Study On Chaotic Characteristic And Prediction Model Of Dynamic Deformation Monitoring Data

Deformation data analysis and prediction is a complex systematic project. With the development of the new technique of deformation monitoring such as GPS, the new methods of extracting deformation information, predicting and evaluating the stability of deformation body based on introducing advanced mathematical theory and signal analysis methods to deeply understand the non-linear and complexity of deformation.The main results of this study and the specific contents are as follows:(1) Introducing the judging method of chaotic behavior of several types of deformation body. Determining the non-linear judgment of deformation monitoring combined with examples of actual projects. Pointing out that the Lyapunov-index can serve as a deformation evaluation criteria in judging different part of dynamic deformation body.(2) Wavelet transform has a strong ability to remove the noise of the chaotic data, however empirical mode decomposition(EMD)does well in extracting of the information. In this paper, comparing the two methods and pointing out their advantages and disadvantages with the help of simulation. And proposing Wavelet-EMD coupled model to extract the deformation information and remove the noise. The results in real project show that the model is very useful in continuously dynamic GPS deformation monitoring data.(3)Kalman filter has an original advantage in eliminating the white noise and predicting future systematic state of deformation data. The key research has analyzed the noise reduction capabilities of Kalman filter and EMD. The dynamic GPS data of a surface mining area in Mongolia was processed and analyzed to prove the effectiveness in practical application of the model.(4) Chaotic time series of deformation body was multi-step predicted and obtained ideal prediction results through weighted local-region model can be gotten, which based on chaotic non-linear analysis method. Multi-steps neural network method was studied by phase space reconstruction and time delay BP neural network prediction model was reconstructed based on chaotic phase space reconstruction method, when chaotic dynamics parameter m is used as the number of input layer modes, and prediction experiment was made with an engineer example.(5)EMD method can decompose the non-linear data to a series components from the high-frequency to low-frequency, and the signal stationary be significantly improved. The EMD multi-scale prediction model was put forward based on the multi-scale characteristics of non-linear time series, and had high prediction accuracy. The integration deformation prediction model of EMD multi-scale decomposition and SVM was researched and established, which was suitable to process non-linear deformation time series.(6)The chaotic theory and EMD analysis technology were introduced to process deformation data, make deformation prediction and reconstruct dynamic process of ground surface settlement of old goaf. Studying and analyzing on the reclamation field of old goaf in Yanzhou coal mine show that ground surface deformation of the old goaf has chaotic characteristics. In this paper, EMD technique was first introduced to analyze deformation data. Experimental results show that EMD analysis technology is an effective method to research and judge the stability of the ground surface of old goaf.