【摘要】 减灾选线是从源头上降低铁路工程全生命周期地灾风险的总体性工作，其智能化转型是复杂山区铁路建设的重要技术需求。目前，尽管人工选线已经积累了大量减灾选线的经验，但这些经验通常以非结构化文本或隐性知识形式分散于规范文件及专家认知中，难以直接被智能选线程序应用，严重制约智能选线系统对多地灾耦合风险的动态响应能力。为此，综合考虑岩溶、崩塌、滑坡、泥石流、冰川、软土、地震及断裂带等地质灾害对铁路线路的威胁，建立基于超图语义的铁路减灾智能选线知识图谱(hazard reduction railway alignment design knowledge graph, HRRAD-KG)。在此基础上，综合考虑铁路建设成本、地震概率风险和典型地灾风险，构建风险-成本多目标铁路线路优化模型。为求解上述优化模型，提出一种HRRAD-KG驱动的粒子群优化算法，并采用多准则锦标赛决策(multi-criteria tournament decision, MTD)算法实现多目标间的综合决策。最后，通过一个我国山区的真实铁路案例，验证本方法在辅助实际铁路选线项目中的有效性。研究结果表明，HRRAD-KG驱动的铁路智能减灾选线方法，能够压缩穿越岩溶区的隧道长度，降低隧道遭受突水突泥的风险。与人工设计的最优方案相比，该方法的建设成本降低了2.94%，总地灾风险降低了16.12%。研究成果可为进一步提升铁路智能选线方法的质量和效率，保障复杂山区铁路建设和运营安全提供参考。更多还原

【Abstract】 Hazard reduction railway alignment design is the overall work to reduce the life-cycle geological hazards of railway engineering from the source, whose intelligent transformation is an important technical requirement for railway construction in complex mountainous regions. Currently, although a lot of experience of hazard reduction railway alignment design has been accumulated in manual design, these experiences are often dispersedly stored as unstructured specification texts or tacit knowledge in the experts ’ cognition, making it difficult to be directly applied in railway alignment optimization programs. Meanwhile, it seriously restricts the dynamic response ability of alignment optimization system to multiple coupling hazards. To this end, the hazards of karsts, collapses, landslides, debris flows, glaciers, soft soil, earthquake and fault zones to railway alignment were comprehensively considered in this research. The Knowledge Graph of Hazard Reduction Railway Alignment Design(HRRAD-KG) was established based on hypergraph semantics. Based on this, a hazard-cost multi-objective railway alignment optimization model considering railway construction cost, earthquake probability risk and typical geological hazard, was constructed. To solve the above optimization model, a HRRAD-KG driven particle swarm optimization(PSO) algorithm was proposed. Meanwhile, a multi-criteria tournament decision(MTD) algorithm was incorporated to realize the comprehensive decision among multiobjectives. Finally, the effectiveness of the devised method was verified via a real-world railway case in mountainous regions. The results indicate that the HRRAD-KG-driven hazard reduction railway alignment optimization method can compress the tunnel length through the karst regions, which can reduce the possibility of tunnels being threatened by water outburst. Compared with the best manual alternative, the construction cost and the total geo-hazard of alternative produced by the proposed method are reduced by 2.94% and 16.12%, respectively. This research can offer practical guidance to enhance the precision and efficiency of railway alignment in complex mountainous regions, thereby guaranteeing the safety of railway construction and operation.更多还原