Research On Intelligent Diagnosis Of Civil Aircraft Hard Landing

By study and gathering information, it was found that hard landing would cause a strong impact on the structure of aircraft, resulting in damage to the aircraft structure or even make it rupture. It was an important point to determine whether the landing was hard landing. In the base of investigation in the airways and analysis of related regulations, it was found a lot of deficiencies existed in the hard landing diagnosis method of domestic airlines. Therefore, the related factors that influence hard landing were analyzed, and the principle of hard landing appearance was revealed based on mechanics. The intelligent diagnosis model with multi-parameter information is established based on Support Vector Machine (SVM). Comparing with other methods of verification, the results show that the model could distinguish the hard landing effectively.Firstly, the formation of hard landing was a complex process, maybe there was a potential threat when the aircraft began to decline. Therefore, with aerodynamics of the aircraft and flight principle, from the velocity of approach, approach trajectory controlling, level off and grounding, the paper described the reasons in detail, and affirmed 5 parameters of the hard-landing diagnosis model. Confirmed the related parameters and classified them, it can help expert analyze the cause of event.Secondly, the record form and classification of flight data was briefly described. On these bases, the decoding algorithms of several hard-landing risk parameters are researched. Through the research on the procedure of flight data collection and the data error sources, the data preprocessing is studied. This paper reproduces the parameter curve at the moment of the aircraft touchdown, using the data fitting method.Thirdly, the intelligent diagnosis model of hard landing based on Support Vector Machine is studied. The model is trained and certified by the data sample from the B737, the results show that the model based on neural network could distinguish the hard landing effectively, which is feasible and intelligent. In the context of the application, the intelligent diagnosis results of the Support Vector Machine are contrasted with that of the Neural Network and the nearest neighbor algorithm. Experiment results show the validity and practicability of the model.