基于神经网络的公路补充定额编制方法及应用研究

发布时间：2018-05-17 01:30

本文选题：定额编制 + 定额影响因素　；参考：《长沙理工大学》2012年硕士论文

【摘要】：随着科学技术的飞速发展，新工艺、新设备、新技术、新材料不断的更新换代，现有的公路工程定额通常难以完全满足各地区的公路建设要求，致使一些项目投资控制缺少依据，从而失去对工程计价的快速反映，造成工程造价管理失控。因此急需编制针对各地区的补充定额。本文首先阐述了国内外定额的研究现状，并分析对现行定额的编制方法及编制补充定额的重要性及现实意义。通过AHP法对众多影响因素进行筛选，确定出5个主要的定额影响因素。通过对公路补充定额传统编制方法研究，构建神经网络定额生成系数模型，计算出能更加真实反映生产管理水平的数据，为整个定额体系的合理编制奠定基础。根据定额编制原理及现有的定额编制方法，分析了现有补充定额编制的滞后性，分析、筛选定额数据的影响因素，确定了神经网络的5个输入神经元。运用BP神经网络结构和算法，阐述神经网络的输入层定额生成系数的内涵，并应用MATLAB对定额原始数据进行模拟和预测。在建模过程中，采用改变模型中隐含层神经元个数和训练函数类别的试凑方法，分别进行拟合预测，通过比较均方差，发现当采用隐含层神经元个数为14、训练函数为traingda、激活函数为pureline的BP神经网络模型进行训练样本的检验时，所得的预测误差很小。再通过4组检测样本的验证说明该模型已训练成功。最后结合《云南公路工程定额编制规程的关键技术研究与应用》研究课题，，运用构建的计算模型进行实证分析，通过震颤标线、路堑土工格室三维植被网、特殊路基防排水处理三个定额子目验证模型能够推导出定额生成系数，并且通过定额生成系数得到补充定额。通过实例分析，验证了本文所建模型的科学性和可行性。
[Abstract]:With the rapid development of science and technology, new equipment, new technology, new materials are constantly updated, the existing highway engineering quota is usually difficult to fully meet the requirements of highway construction in various regions. As a result, some project investment control lacks the basis, thus loses the quick response to the project valuation, causes the project cost management to be out of control. Therefore, there is an urgent need to compile supplementary quotas for each region. In this paper, the present situation of quota research at home and abroad is first expounded, and the importance and practical significance of compiling and supplementing quota are analyzed. Through the screening of many influencing factors by AHP method, five main quota influencing factors were determined. Based on the study of the traditional compilation method of highway supplementary quota, the neural network quota generating coefficient model is constructed to calculate the data which can reflect the production management level more truthfully, which lays the foundation for the rational establishment of the whole quota system. According to the principle of quota compilation and the present method of quota compilation, the lag of existing supplementary quota compilation is analyzed, the influencing factors of quota data are analyzed, and the five input neurons of neural network are determined. By using BP neural network structure and algorithm, the connotation of input level quota generating coefficient of neural network is expounded, and the original quota data are simulated and predicted by MATLAB. In the process of modeling, the method of changing the number of hidden layer neurons and the category of training function in the model is used to fit and predict respectively, and the mean square deviation is compared. It is found that the prediction error is very small when the BP neural network model with 14 hidden layer neurons traingda training function and pureline activation function is used to test the training samples. Through the verification of 4 groups of test samples, the model has been trained successfully. Finally, combining with the key technology research and application of Yunnan highway engineering quota compilation regulation, using the constructed calculation model to carry on the empirical analysis, through the tremor mark line, the cutting geogrid three-dimensional vegetation network, The model of three quota subitems for special subgrade waterproofing and drainage treatment can deduce the coefficient of quota generation and obtain the supplementary quota through the coefficient of quota generation. The model is proved to be scientific and feasible by an example.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F542;F285;F224

【相似文献】