基于神经网络的鸟撞预测模型应用研究

Bird-Strike Prediction Model Application and Research Based on Neural Network

通过对某机场地理环境、生境、鸟类及鸟类活动、鸟情信息进行综合分析,在收集、规范、完善鸟情数据集的基础上,文中给出了鸟情数据预处理方法,提出了一种基于神经网络的鸟撞预测模型,确定了鸟撞预测模型输入、输出参数,并分别采用BP神经网络和径向基网络(RBF)两种方式建立预测模型。运用Matlab对两个预测模型性能进行仿真实验,仿真实验结果表明BP神经网络比RBF神经网络预测具有更高精度、更小误差。最后,通过机场实际观察数据对训练好的鸟撞模型做仿真预测,结果进一步验证了BP神经网络鸟撞预测模型具有很好的预测效果。

Through comprehensive analysis to geographical environment, habitats, birds and their activities, bird informations to an airport, on the basis of collection,standardized,improved data set,it givexl the processing method of bird information,proposed a bird-strike pre- diction model based on neural network, determined the input, output and other parameters, moreover using the BP neural network and RBF neural network methods to establish it. Using the Matlab simulated to the two prediction models,simulation results show that the BP neural network prediction is more accurate and effective than RBF neural network. Finally, through the observed data trained on the airport simulate to the bird beat model ,examples verificated the bird-strike prediction model of BP neural network has good prediction results.