基于人工神经网络的机场鸟击风险预测

鸟击是影响民航机场运行安全的重要因素之一。为研究鸟类活动特征并预测其活动所造成的风险,基于某机场长期鸟类活动调研数据分析了鸟类活动与鸟击风险的关系,得到了4类12项对机场鸟击影响最紧密的风险因素。将上述因素作为输入向量,以鸟击风险作为输出变量,通过人工神经网络法(ANN)建立了机场鸟击风险预测模型,以长期鸟类活动风险记录作为数据库来训练和验证ANN模型的准确性,为提高预测的可靠性,又进行了模型优化和机场实地验证。结果表明,建立合理的ANN模型并利用长期鸟类活动数据加以训练,可有效预测机场鸟击风险。选择合理的输入变量可提高分析速度和预测精度,不仅有助于ANN模型的优化,还可以扩大ANN模型的应用领域以适应不同机场的鸟类活动特征,做出更合理的鸟击风险预测,为机场运行保障提供有力的技术支持。综合考虑"鸟""机""环""管"等因素建立的飞行区鸟击风险预测模型表明,飞行速度(V)对鸟击风险因素评价的影响最小,飞行高度(H)和飞行路线(R)对鸟击风险因素评价的影响相对较小。

Prediction model of the bird strike risk in the airport sphere by using the artificial neural network

The paper is aimed at tracing and examining the bird strike in the civil aviation airport.As is well known,the bird strike is one of the most critical factors that may affect the operational safety of the civil aviation airport.For the above said purpose,we have made an investigation of the features of the bird flying activities and their relation with the safety of the airport as well as the aircraft flying safety based on our long-term observation and the corresponding data collection.In this method,we have to check the 4 categories,i.e.the birds,the flying machine,the environment and the management,along with the 12 major factors,that is,the season,the number of the airplanes,the flying time,the sensitivity,the flying height,the flying frequency,the weight of the passengers and boarding wares,the flying speed,the flying destinations,the flying routes,and the pathways suggested,all of which may involve the bird strike liabilities obviously.And,therefore,they should be put into the variables of the flight safety.Among the said variables,the bird strike probability index can be adopted as the output one.To establish the prediction model of the airport bird strike by the artificial neural network(ANN),a long-term record of the bird activity risks of the airport has been used as the database to train and verify the accuracy of the ANN model.Besides,the ANN model has to be optimized by reducing the numbers of the input variables in order to improve the prediction reliability,and the optimized ANN model has also to have been tested by the airport research data on site.The results of our investigation show that it is an effective method for the bird strike risk prediction by establishing a reasonable ANN model based on the long-term record of the bird strike.Needless to say,the ANN model can be optimized in analyzing the efficiency and prediction precision by choosing the reasonable input variables.With the help of the input variables,it would be possible to broaden the application area of the suggested ANN model and enhance the prediction effect and efficiency of the bird strike risk prediction according to the different bird activity features of and around the airport.Taking into careful account the said 4 major factors,the ANN model shows that the flying velocity(V)may have the least effect on the evaluation efficiency of the bird-strike risk factors,and in turn make much less impact of the height(H)and the route(R)on the flight-risk liability.