STRUCTURAL PARAMETER IDENTIFICATION APPROACH BASED ON ARMA AND NEURAL NETWORK MODEL
Neural network can act as a nonparametric model of a structure by forecasting structural dynamic responses according to them in the past consequent time steps and neural networks-based structural parametric identification method using vibration measurements without any mode shapes and frequency extraction have been proposed and validated with numerical simulation in recent years.As another time-domain method, the auto-regressive and moving average (ARMA) model has been proposed employed to identify structure parameters. Because the neural network model and the ARMA model have the same physical meaning and the matrices in the ARMA model are fully determined by structural mass, stiffness and damping matrices, it a possible to determine structural matrices by the weights and thresholds of the neural network model of the structure. In this study, a novel structural parameters identification methodology by matching ARMA and neural network model was proposed and the accuracy and efficacy of the proposed strategy were validated with numerical simulation for a multi-storey frame structure. Simulation results show that the proposed methodology presents a new way for structural parameter identification using structural response measurement and excitation time series only.
BP neural network ARMA Time series Multi-storey frame structure Dynamic response
An-Su Gong Bin Xu
College of Civil Engineering, Hunan University, Changsha 410082, P.R.China College of Civil Engineering, Hunan University, Changsha 410082, P.R.China Key Laboratory of Buildin
国际会议
第十届国际结构工程青年学者研讨会(The Tenth International Symposium on Structural Engineering for Young Experts)
长沙
英文
1522-1527
2008-10-19（万方平台首次上网日期，不代表论文的发表时间）