Prediction of Low Velocity Impact Performance of Carbon/Flax Bio-Hybrid Composites Using Experimental and Neural Network Approaches

In this paper., the effect of fiber hybridization and stacking configurations on the low-velocity impact behavior of carbon/flax bio-hybrid composite laminates was investigated. The methodology was developed based on experimental testing and then utilizing an artificial neural network-based machine learning regression model. Five different stacking configurations of carbon/flax material were manufactured using the vacuum bagging method and then subjected to low-velocity impact with impact energies ranging from 15J to 90J. Data in the form of peak impact force was recorded for each stacking configuration. A symmetric configuration with a uniform distribution of flax layers across the thickness exhibited enhanced performance. Two machine learning approaches—deep neural network (DNN) with Adam optimizer and DNN with stochastic gradient descent (SGD)—were applied. The performance metrics used to evaluate the performance were the coefficient of determination $(\mathrm{R}^{2})$, mean square error (MSE)., and mean absolute error (MAE). The best-performing model for the prediction of peak impact force was DNN-SGD., with three hidden layers and 80 neurons in each hidden layer. The highest correlation and the smallest MSE and MAE values were attained using the DNN-SGD model. The developed methodology and the model serve as powerful tools for the prediction of the mechanical properties of bio-hybrid composite laminates., utilizing minimal resources and saving time as well.