Diagnosis of Array Antennas Based on Phaseless Near-Field Data Using Artificial Neural Network
Xin Wang, Keisuke Konno, Qiang Chen
Abstract
Diagnosis of array antennas based on phaseless near-field data is a practically important nonlinear inverse problem. One of the biggest challenges for the nonlinear inverse problem is to alleviate an ill-posedness resulting in poor accuracy. In this article, a novel source reconstruction method, which is based on an artificial neural network (ANN) enhanced by eigenmode currents, is proposed. Eigenmode currents, which work as macro basis functions, can be obtained numerically once precise geometry of the array antennas is found. The proposed source reconstruction method is named an ANN-EC (eigenmode currents) and is applied to diagnosis of array antennas based on phaseless near-field data. The ANN-EC has two advantages over conventional source reconstruction techniques purely based on the ANN. The first one is enhancement of accuracy and the second one is robustness to noise. Both of these advantages stem from reduction of the number of eigenmode currents using source reconstruction. Accuracy of the reconstructed currents is evaluated and these advantages of the ANN-EC over conventional ANN are demonstrated. To the best of the authors' knowledge, this is the first paper demonstrating the effectiveness of the eigenmode currents on the nonlinear inverse problem.