基于神经网络技术的结构损伤探测

Detection of structural damage using neural network technique

理论分析表明 ,工程结构损伤前后的固有频率的变化包含了结构损伤位置和程度的信息 .在此理论基础上 ,构造了改进型 BP神经网络的输入参数 .分别对一个框架模型和一个桁架模型进行了损伤数值模拟计算 .首先提取结构固有频率的变化 ,对神经网络进行训练 ;然后分别对结构的损伤位置和损伤程度进行识别 .计算分析结果表明 ,该方法在结构损伤检测中具有较好的识别效果 .

On the basis of a theory that the variation of natural frequency can be employed to determine the locations and degrees of damage of structures, a BP artificial neural network technique with an improved input structure is developed. The positions and degrees of damage are detected by a BP network trained with the variations of natural frequencies of structures. Structural damage based numerical simulation, including a frame and a truss, gives satisfactory results.