Data Processing Model Based On The The Wavelet Theory Of Deformation Monitoring

Data processing is a part of deformation forecasting work. It defines that founding a corresponding model algorithm to predict for the deformation of inspection target through analyzing the inspection data gained. Generally speaking, the quantity of the deformation forecasting data is very large, so it provides important scientific basis for the stability and security of inspection target to research how to collect useful information from the large quantity of inspection data to analyze and predict in time and efficiently. The paper is aimed to research the model of data processing based on wavelets theory, its main concepts are as follows:1. Researching pretreatment methods of deformation forecasting data—wavelet packet de-noising method. It improves the principle of threshold value's selection mainly based on traditional wavelet packet de-noising, then compares the improved method with the traditional one to get the superiority of wavelet packet de-noising method.2. Researching wavelet neural network model's predication for deforming. It improves the convergence speed and approximate accuracy of wavelet network by synthesizing the advantage of wavelets theory aiming at traditional BP neural networks algorithm's defects and neural networks theory to founding model of wavelet neural network and introducing optimization method—BFGS used to solve the problem of unconstrained extreme. It ensures convergence of the internet by introducing a kind of auto correlative revise random aiming at the blindness of wavelet neural network parameters initialization assignment.3. Applying the improved wavelet neural network and traditional BP neural network into different projects:Predication for the sink of tunnel's top and the landslip of reservoir dam. We find that the convergence speed and predication accuracy of the improved wavelet neural network are more superior than traditional BP neural network through the experiment's compare and analysis.