基于小波神经网络的RC梁损伤识别试验研究

简要阐述了小波包分析及BP神经网络理论，利用小波包变换对获得的加速度信号进行分解和重构，求解各频带内的信号能量，将其作为神经网络输入参数，对神经网络进行训练，然后利用该网络进行损伤检测。按上述方法，通过有限元分析对RC梁进行损伤模拟，建立了结构损伤识别神经网络。按同样的构造制作了RC试件，并进行了损伤试验研究，对试验中在不同损伤情况下采集的加速度信号进行小波包分解和重构，将得到的能量向量输入已建网络判断结构的损伤。从试验结果可以看出诊断误差很小，能够满足实际工程要求。

Experimental Research on Damage Diagnosis for Reinforced Concrete Beam Based on Wavelet Analysis and Neural Network

The theories of wavelet packet analysis and BP network are summarized in this paper.The signals of acceleration which gained in tests are decomposed and reconstructed by using wavelet packet transform and energy of frequency bands can be gained.It is used as the input parameter for the neural net.Then the neural net is trained and used to detect structure damages.The damage states of a reinforced concrete beam are simulated by the finite element method,and the neural net is constructed.Damage tests are performed in laboratory for a beam with the same constructional detail.For different kinds of damage,wavelet packet transform is used to decompose and reconstruct the signal of acceleration which gained in tests.Vector of energy is input to the network to detect damage.The results prove that the method can detect structural damages effectively.