基于小波去噪的灰色关联BP神经网络在大坝沉降变形监测中的应用

Grey Relational BP Neural Network Based on Wavelet Denoising in the Application of Dam Settlement Deformation Monitoring

针对大坝变形中存在多种影响因子干扰的问题,提出了基于小波去噪的灰色关联BP神经网络模型.首先利用小波阈值去噪方法对数据进行处理,然后通过灰色关联分析对大坝沉降变形监测中多种影响因素计算分析,得出关联度大的影响因子,将灰色关联分析和BP神经网络相结合建立模型,最后与未经数据处理的灰色关联BP神经网络以及卡尔曼滤波模型进行对比分析,得出结论.实验表明,经过小波去噪数据处理后,提高了灰色关联BP神经网络模型的预测精度和可靠性.通过灰色关联分析,可以对大量的输入变量进行处理,而不必经过主观的筛选,进而增加了BP神经网络的适应能力,同时预测的结果比其他单一模型更加接近最终实测值,具有更高的准确性与可信度.

Aiming at the disturbance of many factors in dam deformation,a grey relational BP neural network model based on wavelet denoising is proposed in this paper.Firstly,wavelet threshold denoising method is used for data processing,and then the grey relational analysis is used to calculate and analyze the factors affecting the deformation monitoring of dam settlement,and obtain the influence factors of greater relevance.A model is established by combining grey relational analysis and BP neural network,and finally it is compared with the grey BP neural network without data processing and Calman filter model to obtain the conclusion.The experimental results show that the prediction accuracy and reliability of the grey relational BP neural network model are improved after the wavelet denoising data processing.Using grey relational analysis can deal with a large number of input variables,without subjective screening,and then increase the adaptive ability of BP neural network,the prediction results than other single model closer to the final measured with higher accuracy and reliability.