Anomaly Detection in Iranian airport’s Aviation using Deep learning

Abstract

Aviation is one of the essential transportation elements in each country, which has always been the center of attention. One of the greatest damages to this industry, is the events that happen for the airplanes, as the most important element of this industry. These events might occur due to human error, unwanted hardware failure, felonious and terroristic deeds, which will change the usual behavior of the airplane; thus, it is expected that the anomaly detection methods might be useful in finding these events. The anomaly detection methods based on human supervision are very constrained. One of these constraints is human error. Another constraint is the heavy cost imposed on the aerospace industry. Finally, the most important constraint is the analysis of human force performance in checking all flights, which causes major problems in airports with heavy traffic. The automatic anomaly detection methods resolve the constraints and obstacles based on human supervision. The proposed method, which employs deep convolutional neural networks and data of three large airports, including Mehrabad (Tehran), Hasheminezhad (Mashhad), and Shahid Sadooghi (Yazd), can significantly reduce the events resulting from human error; an accuracy of 99.13% verifies the above claim.