Wind Load Time-history Simulation Study Based On Neural Network

In the time domain of structure of the wind-induced vibration analysis, the primary issue is to determine the structure of the wind load. But relying solely on the existing record and observational data as input load due to many conditions often can not meet actual needs of the project, therefore, need to develop the artificial wind load simulation methods.However, the existing simulation method to conduct a large number of iterative and recursive computation, especially when more points simulation, a time-consuming and relatively low efficiency.In this paper, using the existing research results and artificial neural network theory of the characteristics, on the basis of use of scientific computing software of Matlab, exploration and research on the use of neural network artificial simulation of wind loads:(1) Realized the simulation of wind loads model based on neural network, the simulation results better, and have higher efficiency.(2) Discussed the neural network simulation modeling the type of network, the number of neurons, transfer function, the training sample sets and training algorithm, and the corresponding solutions, detailed analysis of the neural network model in the Matlab neural network toolbox of methods and steps.(3) Doing a detailed comparison and analysis on BP neural network training algorithms different results, identify the most effective training methods.(4) Through two examples, verified the validity of the model.Neural network simulation is essentially matrix operations of input vector with the network-weight-matrix and the threshold-vector, therefore, does not require iterative and recursive computation, and can greatly reduce time-consuming. The results of this study also proved that the use of artificial neural network method can be an effective solution artificial simulation of wind loads, and with high efficiency.