基于GM-ANN模型的建筑物沉降量变化趋势预测方法  

Prediction method on change trend of building settlement based on GM-ANN model

作　　者：杨彪[1] 李慧民[2] 孟海[3] 裴兴旺 

机构地区：[1]西安建筑科技大学材料与矿资学院,西安710055 [2]西安建筑科技大学土木工程学院,西安710055 [3]中冶建筑研究总院有限公司,北京100088

出　　处：《中国安全生产科学技术》2016年第10期149-153,共5页Journal of Safety Science and Technology

基　　金：国家自然科学基金项目(51178386);住建部科技项目项目(2015-R3-003)

摘　　要：建筑物沉降观测结束之后,为降低和预防因地基不均匀沉降等因素造成的不安全事故发生率,准确预测建筑物沉降量变化趋势已引起相关科研单位的重视。首先,将人工神经网络数据分析与灰色GM(1,1)模型相结合,提出GM-ANN预测模型。然后,结合工程实例验证模型对监测沉降危险点数据变化的准确性,形成Matlab拟合曲线和预测趋势图。最终,结果表明仅考虑时间因素,GM-ANN模型明显优于灰色GM(1,1)模型,可使预测精度提高将近三倍。因此,利用GM-ANN预测模型可以对建筑物安全性进行有效预测。After the observation of building settlement,in order to reduce and prevent the occurrence rate of unsafe accidents caused by the uneven settlement of foundation,the accurate prediction on change trend of building settlement has attracted the attention of relevant research institutions. A GM- ANN prediction model was proposed by combining the data analysis of artificial neural network with grey GM（ 1,1） model. Combining with engineering example,the accuracy of the model for monitoring the data change of the settlement dangerous points was verified,and the MATLAB fitting curve and the prediction trend chart were obtained. The results showed that the GM- ANN model is obviously better than the grey GM（ 1,1） model only considering the time factor,which can improve the accuracy of prediction by nearly three times. Therefore,the GM-ANN model can be applied to effectively predict the safety of buildings.

关 键 词：建筑物 沉降观测 GM-ANN模型 MATLAB仿真 安全预测 

分 类 号：X947[环境科学与工程—安全科学]