Neural network approach to optimizing tensile strength of FDM-printed carbon fiber polyethylene terephthalate glycol

FDM is one of the most common 3DP techniques in which three-dimensional objects are built by successively layering thermoplastics materials. This paper mainly addresses the process of creating CF-PETG specimens through FDM machine. the Taguchi approach was used to identify suitable printing parameters and assess their effects on the UTS of the printed parts. Four control variables which include infill density, printing speed, infill pattern and layer height were used in the experiment to create specimens for evaluating their UTS. In order to bring prediction values closer to reality, an Artificial Neural Network (ANN) model was also created and designed based on the experimental results. The ANN model revealed high reliability in estimating UTS with the best validation performance being at epoch 7 and the best validation value as 11.3256 at epoch 1. When comparing the experimental results with those obtained using ANN for the training, validation, and testing sets, the regression coefficients calculated were 0.94854, 0.99848, and 0.97596 respectively. Consequently, this study offers the proof-of-concept of ANN models in terms of their ability to estimate the mechanical properties of FDM printed parts and identify directions by which 3D printing processes can be fine-tuned to attain the best material characteristics.