基于深度学习的发动机叶片故障检测技术

为了解决航空发动机叶片故障检测中存在的检测精度欠佳、检测效率不高的问题,提出了一种基于深度学习的目标检测方法。针对小样本数据集检测精度低、模型训练速度慢等问题,对Faster R-CNN目标检测算法进行结构优化,引入Res2Net结构,通过分割串联的策略强化残差模块的卷积学习能力,搭建了细粒级的多尺度残差模型Res2Net-50,以提升模型的特征提取能力。同时,在网络的训练过程中,采用多次余弦退火衰减法对学习率进行调整,以加快模型的训练速度,提升模型的训练质量。针对航空发动机叶片裂纹和缺损2种故障类型进行网络训练与检测试验,试验结果表明:优化后的模型识别准确率提高了0.7%,模型的平均检测精度提高了1.8%,训练时间缩短了5.56%,取得了比较好的检测效果。

Fault Detection Technology of Engine Blade Based on Deep Learning

In order to solve the problems of poor detection accuracy and low detection efficiency in engine blade fault detection,an aeroengine blade fault detection method was used based on deep learning.Aiming at the problems of low detection accuracy and slow model training speed of small sample data set,the structure of Faster R-CNN object detection algorithm was optimized.Res2Net structure was introduced to strengthen the convolution learning ability of the residual module through the strategy of segmentation and series,and a finegrained multi-scale residual model Res2Net-50 was built to improve the feature extraction ability of the model.At the same time,in the training process of network,the multiple cosine annealing attenuation method was used to adjust the learning rate,so as to accelerate the training speed of the model and improve the training quality of the model.Network training and detection tests were carried out for two fault types of aeroengine blade cracks and defects.The test results show that the recognition accuracy of the optimized model is improved by 0.7%,the mean Average Precision(mAP)of the model is improved by 1.8%,and the training time is shortened by 5.56%.