Prediction of the material consumption of PLA plus fused deposition models using artificial neural network technique

Fused Deposition Modelling (FDM) is a complex additive manufacturing (AM) process involving multiple process parameters incapable of being modelled with conventional methods such as regression and mathematical modelling. The goal of the study is to develop an Artificial Neural Network (ANN) model that can accurately predict the material consumption of FDM printed parts considering the effect of process parameters such as layer height, infill density, printing temperature, and printing speed to create an ideal model that can optimize the use of resources and reduce material. The experiment was designed using face centered central composite design (FCCCD) yielding 78 specimens that were weighed using a densimeter to identify material consumption. Then, three networks with a different number of hidden layers and neurons were trained to identify the best-performing ANN structure with the lowest mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and highest coefficient of determination (R2). The fittest models were modelled and compared to identify the best-performing structure. Results indicated that the ANN model with double hidden layers with 19 and 14 neurons each showed the most precise prediction in modelling material consumption with the lowest MSE of 0.00096.