A Comparative Study of Artificial Neural Network and Regression Model for Hybrid Additive Manufacturing of Ti6Al4V Parts and Microstructural Analysis

In the present work, a comparative study of artificial neural network (ANN) and regression model for hybrid additive manufacturing (HAM) of Ti6Al4V parts is performed. This study is carried out to identify the best model for the prediction of optimal process parameters. HAM is a fusion of vat photopolymerization (VPP)-based 3D printing and powder metallurgy-based pressureless sintering (PS) technique. Microstructural studies of the fabricated parts are performed using SEM micrographs, and further XRD analysis is used to confirm the phases and their constituents produced during fabrication. To formulate the ANN model, 70% of the datasets are utilized for training, 15% for cross-validation, and the remaining 15% from a total of 20 experimental datasets for testing purposes. A backpropagation training algorithm is used to train and fit the ANN model. Further, a central composite design (CCD)-based regression model is generated to compare it with the ANN model. The findings of predicted values from both regression and the developed ANN model strongly align with experimental outcomes. Moreover, the ANN model demonstrated superior predictive performance, as evidenced by lower mean absolute percentage error (MAPE), mean square error (MSE), and mean absolute error (MAE) values compared to the multiple regression. Additionally, the coefficient of correlation (R) value for ANN model is close to 1, signifying a highly favorable correlation between the experimental and predicted outcomes. Microstructural evaluation reveals a typical Widmanstatten structure with both alpha + beta colonies with fine beta colonies trapped inside large alpha grains leading to lamellae structure as observed in SEM micrographs. Moreover, the dimples in the microstructure contribute to the ductile type of fracture, as observed in the failure analysis.