In-situ Characterization and Machine Learning Applications for Composite Processing

Numerous aircraft and spacecrafts utilize carbon fiber-reinforced polymer structures to significantly enhance their operational efficiency and overall performance. Autoclave composite processing offers a solution for the intricate design of complex structures. However, the rapid temperature and strain fluctuations experienced during the processing gives rise to a multitude of defects and residual stresses in the composite. In this study, we have devised an in-situ methodology that leverages Digital Image Correlation (DIC) and Machine Learning Applications to effectively observe and track defect deformations occurring throughout the process. This process presents a dataset derived from the curing process of 40 carbon fiber-reinforced polymer (CFRP) samples within an autoclave. The dataset encompasses monochromatic images acquired at 30-second intervals during the curing process through a pair of cameras. Additionally, it comprises contour plots representing three distinct types of surface strain, which are superimposed on the images using the VIC-3D software. Furthermore, the dataset contains temperature and strain data presented in comma-separated value (.csv) format. Lastly, it encompasses results obtained from a machine learning algorithm specifically trained for the identification of defects within the CFRP samples. The purpose behind releasing this dataset is to stimulate and facilitate further research and advancement in the fields of digital image correlation and machine learning technologies, particularly in their application for detecting irregularities within engineering materials.