Research And Application Of Building Foundation Settlement Prediction Base On The Theory Of Wavelet Neural Networks

In order to solve the problem of uneven settlement of building,settlement monitoring of buildings is the method of obtaining subsidence data,but the real purpose is to get reasonable research and analysis of having access to the settlement data.And then that provide reasonable scientific basis for construction decisions.On account of the construction foundation settlement of the specific engineering instance,and integrating the current research methods in the field of settlement prediction analysis,this article is using a self-organizing learning and characteristic of nonlinear approximation of intelligent neural networks.Through denoising principle in the theory of wavelet analysis deal with the noise for monitoring to the building foundation settlement data,and then establishing the BP neural networks,RBF networks and the local improvement of CVRBF networks model,this article use the established three models to predict the processed settlement data.On the basis of analysis of the forecast results,that demonstrates the superiority and feasibility of the CVRBF neural network in the aspect of building foundation settlement prediction.This paper mainly studied the following aspects:(1)This paper introduces BP neural networks and RBF neural networks.Through the analysis of the two network algorithms,in theory shows that RBF neural networks are superior to the BP neural networks in learning rate,approximation ability,precision,and functional aspects.(2)According to the concrete construction projects,which monitors the building foundation settlement and gets the building settlement data.Because the construction monitoring in the process of foundation settlement will be affected by the factors mutual influence and have the certain error between the monitoring settlement data and the actual settlement data.Through the denoising principle of wavelet analysis theory,this article has carried on the denoising processing for the monitoring data.The subsidence data after processing more consistent with the actual situation.(3)According to the theory of intelligent neural networks,this paper respectively established the BP neural networks,RBF networks and CVRBF networks prediction model.Through setting and training the network model parameters,getting the network model is suitable for building foundation settlement prediction.Using the foundered complete forecasting model to forecast the foundation settlement,and make the data comparison.(4)Through the results contrast analysis,in the model stability,data precision,etc,CVRBF neural networks is superior to BP neural networks and RBF networks.In the theory and the actual data,which demonstrates the CVRBF neural networks has more advantages in terms of building foundation settlement prediction.Playing the role of a forecast on the building's construction,schedule control,security and so on,this conclusion provides practical basis for the establishment of safety monitoring scheme and has important and far-reaching application value in the process of engineering construction in the future.