基于神经网络的柔性结构损伤模式识别方法

Identification of Flexible Structure Damage Modes Based on Artificial Neural Network(ANN)

提出一种基于人工神经网络 ( ANN—— Artificial Neural Network)的柔性结构损伤模式识别方法 ,将神经网络用做损伤模式的分类器。引入通过对输入模式的增强来实现多种模式分类的函数连接型神经网络构成损伤模式识别方法。通过对柔性悬臂梁结构的损伤识别实验表明 ,该方法运用于模式较少的场合非常有效 ,并且网络结构简单 ,学习速度快。实验中采用 PVDF压电薄膜作为信号获取元件 ,并且以柔性结构多个位置传感器的输出值的相对值作为模式识别中的特征向量 ,识别时对梁的振动幅度没有固定要求。实验结果表明 。

An identification method for flexible structure damage modes based on Artificial Neural Network (ANN) is put forward, where the ANN is used as a damage modes classifer. In the work, a function linked type ANN with strengthered input modes is adopted. Experimental results of flexible cantilever beams show that the method is quite effective in case of the structure with a few damage modes, and the network is noted for its simple structure and rapid learning rate. In the experiment, the PVDF piezoelectric films are used as sensing elements, and because the relative output values of the sensors fixed at the surface of the structure are taken as the input characteristic vector, so the classification is not related to the absolute value of vibration amplitudes.