期刊文献+

基于卷积神经网络的高速公路建筑安装工程造价估算研究 被引量:4

Cost Estimation of Expressway Construction and Installation Engineering Based on Convolutional Neural Networks
下载PDF
导出
摘要 为在投资决策阶段快速、准确地得到拟建高速公路项目的造价,本研究基于历史高速公路案例资料,应用卷积神经网络算法,建立了一种高速公路建筑安装工程造价估算模型。首先通过文献研究与专家访谈确定了影响高速公路造价的工程特征属性指标,并对其进行量化用于模型计算;然后将关键指标重构为二维矩阵,基于卷积神经网络算法,建立高速公路建筑安装工程造价估算模型;最后利用已建高速公路项目信息验证该模型的有效性,并对比分析卷积神经网络模型与反向传播神经网络模型预测效果。分析结果表明,使用卷积神经网络建立的高速公路造价估算模型相对误差在4%~7%之间,而反向传播神经网络的相对误差在4%~15%之间,故本研究构建的模型能较准确、高效地预测高速公路工程项目造价,为工程项目的前期投资决策、全寿命周期造价管理提供一定参考。 Investment estimation is an important basis for the technical and economic evaluation and investment decision-making of highway engineering.The key to saving investment lies in the pre construction stage.Investment estimation,as a key work in the investment decision-making stage,has a great impact on cost management and has significant implications for project cost control.In order to quickly and accurately obtain the cost of the proposed highway project during the investment decision-making stage,carry out pre funding planning,and reduce funding risks,this study is based on historical highway case data and applies convolutional neural network algorithm(CNN)to establish a highway construction and installation engineering cost estimation model.Firstly,through literature research and expert interviews,a total of 11 engineering characteristic attribute indicators affecting the cost of highways were determined,including the total mileage of the main line,pavement type,number of bridges per kilometer,number of tunnels per kilometer,roadbed width,earthwork volume,number of separated interchanges per kilometer,number of interchanges per kilometer,number of channels per kilometer,terrain and topography,and the location.These indicators were quantified for model calculation;Then,the key indicators are reconstructed into a two-dimensional matrix.Based on the convolutional neural network algorithm,a cost estimation model for highway construction and installation engineering is established.The parameters of the highway engineering investment estimation prediction model are optimized using the Adam algorithm to reduce the complexity of the model while ensuring its prediction accuracy;Finally,the effectiveness of the model was verified using information from existing highway projects,and the traditional BP neural network was applied to predict the cost of highway engineering.The prediction performance of the CNN model and the traditional BP neural network model was compared and analyzed.The analysis results indicate that the CNN estimation model has a higher degree of fit and closeness between the predicted curve and the true value curve of the test set samples compared to the BP neural network.The predicted cost values of each sample obtained are relatively close to the actual values;The CNN estimation model uses convolutional neural networks to establish a highway cost estimation model with a relative error between 4%and 7%,which meets the error requirements for investment estimation.However,the relative error of the BP neural network is between 4%and 15%.Therefore,the model constructed in this study can accurately and efficiently predict the cost of highway engineering projects in the investment decision-making stage where the engineering data information is not detailed and clear enough,which is helpful for the early investment decision-making of engineering projects The full life cycle cost management provides a certain reference,a fast and accurate theoretical basis for the preparation of engineering feasibility studies and investment and financing decisions,and also provides a new idea for highway cost estimation.The model can be optimized in the future to further improve its prediction accuracy.
作者 袁剑波 殷婵 Yuan Jianbo;Yin Chan(School of Traffic and Transporatation Engineering,Changsha University of Science and Technology,Changsha 410114,China)
出处 《工程研究(跨学科视野中的工程)》 2023年第5期446-455,共10页 JOURNAL OF ENGINEERING STUDIES
基金 湖南省交通运输厅科技进步与创新项目(202039) 湖南省研究生科研创新项目(CX20200826)。
关键词 高速公路 造价估算 预测模型 卷积神经网络 expressway cost estimation predictive model convolutional neural network
  • 相关文献

参考文献20

二级参考文献168

共引文献194

同被引文献21

引证文献4

二级引证文献9

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部