Nonlinear Analysis Model For Safety Monitoring Of Large Dam Dynamic System

In consideration of the imperfection in dam safety monitoring, the dam and slope are regarded as a dynamic system, the nonlinear science such as chaos theory, neural network and wavelet analysis is introduced to dam safety monitoring, and related models and methods are proposed hi this dissertation. The main contents are as follows.(1)According to chaos theory, the phase space reconstruction of time series is performed. The chaotic invariants of measured time series of dams such as correlation dimension, the Lyapunov exponent and the Kolmogorov entropy, are calculated. The results show that the dam observation data is a chaotic time series.(2) Based on the phase space model of the zeroth and first order approximation, a regression-autoregression model and a regression-local linear regression model are proposed, which provided new method for dam safety monitoring.(3) Based on the study of several kinds of neural network, several phase space models of dam observation time series using neural network are established. The validity of the models is proved by an example. A neural network method for determining the component percentage is given.(4) The theory and methods of wavelet analysis are introduced to the observation data analysis for dam safety, and the methods for detecting outliers, abstracting trend component, and denoising and detecting missing data by use of wavelet is developed. Examples show that the methods are effective.(5) Based on the variation principle, the energy criterion of instability is established. The strain softening of dam body and foundation is the necessary condition for dam instability. According to observation data of dam cracks, a gray cusp catastrophe model is established to judge the stability of the crack. According to the slope observation data of a certain hydropower plant, a nonlinear dynamic model is developed by means of reversion and the Lyapunov exponent spectrum, and the Lyapunov information dimension are calculated for judgement of the stability of the slope.