摘要
针对大坝观测数据常规模型训练后的残差混沌效应及模型回归方法的拟合度等问题,文中融合遗传算法与神经网络的数据训练优势,通过构建的遗传神经网络(GA-BP)算法对大坝变形观测序列资料进行回归提取残差序列.基于位移回归残差序列的混沌特性,利用混沌理论对其残差序列进行数值分析,并将残差预测结果与GA-BP预测模型进行叠加.据此,提出了考虑大坝变形残差序列混沌效应的GA-BP监控预测模型.实例表明,文中建立的预测模型的计算精度及收敛速度均得到提高,且考虑残差影响的大坝监控模型的预测效果得到了有效的提升.该模型的建模方法亦可推广应用于边坡及其他水工建筑物的安全预警.
Aiming at the problems of residual chaos effect in dam common monitoring model and the fitting degree of model regression methods, the dam deformation observation data fitting residual was extracted with genetic neural network （GA-BP） algorithm by combining the advantage of genetic algorithm and neural network. For the chaotic characteristics of deformation fitting residual, chaos theory was used to analyze residual series, and the residual prediction results were added in GA-BP prediction model. Hereby, the GA-BP prediction model considering chaos effect of dam displacement residual was established. Example showed that the calculation accuracy and convergence rate of this prediction model were all improved, and the prediction effect of this model was better. Furthermore, the modeling methods introduced in this paper could be applied to security early warning in slope and other hydraulic structures.
出处
《中国科学：技术科学》
EI
CSCD
北大核心
2015年第5期541-546,共6页
Scientia Sinica(Technologica)
基金
国家自然科学基金(批准号:51409139)
江西省教育厅青年科学基金(批准号:GJJ14223)资助项目
