地采诱发建筑物损害识别的进化支持向量机模型  

Model of nonlinear SVM with GA to estimate building damage due to mining-induced ground movements

作　　者：蒋成荣[1] 彭平安[2] 王李管[2] 

机构地区：[1]攀枝花学院资源与坏境工程学院,四川攀枝花617000 [2]中南大学资源与安全工程学院,湖南长沙410083

出　　处：《中南大学学报（自然科学版）》2015年第9期3368-3376,共9页Journal of Central South University:Science and Technology

基　　金：国家高技术研究发展计划(863计划)项目(2011AA060407);“十一五”国家科技支撑计划项目(2006BAB02A02);国家自然科学基金资助项目(50774092)~~

摘　　要：针对采动区建筑物损害程度的影响因素较多且各因素呈现非线性、多重共线性等特点,应用支持向量机理论并结合工程实际,提出基于支持向量机理论的地采诱发建筑物损害效应预测方法。综合考虑地质采矿方面和建筑物本身因素,选取10个影响砖混结构建筑物采动损害程度的因素作为模型的输入,将砖混结构建筑物的损害等级及建筑物的最大裂缝宽作为模型的输出,以32个建筑物采动损害的工程实例作为学习样本进行训练;采用RBF核函数,建立建筑物损害等级预测的支持向量机分类和最大裂缝宽回归模型;为提高预测模型的泛化能力和预测精度,应用遗传算法选择支持向量机的模型参数,并对6组待判样本进行判别。研究结果表明:建立的遗传算法优化支持向量机分类与回归模型对地采诱发建筑物损害效应预测效果良好,评估结果与实际结果相吻合。Considering that the evaluation of the effects of builting structures damaged by mining of many factors are not deterministic, nonlinear and has multiple linear features, a new method of the support vector machine（SVM） to predict the effects of the structures damaged by mining was proposed based on the statistical learning theory and the actual characteristics of the project. Considering the geological mining and building factors, ten large factors affecting buildings damage of brick and concrete structure were selected as the proposed model input variables, the damage level of the brick and concrete structure buildings as well as structures of the largest crack width were taken as the proposed model output values. Twenty-two typical cases of buildings and structures damaged by mining were used for training data by introducing radial basis function（RBF） kernel function. For enhancing the generalization performance and prediction accuracy, genetic algorithms（GA） were applied to select parameters for SVM model, thus the damage level prediction of building GA-SVM classification and the maximum crack width with GA-SVM regression model were established,respectively, and another six group cases were sentenced to distinguish samples for further study of the effectiveness and practicality of the proposed model. The results show that the establishment of support vector machine classification and regression model prediction of the effects of building structures damaged by mining can achieve high accuracy, which coincides with the actual results.

关 键 词：地下开挖 建筑物损害 损伤评价 支持向量机 遗传算法 预测 

分 类 号：TD229[矿业工程—矿井建设]