Research on the Detection Method for Missing and Incorrect Installation of Aviation Lock Washers Based on Improved DeepLabV3+

To address the frequent issues of missing and incorrect installations in aviation lock washer assembly, a visionbased inspection method based on an improved DeepLabV3+ model is presented. Firstly, by systematically analyzing the assembly characteristics of hex-head bolt lock washers, the assembly defects are categorized into three typical types: structural absence, tail looseness, and claw looseness. Secondly, the model is optimized by introducing a weighted loss function, significantly enhancing the segmentation accuracy. By leveraging the high-precision mask images generated by the model, a determination method for detecting missing and incorrect installations is developed. Finally, experimental results demonstrate that the optimized DeepLabV3+ model achieves excellent segmentation performance, with MPA and MIoU reaching $96.25 \%$ and $92.75 \%$ respectively. Furthermore, the proposed determination method attains a $100 \%$ detection rate in identifying the three defect types in vertical-view images, effectively overcoming the low efficiency and high false-positive rates of traditional manual inspections. These findings provide an efficient and reliable solution for aviation lock washer quality inspection, offering significant engineering application value.