论文部分内容阅读
提出了一种结合经验模态分解(EMD)与极限学习机(ELM)的时间序列多步预测方法,对近断层强震作用下弹塑性高层框架结构的顶层加速度和位移响应时程进行了多步预测。首先,利用经验模态分解技术将高层结构非线性、非平稳地震响应分解为一系列具有不同特征尺度的固有模态函数序列(IMFs)。然后,利用极限学习机分别对固有模态函数子序列进行多步预测,再将各子序列的预测值叠加得到最终的预测值。预测结果表明,EMD-ELM预测方法能够高精度地实现强震作用下高层建筑动力响应的多步预测。建筑结构地震响应时程的短期预测可为主动、半主动控制系统预先提供准确的动力响应,从而有利于实现工程结构的在线实时减震控制。
A time series multi-step prediction method based on Empirical Mode Decomposition (EMD) and Extreme Learning Machine (ELM) is proposed. The top acceleration and displacement response time history of elasto-plastic high-rise frame structure subjected to near-fault strong earthquake Step prediction. First, the empirical mode decomposition technique is used to decompose the nonlinear and non-stationary seismic responses of high-rise structures into a series of intrinsic mode function sequences (IMFs) with different characteristic scales. Then, using the limit learning machine, the predictive value of each sub-sequence is superposed to obtain the final predictive value by multi-step prediction of the intrinsic mode function sub-series. The prediction results show that the EMD-ELM prediction method can accurately predict the multi-step dynamic response of tall buildings under the action of strong earthquakes. The short-term prediction of the seismic response time history of the building structure can provide an accurate and dynamic response to the active and semi-active control systems in advance, which is in favor of real-time on-line damping control of the engineering structure.