Deep Learning Based Antenna Array Fault Detection

Abstract

MmWave communication is a promising communication technique of future 5G system. Large bandwidth and high directional gain are the two advantages of mmWave communication solutions. High directional gain leads the usage of massive units. The system maintenance and status management play more and more important role in future wireless communication systems and contribute higher operation cost than before. In our paper we proposed a fault finding and location methods based on two deep neural network (DNN) with different complexity, the simple network is designed to fault finding with low cost, the other precision network start to fault location when the former one has detected the faults. Simulation results show that our network can work well at low SNR region without manual inspection involved, thus the operation cost is reduced.