Research On Nonlinear Fault Modelling And Diagnosis Method Of Satellite Receivers

With the rapid development of communication technology,satellite communication equipment is gradually evolving in the direction of complexity,precision and integration.While improving the efficiency of equipment,it also brings greater challenges to the fault diagnosis of satellite communication equipment.As an important part of satellite communication system,the safety and reliability of receiver are very important to the whole communication system.If the internal failure of the receiver is not detected and checked in time,it will not only damage the receiving equipment and affect the safe and stable operation of the system,but also lead to the performance degradation or even failure of the receiving system.Therefore,in order to improve the safety and reliability of the communication system,we should find and identify the potential abnormalities and faults in the receiver as soon as possible.Receiver faults are divided into hard faults and soft faults.Hard fault is the fault caused by device structure damage,which is mainly manifested in equipment function failure,shutdown and other faults with obvious changes;Soft fault is the fault caused by the variation of device characteristics,which is mainly manifested in the slow change faults such as equipment function degradation and unstable operation.Compared with the obvious fault caused by hard fault,the soft fault changes slowly and is not easy to detect,which has a greater impact on the system,and the soft fault in the receiver is mainly the nonlinear fault generated by each device.Therefore,according to the standard processing flow of receiver nonlinear fault diagnosis,this paper focuses on the nonlinear fault modeling of satellite receiver,the fault diagnosis method based on machine learning The receiver fault diagnosis method based on deep learning is divided into three parts.The main contributions and innovations are as follows:1.Aiming at the nonlinear fault in the satellite receiver,the nonlinear fault signal model of the RF front-end of the receiver is established.Firstly,aiming at the internal reflection interference of the filter,the nonlinear fault model characterizing the internal reflection interference of the filter is established by deriving its action form on the amplitude frequency response of the filter;Secondly,since the Saleh model can not accurately describe the nonlinear fault of the amplifier at the supersaturation point,the coefficients of the Saleh model are modified by using the amplitude phase conversion relationship,and the nonlinear fault models representing the amplifier at the supersaturation point are established;Then,aiming at the influence of phase noise on oscillator frequency stability,the first-order autoregressive model is used to characterize the nonlinear fault caused by oscillator frequency offset;Finally,the three are cascaded to establish the nonlinear fault signal model of the RF front-end of the receiver,which provides a theoretical and data basis for the subsequent fault diagnosis of the receiver.The experimental results show that the model can effectively characterize the nonlinear faults in the RF front-end of satellite receiver.2.Aiming at the problem of insufficient separability of feature extraction caused by the concealment and coupling of nonlinear faults in the receiver,a fault diagnosis method based on constellation error and kernel support vector machine is proposed.Firstly,according to the generation mechanism of nonlinear fault of receiver,the visual performance of nonlinear fault of filter,amplifier and oscillator in receiver on constellation diagram is analyzed;Then,the nonlinear faults of each device are decomposed according to the characteristics of the constellation diagram,and the dispersion,average density,average amplitude,average phase angle deviation,average angle difference and average arc length of the constellation cluster are extracted to form the feature vector;Finally,the feature vector is input into kernel support vector machine to realize fault classification and recognition.The experimental results show that the proposed method can jointly extract a variety of nonlinear fault features,such as filter amplitude frequency response distortion,amplifier amplitude compression and phase conversion,oscillator frequency offset and so on.Compared with the existing methods,the accuracy of diagnosis is significantly improved.3.Aiming at the problem that the traditional methods rely on the existing cognitive level and processing ability,and only use the signal single-mode information for fault diagnosis,a fault diagnosis method based on multi-mode information convolution neural network is proposed.In the first mock exam,the mode of the fault signal is converted to the first mode,the instantaneous amplitude and phase are second modes,the spectrum amplitude and the square spectrum are third modes as the information in time domain and frequency domain.Secondly,three one-dimensional convolutional neural networks are constructed to extract the features of the input multimodal data;Then the multi-modal features of the signal are fused by feature fusion strategy to obtain more separable and robust features;Finally,the classification and identification of receiver nonlinear faults are realized.Experimental results show that this method can overcome the limitations of existing cognition,effectively extract the fault features of complex data,and significantly improve the accuracy of diagnosis.