Verification and Validation of Residual Stresses in Simple Composite Structures

Abstract

Process-induced residual stresses commonly occur in composite structures composed of dissimilar materials. These residual stresses form due to differences in the composite materials’ coefficients of thermal expansion and the shrinkage upon cure exhibited by polymer matrix materials. Depending upon the specific geometric details of the composite structure and the materials’ curing parameters, it is possible that these residual stresses could result in interlaminar delamination or fracture within the composite. Therefore, the consideration of potential residual stresses is important when designing composite parts and their manufacturing processes. However, the experimental determination of residual stresses in prototype parts can be time and cost prohibitive. As an alternative to physical measurement, it is possible for computational tools to be used to quantify potential residual stresses in composite prototype parts. Therefore, the objective of this study is the development of a simplistic method for simulating the residual stresses formed in polymer matrix composite structures. Specifically, a simplified approach accounting for both coefficient of thermal expansion mismatch and polymer shrinkage is implemented within the Sandia National Laboratories’ developed SIERRA/SolidMechanics code Adagio. Concurrent with the model development, two simple, bi-material structures composed of a carbon fiber/epoxy composite and aluminum, a flat plate and a cylinder, are fabricated and the residual stresses are quantified through the measurement of deformation. Then, in the process of validating the developed modeling approach with the experimental residual stress data, manufacturing process simulations of the two simple structures are developed and undergo a formal verification and validation process, including a mesh convergence study, sensitivity analysis, and uncertainty quantification. The simulations’ final results show adequate agreement with the experimental measurements, indicating the validity of a simple modeling approach, as well as a necessity for the inclusion of material parameter uncertainty in the final residual stress predictions.