摘要
利用Hilbert变换在计算原时间序列瞬态属性方面的特性 ,把冲击应力波的时域信号变为连续的瞬态时序信号 ,易于提取与冲击定位相关的特征信息训练人工神经网络 ,实现冲击位置的自动识别。实验表明 ,以应力波信号的起始时间、峰值以及到达峰值的时间为特征向量具有较高的冲击定位精度 ,该方法可用于结构低速。
The signals of impact stress wave in time domain are transformed into the analytic ones with the features of momentary sequential time series using Hilbert transform. Neural networks are established to take the different characteristics out of these signals for automatic identification of impact location. Experimental results show the impact location method, which makes use of the peak value and the starting time of the envelope as features, has high precision. The method can be applied to the health monitoring of structural damage caused by low velocity impacts.
出处
《振动与冲击》
EI
CSCD
北大核心
2004年第3期124-127,共4页
Journal of Vibration and Shock
基金
国家自然科学基金项目 (编号 :1 0 0 72 0 2 6和 50 1 350 30 )
