Validation of Thermoforming Simulation Models Prior to Parameterization Using Covariance-Based Input-Output Statistics – Assessing the Role of Thermomechanical Material Modeling

Accurate yet computationally efficient simulation models are essential for the virtual design and optimization of thermoforming processes for fiber-reinforced composites. Selecting an appropriate material model remains challenging, particularly when balancing model fidelity against computational cost. In this work, a framework is developed to validate material models used in thermoforming simulations for fiber-reinforced composites. The framework evaluates model performance based on time-series data using covariance-based input-output statistics, without prior calibration. Two numerical studies of increasing complexity demonstrate the versatility of the approach. First, the framework is applied to one-dimensional rheological models, verifying its applicability to mechanical problems relevant to thermoforming simulation. These insights are then applied to complex finite element thermoforming simulations to assess the ability of isothermal material models to predict wrinkling behavior in comparison to a fully coupled thermomechanical reference model. A curvature-based method is introduced to quantitatively evaluate wrinkling severity relative to natural curvatures induced by the tool geometry. The results show that isothermal models are sufficient for short total process times with minor temperature-driven effects, whereas longer total process times with pronounced thermal effects require thermomechanical models to ensure accurate predictions. The findings offer practical guidance for selecting appropriate material models based on specific process conditions, as well as objective criteria for assessing model validity in virtual process design.