军事知识服务平台
基于CBR-IRBFNN的高原山区铁路路基绿色施工措施场景适配决策方法
Scenario adaptation decision-making method for green construction measures of railway subgrade in plateau mountainous areas based on CBR-IRBFNN
摘要 山区铁路路基若未有效践行绿色施工,易因施工中的不当扰动显著破坏植被生态、地形地貌,加剧区域水土流失,且后期修复困难.面向不同特征场景下差异化的绿色施工需求,提出了一种基于CBR-IRBFNN(case-based reasoning-improved radial basis function neural network)的高原山区铁路路基绿色施工措施场景适配决策模型.首先,在系统性收集相关绿色施工实例的基础上,采用文本挖掘方法识别出562个特征场景案例,通过结构化表示与量化分级构建绿色施工场景案例库,并通过K-medoids聚类方法验证了场景案例库的代表性与均衡性.进一步,考虑到已构建场景案例属性表示的多元异构性,基于RBFNN(radial basis function neural network)的局部响应特性与局部拟合能力,构建K-medoids与Gower距离方法相结合的场景案例智能检索模型,通过网格搜索结合交叉验证的方式确定最优的训练参数,在训练集和测试集上的准确率分别达到94.9%、94.4%,模型总体训练结果良好;相较于传统RBFNN方法,本研究方法在准确率、精确率、召回率、F1分数4个性能指标方面分别提升了8.7%、10.8%、8.8%、9.9%.最后,依托在建山区铁路路基工程实例,通过识别目标场景的具体特征及绿色施工需求,动态灵活调整检索策略,模型共检索推荐26项绿色施工措施,有效决策支持率达到92.3%,实现了绿色施工措施对目标场景的动态适配,反映了模型能充分贴合目标场景实际特征,具备良好的适用性和准确性.
Abstract:
If the green construction is not effectively implemented for the railway subgrade in mountainous areas,improper construction may significantly disrupt the vegetation ecology and topography,intensify regional soil erosion,and make later restoration difficult.In response to the differentiated green construction demands in various characteristic scenarios,this paper proposed a scenario adaptability decision-making model for green construction measures of plateau mountain railway subgrades based on CBR-IRBFNN.Firstly,based on the systematic collection of relevant green construction examples,562 characteristic scenario cases were identified using text mining methods.A green construction scenario case library was constructed through structured representation and quantitative grading.The representativeness and balance of the scenario case library were verified through K-medoids clustering methods.Further,considering the multi-heterogeneity of the constructed scenario case attribute representation,based on the local response characteristics and local fitting ability of RBFNN,a scenario case intelligent retrieval model combining K-medoids and Gower distance method was constructed to determine the optimal training parameters through grid search combined with cross-validation.The accuracy on the training set and the test set reached 94.9%and 94.4%,respectively.The overall training results of the model were good.Compared with the traditional RBFNN method,the proposed method improved by 8.7%,10.8%,8.8%and 9.9%respectively in terms of accuracy,precision,recall,and F1 score.Finally,based on the example of the railway subgrade project in the mountainous area under construction,by identifying the specific characteristics of the target scenario and the green construction requirements,and dynamically and flexibly adjusting the retrieval strategy,the model retrieved and recommended 26 green construction measures,with an effective decision support rate of 92.3%.It could achieve the dynamic adaptation of green construction measures to the target scenario.It can reflect that the model can fully adapt to the actual characteristics of the target scenario and has good applicability and accuracy.
刊 名 铁道科学与工程学报 ISTICEIPKU
年,卷(期): 2026, 23(9)
分类号: U215
机标分类号: TP391 F252.14 C934
在线出版日期: 2026年10月27日
基金项目: 2022:中国中铁股份有限公司科技研究开发计划项目,