Real Time Cloud based Fault Detection and Alert for Antenna Array Using CNN

Abstract

The rise of smart antennas made it imperative to use adaptive beam forming, but single-element failure in array antenna disturbs the radiation pattern. Hence, to achieve adaptive beam forming, detecting and locating faults in an antenna array is a must for correction in the radiation pattern. The presented work in this paper provides real-time monitoring and alert for fault occurrence in antenna arrays. Fault detection is performed by an optimized Convolutional Neural Network (CNN), and the Thingspeak web service is utilized to store fault detection information on the cloud to get the visualization of faults. Voice over Internet Protocol (VoIP) Application programming interface (API) provides an alert in real-time once the fault is detected. The filter kernel parameters, number of convolutional layers, and dense layers of the fully connected neural network are tuned to minimize the mean square error (MSE). 16 element (4 × 4) planar microstrip array antenna is simulated on Ansys HFSS, and by introducing discontinuity, i.e. one type of defect in the corporate feed network, a training and testing set of 374 samples considering single element fault are used to train the CNN Model with a 0.3 validation split. Further performance of the same model by changing the last layer of a fully connected layer is evaluated to check the effectiveness of model when two elements are faulty.