The Earth-rockfill Dam Safety Monitoring Model Based On Support Vector Machine And Fuzzy Evaluation Of Safety State

In recent years, hydraulic and hydropower projects are in the ascendant, especially in China's southwestern region with rich water resources. There are still some large-scale hydraulic and hydropower projects in the construction phase. With "the Belt and Road" strategy,more hydraulic and hydropower builders will go abroad, to help other countries. In order to play the huge benefits of the projects and ensure the smooth operation of the projects, dam safety is particularly important, especially for the majority of the earth-rockfill dam, while safety monitoring is a significant way to reflect the safety of the earth-rockfill dam and reduce the occurrence of engineering accidents.Building the earth-rockfill dam safety monitoring model to analysis the monitoring data and judging the earth-rockfill dam safety performance are important parts of safety monitoring work. Support vector machine (SVM), as a new method in data mining, can deal with non-linear, high-dimension and small sample problems with good generalization ability and be applied in dam safety monitoring area. But the selection of the model parameters is the key to the merits of the model performance. The particle swarm optimizition algorithm (PSO) and the artificial bee colony algorithm (ABC) are used to optimize the parameters. Finally, a fuzzy comprehensive evaluation model of the earth-rockfill dam safety state is established, and the seepage safety grade of an earth-rockfill dam is evaluated by combining the seepage monitoring data. The main contents and results as follows:(1) This paper expounds the principle, characteristics and parameters' influence of the support vector machine model. Based on the analysis of the relationship between the effect quality and the environmental variables of the earth-rockfill dam, a monitoring model based on support vector machine is established.(2) The theory, characteristics and parameters of the particle swarm optimizition algorithm and the artificial bee colony algorithm are introduced. In order to solve the problem that the model accuracy is not high while selecting model parameters by using the grid search method in SVM, the particle swarm optimization algorithm and artificial colony algorithm with global optimization capability are introduced in support vector machine to optimize the model and establish the PSO-SVM and ABC-SVM monitoring models of the earth-rockfill dam.(3) The PSO-SVM and ABC-SVM models are applied to the project examples. The simulation and prediction of the displacement and seepage monitoring data of an earth-rockfill dam are carried out and compared with the support vector machine model. The results show that the PSO-SVM and ABC-SVM models have higher accuracy and less error. And the ABC-SVM model has better effect than PSO-SVM model, indicating that ABC algorithm has stronger global optimization ability.(4) The fuzzy evaluation model for safety state of the earth-rockfill dam is established.First draw out the monitoring indicators, then use the fuzzy mathematics method to determine the membership degree matrix of monitoring data for each monitoring point, introducing the game theory to determine the weight of monitoring points, through the operation to obtain the results of the judgments, according to the principle of maximum membership to determine the earth-rockfill dam safety grade. The seepage monitoring measured data of an earth-rockfill dam are evaluated, and the conclusion of the seepage safety grade is in accordance with the dam safety check report.