改进U-Net模型的A356合金铸造轮毂内部缺陷检测

Internal Defect Detection of A356 Alloy Casting Hub Based on Improved U-Net Model

A356合金轮毂在低压铸造生产过程中,容易出现气孔、缩孔等内部缺陷,需要使用X射线无损探伤设备来检测。轮毂X射线检测的核心问题是缺陷自动判定,引入深度卷积神经网络中的U-Net模型作为框架,结合VGG模型改进编码器和解码器,在构建的数据库上展开缺陷自动识别研究。在准确定义轮毂缺陷检测识别率、误判率和漏判率的基础上,发现改进U-Net模型的性能与原始U-Net模型相比明显提升,缺陷识别率达到96.15%,满足实际检测需求。

In the process of low-pressure casting,A356 alloy hubs easily generate internal defects such as gas hole and shrinkage parosities,which need to be detected online by X-ray nondestructive testing.The key problem of hub X-ray detection is the automatic determination of defects.The U-Net model in deep convolution neural network is introduced as framework,and the encoder and decoder are improved combined with VGG model.The automatic identification of internal defects was carried out on the constructed database.On the basis of accurately defining the recognition rate,misjudgment rate and missed detection rate,it is found that the performance of the improved U-Net model is significantly improved compared with the original U-Net model,and the defect recognition rate reaches 96.15%,meeting the actual detection requirements.