基于水浸超声技术的涡轮叶片裂纹损伤检测方法研究OA

To address the difficulty in detecting surface micro-crack damage on coated turbine blades of a certain aero-engine,a crack detection method is proposed based on a one-dimensional convolutional neural network(1D-CNN)and water immersion ultrasonic testing.By analyzing the characteristics of abnormal signals during detection,a novel defect detection network is constructed using a 1D-CNN architecture that incorporates all feature mappings,along with an adaptive defect feature extraction module tailored for water immersion ultrasonic signals of turbine blades.To overcome the scarcity of real damage samples in practical detection scenarios,a data augmentation approach is employed to generate synthetic samples,thereby improving the detection accuracy of the network model for turbine blade inspection.Specialized tooling for turbine blades is developed,and an experimental platform for water immersion ultrasonic testing is established to validate and optimize the proposed detection model.Experimental results demonstrate that the proposed method achieves a detection accuracy of 96.8％for turbine blade damage,outperforming conventional detection models.This provides a novel solution for detecting surface micro-cracks in coated turbine blades.