A lack-of-fit test can be used to determine whether a finite element (FE) material model, or experimental results should be the focus to improve overall model accuracy. The lack-of-fit test compares the unbiased estimate of variance to the estimated variance from noise within data. GSJ and Hart methods from literature are provided to estimate the noise within data, and a generalized lack-of-fit test is described. The variance estimators are compared and convergence is demonstrated on a simple regression problem. The lack-of-fit test is then applied to a material calibration problem using the FE method. The lack-of-fit test indicates that the shear component of a non-linear orthotropic material models could be improved, despite already being considered an excellent fit. Additionally the lack-of-fit test is used with a load-dependent Poisson's ratio to demonstrate that the variance in the experimental data should be reduced in order to improve the model.
展开▼
