摘要
本文以Easydl深度学习平台为基础,对大型锻件热锻过程中的锻造裂纹检测问题进行研究,结果表明采用Faster_R-CNN-ResNet50-FPN深度学习模型能够获得较好的识别结果,对于一般性锻造裂纹的检测准确率可以达到77%以上,而对于热锻典型环境的裂纹检测准确率能够达到约90%。
In this paper,based on the Easydl depth learning platform,the forging crack detection problem in the hot forging process of large forgings have been studied.The results show that the Faster_R-CNN-ResNet50-FPN depth learning model can obtain better recognition results,and the detection accuracy for general forging cracks can reach more than 77%,while for typical hot forging environment,the detection accuracy can reach about 90%.
出处
《大型铸锻件》
2020年第5期47-50,共4页
Heavy Casting and Forging
关键词
深度学习
锻造裂纹
裂纹检测
物体检测
deep learning
forging cracks
crack detection
object detection