Predictive Model for Creep Behavior of Composite Materials Using Gene Expression Programming

In this article, a new method for establishing creep predictive model using gene expression programming (GEP) is proposed. The three-point bending tests under constant load are carried out to determine time-dependent creep curves of fiber reinforced polymer materials with different lay-up styles, a modeling program is developed to predict creep behavior of composite materials. The creep of fiber is much smaller than that of resin matrix, various fiber layups play a role in constraining the deformation of resin, resulting in differences in creep performance of composites. The mathematical model satisfies the variation law that creep strain monotonically increases with time and tends to be stable. Based on 0 ~ 1000 h experimental data, the creep model is established by GEP, and then utilized to predict creep ranging from 1000 to 2000 h, the predicted values are in good agreement with experimental values. The fitting efficiency and prediction accuracy of GEP model are demonstrated by R 2 , RMSE , MAE and RRSE metrics. Moreover, taking R 2 as a statistical metric, the validity of developed model is verified by comparison with Burgers model, Findley model and HKK model. Creep factor calculated by GEP model is lower than standard specified value, and the relative errors δ of creep deflection are very low, all within about 10%, indicating that GEP model can accurately predict the long-term creep performance of composites.