泵塔区钢结构焊缝表面缺陷的超声红外热成像检测

Ultrasonic infrared thermal imaging detection of weld surface defects of steel structure in pump tower area

利用超声红外热成像装置采集焊缝图像,通过灰度化、消噪以及增强对焊缝图像进行预处理,利用卷积神经网络提取图像的特征向量,并用决策树多分类器模型对缺陷进行分类识别。试验结果表明,该检测技术很好地完成了泵塔区钢结构焊缝表面缺陷的无损检测,检测精度较高,为及时修复泵塔区的钢结构焊缝缺陷提供了条件。

The ultrasonic infrared thermal imaging device is used to collect the weld image, and the weld image is preprocessed by grayscale, denoising and enhancement. Then the convolution neural network is used to extract the feature vector of the image, and the decision tree multi classifier model is used to classify the defects. The test results show that the detection technology can complete the nondestructive detection of the weld surface defects of the steel structure in the pump tower area, and the detection accuracy is high, which provides the condition for repairing the weld defects of the steel structure in the pump tower area in time.