基于多任务学习的公路隧道支护方案优化分析

为提升公路隧道施工过程中支护设计的实时性和合理性，以浙江省龙丽温高速典型隧道的围岩地质特征和支护方案为研究背景，分析围岩特征与支护参数的相关性，确定了以围岩风化程度、坚硬程度、岩体完整程度、地下水状态、围岩稳定情况为输入围岩特征，锚杆间排距、锚杆长度、喷砼厚度、钢拱架型号为输出支护参数的多任务学习模型结构。针对多任务学习模型预测的支护参数，采用层次分析法对比评价了预测与实际支护参数。结果表明预测支护参数综合得分高于实际支护参数，更为优秀。研究方法及成果为实现公路隧道快速支护方案选型和评估有较大参考价值。

Optimization design of Highway Tunnel support Based on Multi-task learning

In order to improve instantaneity and rationality of supporting design in the highway tunnel, the typical surrounding rock and support design features of Long-Liwen tunnel in ZheJiang Province were collected. After a correlation analysis, decay level, hardness, intergrity, water level and rock stability are selected as surrounding rock features. Anchor interval, anchor length, concrete and frame are selected as predict parameters of support design. A multi-task learning model is trained based on these features. The analytic hierarchy method is used to evaluate that model predict and actual parameters. The evaluate result shows that the predict parameters have higher score than the actual one, which is better. The research shows a high reference value for highway tunnel support instant selection and evaluation.