Physics Informed Synthetic Image Generation for Deep LearningBased Detection of Wrinkles and Folds

Abstract

Deep learningbased image segmentation methods have showcased tremendous potential in defect detection applications for several manufacturing processes. Currently, majority of deep learning research for defect detection focuses on manufacturing processes where the defects have welldefined features and there is tremendous amount of image data available to learn such a datadense model. This makes deep learning unsuitable for defect detection in highmix low volume manufacturing applications where data are scarce and the features of defects are not well defined due to the nature of the process. Recently, there has been an increased impetus towards automation of highperformance manufacturing processes such as composite prepreg layup. Composite prepreg layup is highmix low volume in nature and involves manipulation of a sheetlike material. In this work, we propose a deep learning framework to detect wrinklelike defects during the composite prepreg layup process. Our work focuses on three main technological contributions: (1) generation of physics aware photorealistic synthetic images with the combination of a thinshell finite elementbased sheet simulation and advanced graphics techniques for texture generation, (2) an opensource annotated dataset of 10,000 synthetic images and 1000 real process images of carbon fiber sheets with wrinklelike defects, and (3) an efficient twostage methodology for training the deep learning network on this hybrid dataset. Our method can achieve a mean average precision (mAP) of 0.98 on actual production data for detecting defects.