论文部分内容阅读
以表面贴有多片压电片元件(PZT)的焊接铝板为实验对象,通过阻抗分析仪测量PZT导纳信息,构建用遗传算法优化的神经网络,对焊缝裂纹进行定位分析。实验结果表明:PZT导纳谱能灵敏地反映耦合结构状态的微小变化;离裂纹越近的PZT其导纳谱变化越大;构建的遗传神经网络以导纳信息中表征结构健康状况的特征量作为输入向量,通过训练能够实现对焊接结构损伤的监测和定位,且性能优于未经优化的BP神经网络。
The welded aluminum plate with multiple piezoelectric elements (PZT) on the surface was used as the experimental object, the PZT admittance information was measured by impedance analyzer, and the neural network optimized by genetic algorithm was constructed to locate the weld cracks. The experimental results show that PZT admittance spectroscopy can sensitively reflect the small change of the coupling structure state, and the admittance spectrum of the PZT closer to the crack changes more. The constructed genetic neural network uses the admittance information to characterize the structural health status As an input vector, the training can be used to monitor and locate the damage of welded structures, and the performance is better than the un-optimized BP neural network.