Research On Dam Safety Monitoring System Based On Artificial Neural Networks

The security state of dam not only influences the exertion of economic benefits of projects directly, but concerns the safety of life and property of downstream residents and even the stability of the whole society. Therefore, it is very significant to exploit a practical and credible dam safety monitoring system for enterprise, society and the broad masses of the people.Combined with the monitoring projects of the Black River, the Stone River, the Shibianyu River and the Jianyu River, based on analyzing and summarizing the commonness and individuality of these dams, compared with statistic model, researches were carried out on Artificial Neural Networks model and its use in the dam safety analysis. Besides, a dam safety monitoring system was developed, which increases the speed and facility of dam safety analysis greatly.Major contents and findings are as follows:(1) Based on the analysis of monitored data of dam and common arithmetic, a data pretreatment arithmetic was established which could identify and reorganize the data real-timely. The testes indicate that it has great adaptability and perfect treatment effect.(2) The basis of the factor chooses for statistical model of dam monitoring was simply analyzed, as well as its limitation of building model.(3) The part performances of RBF Neural Networks were tested and analyzed theoretically. Based on the characteristics of RBFNN, a celerity off-line and on-line training arithmetic which could add nodes automatically was designed and the test indicates that this arithmetic enhances the velocity and study ability greatly compared with others.(4) The tests on RBF and statistical model in different conditions indicate that the former is insensitive to the factor correlativity of input-data and used simply and conveniently.(5) By using the advanced developing platform and tools, the dam safety monitoring system was developed based on B/S frame. The system has the characters of perfect function, friendly interface, good effectiveness and high automatic degree. In addition, the powerful module structure makes the system have a good reuse and expansion characteristic.