Efeito de spring-back em compósitos curvos : uma análise estatística multivariável

Abstract

This work investigates the spring-in effect as a function of time in L-shaped composites. The spring-in effect represents the angular deformation measured immediately after the manufacture of a curved composite, but this deformation varies over time. The main objective of this study is to reclassify this angular deformation as a spring-back effect, that is, as the return of angular deformation of the spring-in over time. The spring-back effect is measured over approximately 30 months for several different composite families. Statistical results have shown that the degree of cure associated with the cooling rate and extensional rigidity has a significant effect on spring-back. In order to investigate the influence of mechanical variables on spring-in and spring-back effects, forces and moments components were calculated from samples processed from three different autoclave processing cooling rates, different thicknesses, and several stacking sequences. This angular deformation variable tends to stabilize after approximately three years from the manufacture of the composite. The angular deformation observed shows direct dependence on the residual curing process of the resin composing the epoxy polymer matrix of each sample. Thirteen angular deformation measurements were taken over from May 2015 to December 2018. The forces and moments components were obtained indirectly through the Classical Lamination Theory (CLT) for each configuration family. It was proposed a statistical study using: Linear and non-linear multivariable regression analysis, Generalized Additive Models (GAM) and Random forest models. Analyzes made by the first statistical approach investigated the effect of several mechanical and physical properties, as well as the time variable as predictive factors of angular deformation. The other approaches investigated the influence of these components of forces and moments on the spring-in and spring-back effect. Multivariate regression analysis found a direct relationship between degree of cure, thickness, stiffness and time as strong influencers of these angular deformations. The analysis obtained by means of the GAM and Random Forest modeling proved that the Nx force component and the Mx momentum component showed a great influence on the spring-in and spring-back effect. The GAM statistical models predicted the influence of the components of the forces and moments with accuracy close to 90%, while the Random Forest Model explains the angular deformation, on average, in values greater than 85%.