The lightweight YOLOv8n-based algorithm for defect detection in aircraft engine turbine blades

Aircraft engines play a vital role in ensuring the safety of aircraft. In order to solve the problems of low efficiency and low accuracy of traditional pipeline mirror inspection method, this paper improved the method based on YOLOv8n, and proposed a lightweight algorithm of CALP-YOLO aero engine turbine blade defect detection. Firstly, CSPHet network is used instead of C2f network to improve lightweight performance and reduce model size and calculation requirements. Secondly, ADown subsampling module is introduced into the backbone network to enhance the feature extraction capability. Then, shared convolutional LSDECD is used to improve the detection header and enhance the multi-dimensional feature processing capability of the network. Finally, PIoU2, which is highly sensitive to the size and position of the boundary frame, is used as a loss function to improve the detection ability of the model in the case of turbine blade occlusion. The experimental results show that compared with YOLOv8n, CALP-YOLO reduces the number of parameters and the amount of computation by 50.4% and 44.5%, respectively. Recall rates and [email protected] increased by 3.6% and 1.0% to 97.8% and 99.3%, respectively. This model is lightweight and has excellent detection performance, which can meet the deployment requirements in the environment with limited computing resources, and has important engineering application value.