Impact damage prediction of CFRP laminates with rubber protective layer using back-propagation neural networks

The carbon fiber–reinforced polymer (CFRP) structure in the aviation industry is typically subject to low-velocity impact damage during the assembly process, which can have a catastrophic effect on the strength and durability of composite structures. To reduce the impact of this damage on the composite structure, a composite laminate with a rubber protective layer has been proposed and proved to be effective in reducing the low-velocity impact damage. The delamination volume of the composite laminate with a rubber layer under low-velocity impact loadings was predicted in this study using a back-propagation neural network (BPNN). Various factors were considered, which can affect the delamination damage volume including impact diameter, rubber layer thickness, impact velocity, and specimen area. To generate enough training data, hundreds of finite element models have been simulated with various factors mentioned above as input data. Low-velocity impact tests have been conducted to validate simulation results. Simulation results were processed into delamination volume by python as output data, which can describe the damage degree of the composite. These input and output data were trained by a back-propagation network until the learning results meet the expected accuracy. The prediction error with simulation results and experiment results were within 4.3% and 11.2%, respectively.