Research On The Diagnosing Method Of Working Situation For Concrete Gravity Dam With Fusing Multiple Spatio-temporal Information

The establishing of monitoring model and diagnosing of working situation for concrete gravity dam is the most significant content in the dam safety monitoring. However, many shortages are existed in the sides of the accuracy for the monitoring model and diagnosing methods. According to the problems of the currency establishing model methods and diagnosing methods, the passage starts with the constructing fusing diagnosing system, combining with statistical theory、finite element theory、intelligent algorithm、principal component analyzing theory and cloud model. The methods for establishing methods for different models are researched, the optimizing inversing methods for concrete dams’ parameter are explored. Through the analyzing of fitting residual, the cloud model for diagnosing working situation with multiple indicators is constructed. The main researching content are as follows:(1) The selecting principles and evaluating collections are established, which are essential for fusing diagnosing. According to a practical engineering, the fusing diagnosing system are established.(2) The characters of time series and relative methods for parameter estimating are researched. The efficiency of auto-regressive and moving average model is proved with the higher accuracy than normal model of an example.(3) The constructing method of hybrid model for concrete gravity dam is discussed, The optimizing methods of using neural network to fit water pressure component and using neural network and stepwise to establish combination model are promoted. The optimizing inversing methods under the condition of multiple measuring point for displacement is promoted.(4) On the basis of the established models, the distribution regular of fitting residual is researched. The diagnosing cloud model of one measuring point is established. With the using of principal component analysis to extract the weight, the establishing method of multiple dimension cloud model, which is suitable for multiple indicators is promoted.