基于Faster R-CNN的发动机叶片损伤识别方法研究

Research on Engine Blade Damage Identification Method Based on Fast R-CNN

航空发动机是航空飞行器中的重要组成部件,它的机械结构非常复杂,可以产生强大的动力支持飞行器的运行。发动机里面的金属叶片又是整个发动机最核心的部分,叶片将汽油尾气排放出去的同时要负责将飞机外部的空气吸入发动机的燃烧室,一直处于高温、腐蚀性气体的环境中,极易产生裂纹、边缘刻口、表皮脱落等损伤。本文从传统图像处理的知识和卷积神经网络的知识入手,设计并实现了一种基于Faster R-CNN模型的发动机叶片损伤识别算法,能够实现裂纹、边缘刻口、表皮脱落等损伤识别,实验结果表明,该算法识别精度较高,具有良好的鲁棒性。

Aeroengine is an important component of aircraft. Its mechanical structure is very complex and can produce strong power to support the operation of aircraft. The metal blades, which is the core part of the whole engine, are responsible for sucking the air outside the aircraft into the combustion chamber of the engine while discharging the gasoline exhaust. The blade inside the engine is prone to cracks, edge notches, skin peeling and other damages. This paper designs and implements an engine blade damage identification algorithm based on fast R-CNN model, starting with the knowledge of traditional image processing and convolutional neural network. The experimental results show the proposed algorithm can identify cracks, edge notches and skin peeling with high identification accuracy and good robustness.