论文部分内容阅读
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 weights / thresholds of BP neural network. The SMA neural network constitutive model was established. The optimal configuration of SMA was applied to a spatial structure. The Newmark-β algorithm program was used to solve the structural dynamic response. Shaker test results were compared. The results show that compared with the un-optimized constitutive curve of SMA neural network, the optimized constitutive curve can better predict the hypersurvival resilience of SMA under repeated loading. It is a high stability rate-dependent dynamic model Model. After the vibration control of the SMA wire with optimized configuration, the peak simulation results of seismic response of structures are in good agreement with the experimental results, and are effectively suppressed. The applicability of SMA neural network constitutive model and the simulation of SMA passive control by MATLAB are feasible Sex.