为克服传统高速公路隧道运营安全风险评估方法计算过程繁琐、运算效率低及泛化能力差等问题，采用极限学习机（Extreme Learning Machine, ELM）神经网络模型对高速公路隧道运营风险进行评估。首先，基于系统工程理论，分析了高速公路隧道运营风险影响因素，构建了运营风险评估指标体系。然后，以全国126个隧道典型运营事故数据为样本集，基于ELM神经网络算法，对比不同激活函数模型的分类准确率和测试时间指标，选定Sigmoid作为激活函数，训练得到高速公路隧道运营风险评估模型。最后，以该模型为核心算法开发了隧道运营风险评估系统，并依托广东省某高速公路隧道路段开展了工程应用。结果表明，所构建的风险评估模型简化了人工计算过程，可提升高速公路隧道运营风险评估的及时性和有效性。

To overcome the problems of traditional operation risk assessment methods of highway tunnels, such as cumbersome calculation process, low computational efficiency and poor generalization ability, this paper conducted an operation risk assessment model of highway tunnels based on the ELM (Extreme Learning Machine) neural network. Firstly, based on the theory of systems engineering, the factors affecting operation risk of highway tunnels were analyzed, and the evaluation index system of operation risk was constructed. Then, taking the actual operation accident data of 126 tunnels in China as the sample set, the Sigmoid function was determined as the activation function based on comparing the classification accuracy rate and test time of different function. An operation risk assessment model of highway tunnels based on ELM neural network algorithm was trained. Finally, using this model as the core algorithm, an operation risk assessment system of highway tunnels was developed and applied to a highway in Guangdong Province, China. The results showed that the proposed risk assessment model simplified the manual calculation process and could improve the timeliness and effectiveness of operation risk assessment of highway tunnel.