Thermoset fiber-reinforced composites (FRPCs) are widely used in engineering applications because of their specific properties such as strength and stiffness. While these materials are in heavy use in sectors such as aerospace and wind energy, the persistent status quo is to overdesign these materials to avoid failure, resulting in unnecessary costs and manufacturing delays. A major cause of over-conservative design is the lack of confidence in FRPC mechanical properties due to large scatter in experimental data.

The manufacturing process defines the composite microstructure whose imperfections drive failure initiation and propagation. During manufacturing, fibers are compacted, handled, and spooled, which causes the fibers to deviate from ideal axial alignment. The process results in a disorganized structure in which there are matrix rich pockets and off-axial fiber bundles. In turn, fibers mechanically constrain the matrix during curing. Residual stress that builds up in this process can result in matrix cracking and uneven matrix curing. That is, cured matrix exhibits in-situ properties, which are effective mechanical properties of the solidified resin that take into account the presence of residual stresses, local temperatures and imperfections. This causes the matrix material to vary locally from the ideal, bulk matrix properties.

Although the effect of matrix in-situ properties has been correlated to FRPC failure mechanisms through testing at the macroscale, matrix properties variability has not been quantified at the microscale.  Therefore, to understand the source of scatter in composite mechanical properties and eventually optimize the manufacturing process, both fiber spatial distribution and matrix in-situ properties need to be quantified.

This work aimed to establish the curing-microstructure-property relationship of FRPCs through experimental techniques and computational modeling considering both the fiber geometrical arrangement and matrix properties variability. The outcomes can be broken up into three categories:

First, images of FRPC specimens were taken at the micro-scale to develop a feature-based mathematical language to describe the orientation of fibers within the composite.  CT scans and serial sectioning were used on specimens created with common manufacturing methods found in industry.  Algorithms were developed to automate the extraction of fiber positions and connect them within images, where images often contain thousands, tens-of-thousands, or more fibers.  A method of identifying fiber clusters and voids was developed, along with other mathematical quantities to aid in the description of the microstructure.  These descriptors showed the difference between composites made by different methods, which could one day lead to predicting the scatter in mechanical properties based on the manufacturing process employed.    

Second, matrix properties variability was quantified by performing Micro-Raman scattering and using imagining techniques to virtually reconstruct the microstructure during curing. A technique called 2D Raman mapping was ussed to study the behavior of epoxy during the curing process. We developed algorithms to help quantify the amount of cure as it happens. This program helped us analyze 2D and 3D images, allowing us to see differences in the cured and uncured parts within a composite material. We were able to get detailed information about the structure and chemical makeup at the boundary between the fibers and the resin. Our early results suggest that carbon fibers might make the epoxy cure differently at fiber boundaries.

Lastly, a computational curing model was combined with the fiber position work to create strength predictions based on the microstructure.  A workflow was created to go from cross-sectional images at the micro-scale to small representative volume element models that can predict the stresses which occur during curing and the subsequent strength of the composite part.  This provided insight into the correlation between the processing, resulting microstructure, and properties of composites, and improved our ability to simulate and predict the performance of composites virtually. 

Last Modified: 02/19/2024
Modified by: Scott E Stapleton