Intelligent analysis of structural reinforcement strategy based on random forest
-
摘要: 工程结构加固是指为满足新的功能需求、安全标准或抗震设防要求对既有结构进行的加固补强。加固方案的选择往往依托技术人员的专业技术水平和工程经验，具有一定的主观性。本文采用机器学习中的随机森林算法作为数据分析工具，以结构信息和结构问题两方面共14个影响因素为输入特征，以结构整体加固和局部加固方案为输出特征，构建了结构加固策略智能化分析模型，有效解决结构加固方案复杂且组合情况众多而难以构建有效机器学习模型的问题。本文以122个实际工程案例为样本构建数据集，通过5个独立随机森林模型（整体结构、墙、柱、梁、板）的构建与整合实现了结构整体加固和局部加固方案的预测，分析了全部特征、重要特征和简单特征三种不同输入特征组合的随机森林模型预测结果以探究不同输入特征对模型准确率的影响。结果表明：基于全部特征和基于重要特征的机器学习模型的准确率相近，5个模型中准确率均可达到69％以上，其中部分模型可以达到80％左右；而基于简单特征的随机森林模型的准确率较低。由此可见，选取的重要特征与随机森林模型的预测准确率相关性高，而简单特征难以构建可靠的预测模型。本文研究成果可为结构加固领域的智能化决策和分析提供参考。
-
关键词:
Abstract: Engineering structural reinforcement refers to the reinforcement of existing structures to meet new functional requirements, safety standards or seismic defense requirements. The choice of reinforcement program often relies on the professional and technical level and engineering experience of technicians, and has a certain degree of subjectivity. In this paper, the random forest algorithm in machine learning is used as a data analysis tool to construct an intelligent analysis model of structural reinforcement strategy by taking a total of 14 influencing factors in two aspects of structural information and structural problems as the input features, and structural reinforcement as a whole and component reinforcement schemes as the output features, which effectively solves the problem of structural reinforcement schemes that are complicated and difficult to construct an effective machine learning model due to the many combinations.In this paper, 122 actual engineering cases are used as samples to construct the data set, and the prediction of overall structural reinforcement and local reinforcement schemes is realized by constructing and integrating five independent Random Forest Models (RFMs), and the prediction results of the RFMs with different combinations of all features, important features, and simple features are analyzed in order to explore the influence of different input features on the accuracy of the models. The results show that the accuracy of the machine learning models based on all features and important features is similar, and the accuracy of all five models can reach more than 69%, and some of them can reach about 80%; while the accuracy of the random forest model based on simple features is lower. It can be seen that the correlcation between the selected important features and the prediction accuracy of the random forest model is high, while the simple features are difficult to construct a reliable prediction model. The research results of this paper can provide a reference for intelligent decision-making and analysis in the field of structural reinforcement. -
点击查看大图
计量
- 文章访问数: 27
- HTML全文浏览量: 2
- PDF下载量: 9
- 被引次数: 0
下载:
