Research On Dam Foundation Uplift Pressure Monitoring Model Of The Concrete Dam Based On One-to-one Connection BP Neural Network

The dam foundation uplift pressure was one of the most important physical monitoring factors for concrete dam safety.Analyzing dam foundation uplift pressure monitoring data was conducive to grasp seepage condition of the dam foundation and investigate working performance of the impervious curtain and drainage system as well as the dam body's stability.Firstly,the paper established two conventional dam foundation uplift pressure monitoring models—the stepwise regression analysis model and the BP neural network model.The stepwise regression analysis model had simple principles and was convenient to operate,and moreover,it could perform factor separation so as to quantitatively study influence of each factor on the uplift pressure.However,this model failed to simulate a nonlinear system and it had poor accuracy performance of uplift pressure prediction.The BP neural network model implemented a highly nonlinear mapping between the input and output and its prediction accuracy usually could meet the requirements.But it was impossible for it to carry out quantitative research about influence of each factor on the uplift pressure.At the same time,the input factors of conventional models were diverse and numerous.On this basis,a one-to-one connected BP neural network dam foundation uplift pressure monitoring model was proposed.From the two aspects of network structure and input factors,this model changed the BP neural network.(1)The one-to-one connected BP neural network model changed neurons' connection mode between the input layer and the hidden layer from original full connection to one-to-one connection and still kept the full connection between the hidden layer and the output layer.After changing the network structure,the output of the model became the linear summation of the influence of each factor.,which was the fundamental reason why the model could differentiate the influence of input factors.The one-to-one connected BP neural network model could not only achieve nonlinear mapping between the output and input factors,but also separate factors so as to quantitatively study influence of each factor on the uplift pressure.Meanwhile,the change of the network structure made the number of neurons in the hidden layer equal to the number of input factors,which overcame the problem that the number of neurons in the hidden layer of the BP neural network was difficult to determine.(2)The one-to-one connected BP neural network model adopted a new input factor selection method.Considering the continuity of monitoring data over time,the previous water level monitoring value of the pressure measuring hole tube,the corresponding upstream changed water level value,the corresponding downstream changed water level value,the corresponding temperature changed value and the rainfall data in corresponding time serial were used to predict the later tube water level value.The number of factors was reduced to 5 and its content was fixed.The one-to-one connected BP neural network model controlled the number of input factors and avoided content's diversification.Using Danjiangkou dam foundation uplift pressure monitoring data,the paper established the stepwise regression analysis model,the BP neural network model and ?the one-to-one connected BP neural network model in R.The calculation results showed that the factor separation results of the one-to-one connected BP neural network model were reliable and it had the most accurate performance than other two models.The prediction accuracy of BP neural network model was the second and that of stepwise regression analysis model was the lowest.Finally,the reason of better performance of one-to-one connected BP neural network model was analyzed.It was found that the change of network structure would not reduce the prediction accuracy,while the change of input factors increased the model accuracy greatly.However,the above two changes both could reduce the number of parameters in the model which effectively improved the model calculation efficiency.Therefore,using this model to analyze dam foundation uplift pressure data was effective and feasible,which could provide scientific basis for the safety operation of dam.