Artificial Neural Networks Modelling for Surface Finish in Ball Burnishing of Fused Filament Fabricated Parts

Abstract

In additive manufacturing, surface finish and component performance are critical in controlling and determining operating efficiency. The current study employs an artificial neural network (ANN) to develop an appropriate modelling approach and estimate the impact of the ball burnishing on the surface finish of parts produced using fused filament techniques. The experimental data reported in the previous work is used for ANN modelling. Different levels of tool passes, applied force, forward speed, and lateral path width are considered as input conditions, which correlate to the output parameter, surface roughness. A feed forward back-propagation neural network is used to map the relationship between inputs and outputs. The optimal ANN model architecture is four input neurons, a double-hidden layer with seven neurons, and two output neurons (4-7-7-2) based on the average absolute error in prediction (AEP) and mean sum squared error (MSE) of the predicted data. The correlation coefficient of the ANN model's predictions with the experimental data is 0.967, indicating a strong relationship. The results of the study show that the ANN model is highly efficient in predicting the surface roughness of fused filament components.