高速列车侧墙结构的决策树隔声预测模型

Decision tree sound insulation prediction model for side wall structure of high-speed train

为实现对高速列车侧墙复合结构快速、精准的隔声预测,基于已有的前期大量实测数据,构建了高速列车侧墙复合结构隔声样本数据库,并提出了一种基于决策树算法的隔声预测模型,建立了结构的隔声量与材料参数之间的映射关系。首先,介绍了原始实测样本的来源,并进行了材料组成分析、整理和归类;接着,对影响侧墙复合结构隔声性能的主要因素进行分析,开展了显著特征的筛选;最后,基于机器学习中的决策树算法对模型进行了训练和验证,并与传统有限元-统计能量分析预测方法进行了对比。结果表明:相较于传统有限元-统计能量分析模型,该决策树学习模型对高速列车侧墙复合结构隔声特性的预测准确度和预测效率均显著提高。随着将来更多样本的加入,模型还可以被进一步改进和完善,具有较大的工程实用意义和推广潜力。

In order to achieve rapid and accurate sound insulation prediction for the composite structure of high-speed train side walls,we have built a sound insulation sample database based on a substantial amount of pre-existing measured data.Furthermore,a sound insulation prediction model using a decision tree algorithm has been proposed.The model establishes the mapping relationship between the sound insulation of the structure and the material parameters.Firstly,the source of the original measured samples is introduced,and the materials are analyzed,organized,and classified.Next,the main factors affecting the sound insulation performance of the sidewall composite structure are analyzed,and significant features are selected.Finally,the model is trained and validated using the decision tree algorithm in machine learning,and a comparison is made with the traditional finite element-statistical energy analysis(FE-SEA)prediction method.The results show that compared to the traditional FE-SEA model,this decision tree learning model significantly improves the accuracy and efficiency of predicting the sound insulation characteristics of high-speed train sidewall composite structures.With the inclusion of more samples in the future,the model can be further improved and perfected,making it highly practical and promising for engineering applications.