一种地铁塌方风险评估模型的设计仿真

Design and Simulation of a subway collapse risk assessment model

以提升施工安全、避免塌方事故发生为目标,设计基于LightGBM的大跨度地铁塌方风险评估模型。分析大跨度地铁塌方影响因素后,将材料设备、技术、勘察设计、管理、人员以及环境作为准则层,架构塌方风险评估指标体系,根据判定矩阵标度验证判定矩阵合理性,明确指标层中各指标的对应权值。依据梯度提升决策树与叶子分裂模式,经过不断分裂最大分裂增益叶子节点,得到LightGBM综合评估模型。通过样本采集、模型参数训练以及塌方风险分析等阶段,获取塌方风险的评估等级结果。仿真针对某地铁一期工程的实际情况展开塌方风险评估,经对比不同开挖阶段的实际风险等级与评估风险等级,结合AUC、K-S指标结果,检验出所建模型的评估精准度较为理想,具有一定的可行性。

The collapse risk assessment model of long-span subway based on LightGBM is designed to improve construction safety. The influencing factors of long-span subway collapse were investigated in detail. Materials, equipment, technology, survey and design, management, personnel and environment were regarded as the criteria layer. The collapse risk was built to evaluate the index system. The matrix scale was determined to verify the rationality of the matrix, and the corresponding weight of each index in the index layer was also confirmed. Gradient theory was applied to improve the decision tree and leaf splitting model. The leaf node was gained by the maximum splitting, so the LightGBM comprehensive evaluation model was obtained. According to sample collection, model parameter training and collapse risk analysis, the evaluation grade results of collapse risk were obtained. According to the AUC and K-S index results of example analysis, it shows that the model has high evaluation accuracy and excellent feasibility.