In view of the lack of effective intelligent detection methods due to the complex causes of quality defects

the variety of defect features

and the high detection requirements in the development and production of aviation products

this paper first systematically reviews the research status of the intelligent detection technology of aviation equipments

and summarizes the ideas and implementation methods for the research of intelligent detection methods for this application scenario and specific defect characteristics. Secondly

an improved Mask R-CNN algorithm fused with global feature pyramid network is designed

and a digital radiographic detection defect feature dataset for aviation castings is constructed by using data augmentation techniques such as cutting

flipping

Overlap and Mosaic for aviation castings with complex defect features and high detection requirements. Finally

the improved algorithm and the constructed dataset are used to test and verify three types of defects in aviation castings

including porosity

cracks

and high-density inclusions. The experimental results show that the detection accuracy of the improved algorithm is 93.25% and the recall rate is 96.51%

having a good detection effect.