基于Sym小波与BP神经网络的装配柱钢筋套筒灌浆连接缺陷检测方法

A Damage Detection Method for Grout Sleeve Splicingat Assembly Column Based on the Sym Wavelet and BP Neural Network

为检测装配柱钢筋套筒灌浆连接由于施工原因而产生的缺陷程度,采用ABAQUS建立有限元模型,利用Sym8小波基对装配柱加速度响应时程曲线进行小波分解,提取各个频带成分的能量占比变化量。构造特征向量作为BP神经网络的输入向量,定义弹性损伤系数与塑性损伤系数,分别作为神经网络的输出值。通过数值模拟的方法获得多种损伤工况下的训练样本用以训练网络,建立特征向量和装配柱套筒连接段等效弹性损伤与塑性损伤的对应关系。最后,通过有限元模型获得一系列测试样本用以检测网络的识别能力。结果表明,训练得到的神经网络可以对装配柱钢筋套筒灌浆连接的等效损伤程度进行识别,为实际工程中的检测提供了参考。

In order to detect the damage degree of grout sleeve splicing at assembly column connecting section caused by construction reasons,in this paper,a finite element model was established using ABAQUS.The acceleration time-history response signal of assembly column was decomposed by Sym8 wavelet.The energy proportion variation of each frequency band component was extracted.Theeigenvector was constructed as the input vector of BP neural network.The elastic and plastic damage coefficients were defined asthe output values of the neural network respectively.In order to train the network,training samples under various damage conditions were obtained by the means of numerical simulation.Meanwhile,the corresponding relationship between the input eigenvector and the equivalentelastic or plastic damage of grout sleeve splicing was established.Finally,a series of test samples were obtained through the finite element model to test the recognition ability ofnetworks.The results showed that the neural network could be used to identify the equivalentdamage degree ofgrout sleeve splicingand provided reference for the test in practical engineerings.