Low-velocity impact simulation of carbon fiber reinforced composite laminate using IFF-criterion based on BP-ANN

Damage characteristics of composite materials after low-velocity impacts are hard to predict and numerical simulation is time-consuming. This paper explores a study in carbon fiber reinforced composite (CFRP) laminates low-velocity impact (LVI) simulation with Puck inter-fiber failure (IFF) criterion based on the back-propagation neural network (BP-ANN) model. LVI numerical simulations under three different energy levels were conducted and the resulting element stress components were extracted. After data screening and stratified sampling, a training set including five stress tractions as input and the corresponding fracture angle as output was generated. With this dataset, a fracture angle search algorithm based on the BP-ANN model containing three hidden layers was established. Through Matlab coding and ABAQUS subroutine tests, the algorithm exhibits high efficiency compared with other common algorithms and distinctly accelerates the simulation progress (reduce cost time 22.75% and 38.06% compared with Puck's method and SRGSS algorithm). In addition, the simulation results, including the predicted damage size, impact force, displacement and absorbed energy, are in good agreement with experimental results. This work provided a referable thought to accelerate the impact simulation process with the fracture angle-based criteria.