Antenna Array Fault Detection Using Logistic Regression Technique

Array antenna is widely used in 5G wireless networks. The main issue with the array antenna is that, failure or fault in one or more elements disturbs its radiation pattern, and its directivity pattern gets disturbed. Fault in elements of the array antenna enhances the side lobe levels; To tackle this kind of issue, pattern regeneration techniques can be utilized, but for that, the detection of the faulty element is needed. Hence to cater this kind of problem exemplary machine learning technique, i.e., logistic regression classifier, is used in this paper. The eight-element planar antenna is simulated using the Ansys-HFSS tool. To build training and testing datasets, a discontinuity is formed in the array’s feed network to simulate various fault conditions. The array is designed using RT Duriod 5880 substrate with relative permittivity 2.2 and thickness 1.6 mm; the simulated result shows a high gain of 12 dB and S11 of −32 dB for 3.67 GHz frequency. Before applying a logistic regression machine learning algorithm for fault detection in an antenna array, a review of various techniques applied by researchers is carried out. The logistic regression multiclass model with a liblinear solver obtained 95% accuracy over 105 test samples.