For the complex structure of the artillery rocker, the detection of internal defects in the weld seam is not ideal. The ultrason-ic phased array detection technology, which is also applicable to non-planar objects and complex structural bodies, was selected for the detection of defects in the weld seam of the rocker. The obtained ultrasonic phased array mapping was combined with the ResNeXt net-work model to achieve intelligent classification of weld defects. SK convolutional units were also introduced into the ResNeXt network model for the qualitative analysis of rocker weld defects. The improved network model shows 5. 5％ improvement in classification accu-racy compared to the original ResNeXt network, eventually reaching 98. 2％.
维普
万方数据