Researh On Dynamic Model Modification Method Of Arch Dam Based On Intelligent Algorithm

The revision of the finite element dynamic model of the concrete arch dam has important applications in seismic analysis and health diagnosis of structures.In order to monitor the health status of the structure in time and ensure the safety of the structure,it is required to continuously improve the accuracy and calculation speed of the model modification method.The revision of the arch dam finite element dynamic model is based on active monitoring.The acquisition of active monitoring information data is often unable to fully obtain structural dynamic characteristic information due to the unreasonable arrangement of sensors.It is necessary to optimize the layout.Structural modal identification based on vibration monitoring is another key step in model modification.Among the various modal recognition methods,the modal recognition based on the optimization method has the characteristics of high accuracy and robustness,but it also has the disadvantage that it is easy to fall into local optimal and premature convergence for multi-degree-of-freedom systems.The main research contents are as follows:(1)Aiming at the contradiction between maximizing the norm of Fisher information matrix and minimizing the condition number of mode shape matrix considering model errors;a norm based on Fisher information matrix combined with the condition number of mode shape matrix and modal strain energy(MSE),Multi-objective hybrid optimization dynamic sensor layout method(Shape-MSE-Fisher-MAC)of modal confidence(MAC)optimization criteria,through the numerical model of arch dam,the schemes under different optimization layout criteria are studied for different measurement indicators and models The influence of the state recognition results,the analysis results show that the proposed multi-objective hybrid optimal layout method of dynamic sensors is reasonable and effective.(2)For the intelligent optimization algorithm to identify structural multi-degree-of-freedom modes,it is easy to prematurely converge and fall into the local optimal problem.Combined with quantum particle swarm optimization algorithm(QPSO),the use of prior knowledge of the system to narrow the search space and change Partial modal decomposition(VMD)two intelligent optimization modal parameter identification methods that reduce the search space dimension;combined with examples,through comparison with traditional methods,the analysis shows that the two methods effectively overcome the premature convergence and local optimal problems,and the identification results The accuracy is higher than the traditional method,the frequency and damping ratio are high,and the accuracy of the mode identification is low,but it can meet the needs.(3)Aiming at the problem that the traditional proxy model direct analysis method has lower precision than the indirect analysis method under the condition of low modal order and small sample number,a direct model correction method based on convolutional neural network(CNN)is proposed;for CNN The high requirements for the number of training samples are determined by comparing the accuracy of direct analysis and indirect analysis based on correlation vector machine(M-RVM),and then form a modified direct analysis method of CNN model(M-RVM-CNN)based on M-RVM extended data.Combining with the numerical model of arch dam,the analysis results show that the indirect analysis method of M-RVM algorithm is more suitable as a method for expanding the number of training samples.By comparing the correction accuracy of the model under different correction methods and different sensor layout schemes,the sensor optimization proposed in this paper is further verified The superiority of layout and M-RVM-CNN model correction method.