基于超声波法的混凝土内部缺陷智能检测

Intelligent detection of internal defects in concrete based on ultrasonic method

混凝土的内部缺陷严重影响结构的耐久性和安全性。为此,本文提出了一种基于粒子群算法优化的支持向量机(PSO-SVM)模型。首先,对实验室混凝土试样的超声信号进行3层小波分解,得到各子波段的能量;然后,定义了模型的5个输入特征;最后,采用PSO-SVM模型对检测信号进行训练和预测,并与人工神经网络(ANN)模型和支持向量机(SVM)模型这2种传统模型的测试准确率进行了对比。结果表明:定义的5个输入特征能够有效反映缺陷信号与完整信号之间的差异;提出的混合优化模型能够准确识别和分类超声缺陷信号,准确率高达93.33%,优于传统的人工神经网络和支持向量机模型。

The internal defects in concrete seriously affect the durability and safety of the structure.Therefore,a support vector machine model based on particle swarm optimization is proposed.Firstly,the ultrasonic signal of concrete samples is decomposed by three layers of wavelet,and the energy of each sub-band is obtained.Then,five input features of the model are defined.Finally,the PSO-SVM model is used to train and predict the detection signal,and the test accuracy of two traditional models,artificial neural networks(ANN)model and support vector machine(SVM)model,is compared.The results show that the five defined input features can effectively reflect the difference between the defect signal and the complete signal.The proposed hybrid optimization model can accurately identify and classify the ultrasonic defect signals with an accuracy rate of 93.33%,which is better than the traditional artificial neural networks and support vector machine models.