黄土力学数据挖掘系统研究

Data mining system of loess mechanics

岩土工程智能化系统研究是岩土工程领域前缘问题。本文基于信息技术中最先进的数据挖掘技术,研制了黄土力学数据挖掘系统,包括有预处理模块、挖掘操作模块、知识库管理模块,自重湿陷与湿陷性判定决策树模型、判定规则和评价模型的精度以及其他应用软件接口等。可实现数据的归约、聚类分析、分类、预测等。通过工程实例分析,证明了该系统的有效性和实用性。采用59项工程的2766组黄土试验数据对决策树的每条规则进行测试,结果表明自重湿陷性判定规则的准确率为87.3%,湿陷性判定规则的准确率为92.5%。利用挖掘所得到的判定规则和决策树模型,可减少试验工作、降低成本;提出的模型还可用于确定湿陷底界与自重湿陷的底界,从而为地基处理深度的确定提供依据。

The intelligent system of geotechnical engineering is an important research interest in geotechnical engineering field. A data mining system of loess mechanics was developed based on advanced data mining system in information technology. The data mining system of loess mechanics included the preprocessing module, mining operation module, knowledge base management module, and software interface of geotechnical application. The system could be applied to data reduction, cluster, classification, and prediction for loess mechanics. Through a lot of engineering applications, the results indicated that the system was effect and practical for loess collapse. Through the assessment of each rule in decision trees model by using 2766 groups of loess testing data in 59 projects, the results show that the precision discriminating self weight collapse loess is about 87.3%, and the precision discriminating collapse loess is about 92.5%.