Research On Deformation Model Based On Wavelet Analysis And Neural Network

Deformation monitoring throughout the building from the construction to the whole process of production, through the deformation monitoring, can in time to master the deformation law of buildings, the structure of the deformation forecast model is set up, found the abnormal situation rapidly take protective measures, to ensure the safety of buildings can normal operation. Because of the satellite navigation and positioning technology, 3 d laser scanning technology and other advanced technology development and application of deformation monitoring data is more and more complex, how to effectively extract information from a large number of monitoring data and mining useful information, make the deformation prediction in time, it is of great significance. Deformation prediction is an important part of deformation analysis and prediction in front of the data processing is also very important, this thesis mainly studies the deformation analysis of wavelet analysis and neural network model, including:(1) study of nonlinear wavelet threshold denoising method;(2) to study optimization improvement method based on the standard particle swarm algorithm;(3) research based on wavelet analysis and neural network, the combination of particle swarm optimization(pso) combined with the model prediction method.Paper aiming at the complex diversity of deformation data noise, theoretically puts forward the nonlinear wavelet transform threshold denoising new method, by building a new threshold function to achieve better denoising. For different signal denoising experiments and obtain different denoising effect. To obtain a better denoising effect, this paper further studies the optimal wavelet decomposition level of wavelet transform, selection of the optimum wavelet base, and the experimental results.Deformation of the BP neural network prediction model is established, auxiliary combining wavelet analysis and artificial neural network, the combination of wavelet analysis and neural network embedded wavelet neural network model. Although the gradient descent method can be optimized, but it’s easy to fall into local extremum. To overcome this defect, this paper introduce particle swarm algorithm to optimize the wavelet neural network model. For standard particle swarm optimization(pso) algorithm is easy to be precocious defect, to construct a new particle swarm algorithm, combined with the genetic mutation operator should the inertia weight optimization algorithm, the experiment proved that the improved particle swarm algorithm optimization ability is better than the standard particle swarm algorithm.Powerful approximation ability of artificial neural network and wavelet analysis of local amplification function, particle swarm algorithm combining optimization ability, form the integrated model, better realize the prediction of deformation monitoring data analysis, the nonlinear prediction provides a new method.