Research On Fault Diagnosis Methods Of Array Element-Level And Sub-Array Level Planar Arrays

Array antennas are widely used in existing equipment such as radar,sonar and navigation systems.With the increase of array size in various applications,the failure probability of array elements and sub-arrays is also increasing.Usually,large-scale arrays are partitioned by sub-arrays,and related faults often occur at the sub-array level.Therefore,the fault diagnosis problem It can be divided into two types:array-level processing and sub-array-level processing.In this paper,the fault diagnosis method of the array element level and subarray level planar array is the main body of research,focusing on the radiation far field information,considering the failure scenarios of the array element level and subarray level,combining skeletonization,deep learning and other related technologies to carry out research.The research content of this paper is as follows:1.Aiming at the sampling problem in the fault diagnosis of general element-level planar array antennas,a planar array diagnosis method based on skeletal sampling is proposed.A small number of sampling points are selected by using the skeletal method,and matrix transformation is performed at the skeletal sampling points.Calculate the pattern of the entire radiation field.This method can effectively reduce the number of sampling points,and then combine the least squares reconstruction and sparse Bayesian learning algorithms for verification.For ideal conditions,skeletonized sampling can completely replace the diagnostic results under uniform sampling.For low-noise conditions,the diagnostic error rate similar to that of the least squares reconstruction and sparse Bayesian learning algorithms can be obtained after skeletonized sampling reconstruction data.2.Aiming at the fault diagnosis problem of general element-level planar array antenna elements,considering the influence of noise environment or human operation errors,the array has position error or noise influence in actual application,which leads to the difference between the theoretical array response and the actual situation.A dual convolutional neural network joint diagnosis model based on radiation far-field information is proposed.The model divides the radiation far-field information into two parts:amplitude information and phase information.The amplitude information determines the number of failed array elements through a convolutional neural network,and the phase information determines the state of the array elements through another convolutional neural network.Combined with the two parts of the network output Jointly determine the location of the failed array element.Considering the position error of the antenna array in practical applications,an attention mechanism is added after the feature extraction layer of the model to make the model’s diagnostic performance better in the case of position error.Compared with the existing array fault diagnosis methods,the proposed method can detect the failed elements in the array antenna with higher accuracy under the condition of position error of the array antenna.3.The failures faced by the array antenna will be in the form of a subarray structure,such as the failure of the subarray caused by the failure of the network feed system,and the position error that exists at the same time is generally presented in the form of a subarray structure.Aiming at the fault diagnosis problem of large-scale array antennas in such subarray scenarios,an EfficientNet-B0 network joint diagnosis model for sub-array failure problems is proposed.The subarray fault diagnosis problem is divided into two tasks:determining the number of failed subarrays and determining the status of subarrays,and using EfficientNet-B0 network to extract the characteristics of radiation far-field data.According to the characteristics of large-scale array antenna sub-array position errors,the sub-array position is considered to be shifted as a whole.Experimental results show that the proposed method can detect failed sub-arrays in large-scale array antennas with high accuracy under the condition of sub-array position errors,and it is robust to observation noise and position errors.