基于灰色相关分析的GRFM倾斜量预测模型

发布时间：2019-04-28 15:38

【摘要】：针对时间序列的动态性、相关性、小样本性、非线性等特征,利用灰色模型的小样本适用性和神经网络的预测高精度等性能,提出了一种基于相关分析的灰色神经网络组合预测模型.首先,基于灰色相关理论定量分析了倾斜量与沉降观测指标时间序列之间的相关度;然后,采用GM(1,1)模型对原始序列累加求和,降低各因素原始数据的噪声干扰;利用优化径向基神经网络(Radial Basis Function,RBF)多步拟合训练,其中心点和扩展系数初值采用蚁群算法进行优化.最后,将该模型应用到了砖石古塔的倾斜量预测中,设计了沉降综合指数,通过计算,该指数与倾斜量的灰色相关度为0.789 1,采用该模型对某古塔倾斜量进行了预测,平均相对误差为9.056%.实验结果表明,该模型对小样本、非线性的时间序列预测具有高精度和有效性,为古建筑保护中变形预测提供了理论和实践经验.
[Abstract]:In view of the characteristics of time series, such as dynamics, correlation, small sample nature, nonlinearity and so on, the applicability of small samples of grey model and the performance of high precision prediction of neural network are used. A grey neural network combined prediction model based on correlation analysis is proposed. Firstly, the correlation between the inclination and the time series of settlement observation index is analyzed quantitatively based on the grey correlation theory, and then the GM (1,1) model is used to sum up the original series to reduce the noise interference of the original data of each factor. The optimal radial basis function neural network (Radial Basis Function,RBF) is used for multi-step fitting training. The center point and the initial value of the expansion coefficient are optimized by ant colony algorithm (ACA). Finally, the model is applied to predict the slope of the ancient brick and stone pagoda, and the comprehensive index of settlement is designed. Through calculation, the grey correlation between the index and the inclination is 0.789 1. The model is used to predict the slope of an ancient pagoda. The average relative error is 9.056%. The experimental results show that the model has high accuracy and effectiveness for the prediction of small samples and nonlinear time series, which provides theoretical and practical experience for the prediction of deformation in the protection of ancient buildings.
【作者单位】：西安建筑科技大学管理学院;西安科技大学通信与信息工程学院;西安建筑科技大学信息与控制工程学院;西安建筑科技大学土木工程学院;陕西省文物保护研究院;
【基金】：国家教育部归国留学人员科技支撑基金资助项目(K05055) 西安市碑林区科技计划基金资助项目(GX1614)
【分类号】：TU196.3

【相似文献】