Augustin, MJ and Ramesh, V and Prasad, RK and Gupta, N and Kumar, MR (2021) Detection of inclusion by using 3D laser scanner in composite prepreg manufacturing technique using convolutional neural networks. Machine Vision and Applications, 32 (6). ISSN 0932-8092
Full text not available from this repository.Abstract
Among different manufacturing techniques available for composite aircraft structures, prepreg-based manual layup is widely used. During the fabrication process, the protective films of the prepregs or other materials used in the process could get inside as a foreign object between the layers. The present method of finding the inclusions during the prepreg layup is by visual inspection in the cleanroom. Carrying out visual inspection is challenging as the layup is usually carried out on large surfaces and reflective by nature. This paper proposes a 3D laser scanner-based approach for the detection of inclusion on flat and curved surfaces. Using the portable laser scanner, the surfaces of each layer are scanned and compared the resulting point clouds using with a reference layer data. Thicknesses between two surfaces are computed with Cloud to Cloud, Mesh to Cloud and Hausdorff distance to enhance the visibility of inclusions. It was found that this approach could enhance the visibility of inclusions over 50 micron and above. These enhanced features are used to train a multiview convolutional neural network to mark the inclusion regions, which can aid the inspector to identify the inclusion regions in a fast and efficient way.
Item Type: | Article |
---|---|
Uncontrolled Keywords: | Hyperspectral imaging, Deep learning, Convolutional neural networks |
Subjects: | CHEMISTRY AND MATERIALS > Composite Materials ENGINEERING > Lasers and Masers MATHEMATICAL AND COMPUTER SCIENCES > Computer Programming and Software |
Depositing User: | Mrs. Usha Kumari |
Date Deposited: | 13 May 2022 11:05 |
Last Modified: | 13 May 2022 11:05 |
URI: | http://nal-ir.nal.res.in/id/eprint/13507 |
Actions (login required)
![]() |
View Item |