradiation effect is proposed.The algorithm integrates the residual network with LSTM,designs a network model with different downsampling strategies,and uses the characteristics of the two dimensions of fusion space and time as the basis for identification.The experimental results show that with the increase of downsampling,the recognition accuracy of the model is improved.When the residual convolution layer is set to 12,the convolution output is downsampled to 1/16 of the input,and the number of convolution kernels is increased to 32 and the LSTM unit.When the number of neurons in the hidden layer is 128,the recognition accuracy of radio frequency circuits of the same type and batch and different irradiation doses of the four types is 85.52%,and four different annealing time nodes for the RF circuit receiving the irradiation dose of 100Krad.The signal recognition accuracy rate reached 83.71%,and the recognition results prove the effectiveness of the proposed algorithm.