Study On Calculating Methods Of Dam Risk Analysis

Since the population of our country will maintain a large density and the economic will be growing rapidlly in the future, the dam failure will cause much greater loss of life and property, therefore the conflicts between water resources development and public security have become increasingly prominent. As the continuation and supplement of dam safety management, dam risk management has been paid more and more attentions worldwide. Both the safety of dam and the consequence in downstream area are considered to evaluate the risk of dam construction, and it could be an efficient tool to resolve the contradiction between water resources development and public security. The key issues of dam risk management are calculating the possibility of dam failure in a certain mode and evaluating the relevant loss. As a new developing discipline, there remain a number of problems in dam risk management, especially in aspect of quantitative analysis. Based on such background, this dissertation studied several calculating methods of dam risk analysis, and the main contents are summarized as follow:(1) Based on analytic hierarchy progress and improved particle swarm optimization algorithm, an identification model of dam operaing risk factors was established. According to the insufficiency of low accucary and unstable calculating process, the determination on priority weights of hierarchy elements and the consistence check of comparison matrix were attributed to nonlinear constrained optimization problem, which could be solved by improved particle swarm optimization algorithm. The proposed model was employed to recognize risk factors on dam safety, and it was shown that the identification model was efficient with accurate and stable results.(2) For explicit limit state function, a reliability index calculating method was presented according to its geometric meaning. The artificial bee colony algorithm was used to search the minimum distance from the limit state surface to the origin in standard Gaussian space and the reliability index would be obtained. Several testings verified that the proposed method was efficent and stable, especial for high order and complex limit state functions. For implicit limit state function, a failure probability calculating method of was proposed using least squares support vector machine. The results of finite element method were used as training samples to establish least squares support vector machine. Based on the trained learning machine, Monte Carlo simulation was run for computing the failure probability. The proposed method was used to calculate the cracking probability of gravity dam heel, and the results showed the accuracy and enfficiency of this method. (3) During the service period of dam, the structure resistance would deteriorate, and meantime the population and economic in downstream area would increase. Considering the above-mentioned aspects, a time dependent risk analysis model was proposed. On the assumption that the annual maximum water depth belongs in normal distribution, the stochastic processes of resistance and load were discretized into stochastic variables, and the time dependent dam danger degree could be obtained. Then the population and economic status in the downstream area were integrated into time dependent vulnerability. Finally, the time dependent risk could be calculated through mutiplying time dependent danger degree by time dependent vulnerability. The results could provide useful information to structure design and resevoir operation.(4) To solve the conflict between efficiency and precision in dam fragility analysis, a least squares support vector machine based fragility analysis method was presented. The numerical results were used to train least squares support vector machine, and the nonlinear relationship between dynamic load and the response of dam body could be established. Finally, the high accurate fragility curve could be obtained by Monte Carlo simulation based on the trained least squares support vector machine. Using cumulative sliding displacement as the indicator of dynamic stability, the fragility of gravity dam stability was analysed, and the results verified the high efficiency and precision of the proposed method.(5) Aim at the diefficiency of fuzzy c-means clustering algorithm, a new clustering algorithm was presented by combining the gradient information of fuzzy c-means clustering algorithm and the stochastic searching mode of artificial bee colony algorithm. Taking risk index as attribution, the risk level of barrier lakes and quake-damaged reservoirs caused by Wenchuan earthquake were classified the using the proposed clustering method. The examples showed that the presented method could provide scientific and reasonable information for rescue plans. Considering the affections of length of projection value vector, an improved projection pursuit model was presented. The projection values of dam danger and consequence were calculated to describe relative degree of dam danger and consequence respectively and their product was used to curve the relative degree of risk. The prioritization of barrier lakes caused by Wenchuan earthquake was obtained with this method, and the results showed that the proposed method could provide detailed and objective information for rescue plans.