基于深度学习的TA15钛合金显微组织检测

钛合金在航空制造中常用于飞机的主要承力构件,需要对其进行显微组织检测以保证材料性能。为克服钛合金显微组织图像难以自动检测的问题,文章以TA15钛合金为突破口,提出并详细说明了一种基于深度学习的显微组织检测方法,包括采集原始图片时应注意的问题、数据增强时旋转图片导致的纹理不一致问题及解决方法,提出了一种性能较好的神经网络结构以及考虑实际生产置信度的预测结果后处理与管线构建方法。最终实现了从原始图片到显微组织类别的端到端预测。

Titanium alloy is commonly used in aviation manufacturing as the main carrier component of aircraft,whose microstructure needs to be detected to ensure material properties.In order to overcome the problem that titanium alloy microstructural image is difficult to detect automatically,this paper takesTA15 alloy as anexemplar,puts forward and elaborates a detection method based on deep learning,including things that should be paid attention to when collecting the original pictures,problem and solution of texture inconsistion caused by image rotating when doing data augmentation,a neural network structure with better performance,and the post-processing and pipeline construction method considering the prediction results of the actual production confidence.The end-to-end prediction from the original picture to the microstructure category is finally achieved.