Abstract: There is a great potential safety hazard in the special line of Guo-Bai Railway，due to its backward basic equipments and natural serious disaster.Therefore，it is very significant to carry out the safety evaluation on it.Because BP neural network had strong adaptability to the nonlinear and complexity of the safety evaluation of the special line of Guo-Bai Railway，the safety evaluation model of the special line of Guo-Bai Railway was established based on BP neural network.The improved BP neural network algorithm was adopted in the proposed model.Compared with the conventional algorithm，the improved BP neural network algorithm had the advantages of fast convergence，high calculation accuracy and high prediction accuracy.According to the special characteristics of the special line of Guo-Bai Railway，a set of safety evaluation index system easy to assign value and fit the actuality was set up，which increased the reliability of the model.The safety evaluation model of the special line of Guo-Bai Railway based on BP neural network and its prediction results provide reference and basis for the scientific management of the special line of Guo-Bai Railway.

Key words: traffic and transportation engineering, BP neural network, security evaluation, model, reliability

摘要： 郭白铁路专用线基础设备落后，自然灾害严重，存在很大的安全隐患，对其进行安全评价意义重大。由于BP神经网络对郭白铁路专用线安全评价的非线性和复杂性具有较强的适应性，故构建了基于BP神经网络的郭白铁路专用线安全评价模型。模型采用改进后的BP神经网络算法，相较于常规算法，具有收敛快、计算精确和预测精度较高等优点。针对郭白铁路专用线自身的特殊性，建立一套便于赋值和贴合实际的安全评价指标体系，增加了模型的可靠性。基于BP神经网络的郭白铁路专用线安全评价模型和通过模型预测出来的结果，为郭白铁路专用线科学的管理提供参考和依据。

关键词: 交通运输工程, BP神经网络, 安全评价, 模型, 可靠性