High-fidelity computational modeling of scratch damage in automotive coatings with machine learning-driven identification of fracture parameters

Abstract

As one of the key components of a vehicle, automotive coating systems are typical micro-size multilayer composites, which normally suffer from intricate scratch damage. Currently, it is a challenging task to reproduce the complex physical phenomena of automotive coatings with micron thickness using numerical simulations. The purpose of this work is to develop a computational framework for an accurate evaluation of such scratch damage. To achieve this end, a high-fidelity finite element (FE) model is proposed, where a rate-dependent cohesive zone model is developed to account for coating crack behaviors. A reverse identification approach based on the machine learning (ML) method is suggested to determine the rate-dependent cohesive parameters that are difficult to be measured via direct experiments. In the course of recognition, this work develops a two-step regional data augmentation strategy based on the existing single-shot recognition method. The developed strategy is capable of reducing the number of training data samples, while improving identification accuracy and robustness. With the identified cohesive parameters, the proposed finite element model is applied to simulate speed-dependent coating scratch behaviors. The good agreement between numerical and experimental results demonstrates the capacity of our developed computational framework for coating scratch problems.