基于神经网络技术的结构损伤检测

Structural Damage Detection Based on Neural Network Technique

结构损伤会导致其振动频率的变化,因此测量结构振动频率可以判断结构是否存在损伤,同时,基于频率测量的结构损伤识别方法具有测试简便,精度较好的优点。在分析结构固有频率的基础上,把结构损伤识别问题分为损伤辨别、损伤定位、损伤程度标定3个子模块,对每个子模块用模态参数构造对损伤敏感的标识量,并作为特征参数输入到神经网络中实现损伤识别。将BP网络和频率相结合进行了矩形梁的损伤检测,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。最后对神经网络方法在损伤检测中的发展前景作了展望。

Structure Damage Detection Method （SDDM） based on frequency measurement exhibits a feature that natural frequencies of a structure can be measured conveniently with relatively high precision. In this paper, for the first time structural damage detection is divided into three sub-modules., which are structural damage identification module, structural damage localization module and structural damage severity determination module. Damage detection by modal parameters needs to solve intricate mathematical iteration problem, which makes it difficult to realize real-time mapping ability, which can change inverts problem into forward problem. Based on this point, damage features formed by vibration modal parameters are inputted to neural network as parameters for structural health monitoring. Damages of different locations and severity on a beam of rectangular cross-section are simulated with optimized BP （back propagation） neural network. The results as given in conclusion show preliminarily that our method is feasible. Finally, prospect of using neural networks in structural damage detection is set forth.