基于SMA神经网络本构模型的结构地震响应控制

Structural Seismic Response Control Based on SMA Neural Network Constitutive Model

利用遗传算法优化BP神经网络初始权/阀值,建立SMA神经网络本构模型,并将优化配置后的SMA应用到一空间杆系结构,通过MATLAB编写Newmark-β算法程序求解结构动力反应,与振动台试验结果进行对比。结果表明,相比未优化的SMA神经网络本构曲线,优化后本构曲线能更好地预测SMA在反复荷载作用下的超弹性恢复力,是一种稳定性较高的速率相关型动态本构模型。应用优化配置的SMA丝进行振动控制后,结构地震反应峰值仿真结果与试验结果基本吻合,且得到有效地抑制,验证了SMA神经网络本构模型的适用性和采用MATLAB进行SMA被动控制仿真的可行性。

The genetic algorithm was used to optimize the initial weight/threshold of BP neural network. The SMA (shape memdry alloy)neural network constitutive model was established, and the SMA optimal allocation was applied to a space truss structure. A Newmark-βalgorithm program was written by means of MATLAB to solve the dynamic response of the structure. The output of computation was compared with the results of shaking table test. Results show that compared with the SMA neural network constitutive curve without optimization, the optimized constitutive curve can precisely predict the superelastic restoring force of the SMA under cyclic loading. It is a rate-dependent dynamic constitutive model with high stability. After applying optimal allocation of the SMA, simulation results and experimental results of structural seismic response peaks are basically consistent mutually, and the seismic response peak has been effectively suppressed. Thus, the applicability of the SMA neural network constitutive model and the feasibility of the SMA passive control simulation based on MATLAB were verified.