Research On Civil Aviation Hazard Management System And Its Key Technology

With the development of modern economy,more and more people choose the aircraft as a travel tool.In order to enable passengers to safely reach the destination,safety is an important theme in the air traffic management,among which identification and analysis of hazard is an important guarantee of safety.Because of wide variety of environment and complicated equipment and parameters in the flight,there are the large number of hazard and the deeper features are more and have strong correlation.This puts forward a challenge to the hazard identification and analysis.How to deal with the huge quantity of hazard data and analyze the deep characteristics of the hazard is the key point of the research on hazard identification and analysis.Based on the background of civil aviation safety risk management,the hazard management system of civil aviation is designed,and the improved deep extreme learning machine and particle swarm optimization algorithm are used to identify and analyze the hazard.The main research contents of this paper are as follows:(1)The general framework of civil aviation hazard management system and the main function structure is given,and the key technology in the system is provided.(2)A hazard identification algorithm based on deep extreme learning machine is designed.The algorithm is a deep structure composed of multiple stacked extreme learning machine(S-ELM)and a single extreme learning machine(ELM).Multiple S-ELM uses a parallel structure.Each has a different number of hidden nodes,which accepts hazard state information according the field of hazard and its last hidden layer outputs are used as inputs to the ELM.In the single ELM,the improved back propagation algorithm is introduced to improve recognition accuracy of the algorithm.At the same time,the way that generating input weights is enhanced and the method of training S-ELM respectively is used to alleviate the over fitting phenomenon and the excessive memory pressure when facing the high-dimensional datasets.The algorithm is verified on a database of a civil aviation hazard management system.The results show that the algorithm can improve the training speed and the recognition accuracy of the hazard.(3)A kind of hazard cause analysis algorithm based on weighted multi-population particle swarm optimization is designed.The algorithm is divided into preprocessing and the hazard cause analysis.In the preprocessing phase,weights are assigned to the items in hazard transaction database,and calculate the weighted range itemset to generate hazard candidate rules.In the hazard analysis phase,the weighted multi-population particle swarm optimization is used to generate the significant hazard as-sociation rules and analysis results of hazard is generated by tracing back along associate rules.In WMPSO,a parallel search patterns is a dopted,and the interaction mechanism between populations is provided.In order to generate more meaningful hazard association rules,the concept of weight of each particle in the population is introduced.At the same time,particle weight and global local optimization is introduced in the particle velocity updating formula to enhance the interaction between particles.(4)The main functions of civil aviation hazard management system are accomplished and the research productions in this paper are applied to this system.The design of algorithm and the main function modules are given in detail,and the typical interface is given.