Research On The Key Technologies Of Dam Deformation Monitoring Multisource Data Acquiring And Processing In Vietnam

Dams,in their many forms,bring enormous benefits to many aspects in our lives;thus they have been installed in many countries all over the world.Along with the increase in the number,size and types of dams,those countries that have installed dams have attached special importance to the quality of that dam's construction to ensure the safety not only of their project but also those communities that live downstream.Therefore,deformation monitoring as an aspect of preserving and extending the life of a dam is a very important task in the operational phase management of dams and as those dams increase in size so too does deformation monitoring increase in importance.First,an overview of the overseas and domestic status of dam deformation monitoring is briefly introduced for the purpose of outlining their importance and application by others within this field.Current existing methods and instruments of dam deformation monitoring in the world are studied.The methods and corresponding instruments are classified into two groups:Geotechnical and Geodetic and they are followed by the comparative advantages and disadvantages of these two groups on the basis of suitability for use in Vietnam.The regulations and specifications of dam deformation monitoring within Vietnam and other countries,such as the USA and China are considered.Based upon a comparison of these countries,some additions and/or amendments to the current regulations and specifications of dam deformation monitoring in Vietnam are proposed.For the gap in analyzing and forecasting dam deformation,various types of models are studied.In the case of statistical model application,the Seasonal ARIMA model has great potential for analysis but despite this it has not been used in previous studies of dam deformation;whereas the SARIMA model has been exhaustively studied and practical application has shown that it is quite suitable for dam deformation analysis and forecast.The Multiple Linear Regression(MLR)model,which has the main impact factors as reservoir water level,air temperature and aging component of dam,are well established.The experiment shows that SARIMA model analysis has the more precision in comparing to the MLR model.In the case of artificial intelligence techniques,the Multilayer Perceptron Neural Network(MLPs)with Back-propagation algorithm is applied to modeling the horizontal displacement of dam.When the MLR model or SARIMA model is combined with MLPs,the forecast precision can be clearly improved.In addition,a novel hybrid artificial intelligence approach namely PSO-FIS for modeling and forecasting the horizontal displacement of earth-rockfill dams is proposed based on Neuron-Fuzzy Inference System and using Particle swam optimization to search the best parameters of model.Experimental results show that PSO-FIS merging model performed well on both the training and validation datasets.Furthermore,the comparison of four benchmark models shows that PSO-FIS is a promising tool for modeling and forecasting horizontal displacement of earth-rockfill dam.The experiments of this research are implemented using the monitoring data of an earth-rockfill dam in Vietnam named Hoa Binh.It is a hydropower dam that was constructed in the years from 1981 to 1990.Data set covers a period of 11 years,from September 1999 to December 2010 with 131 samples,and consists of reservoir water level,air temperatures and horizontal displacement.Novelties of the dissertation are following:i.Focusing on the current and relevant Vietnamese National Standards and the criteria applied in the People's Republic of China for dam deformation monitoring,compares one with the other to derive a conclusion.Based upon the detailed comparisons,some additions and/or changes should be incorporated into the Standards which are in current use in Vietnam.ii.One kind of statistical model,namely Seasonal ARIMA model,has been built and utilized for analyzing and predicting dam deformation;this model being derived from unique deformation data.iii.The Multilayer Perceptron Neural Network with Back-propagation is implemented for modeling the dam deformation although it has some limitations;but it has been found useful for combining to Multiple Linear Regression model or Seasonal ARIMA model because of defimite improvements in the prediction accuracy.iv.A powerful tool based on ANFIS and PSO has been proposed,which has acceptance of modeling and forecasting abilities.Four benchmark models(Support vector regression,MLP Neural Nets,Gaussian Processes,and DE-FIS)have been used for comparison and to confirm the prediction accuracy of the proposed approach.