Applying the ANN and the GPR models to predict energy consumption for AM-FDM of isovolumetric mechanical components

Abstract

Energy consumption in additive manufacturing processes is critical in improving environmental quality and attaining a sustainable equilibrium between energy use and environmental degradation. The study employs two widely accepted non-parametric machine learning (ML) methods, i.e., Artificial Neural Network (ANN) and Gaussian Process Regression (GPR), to predict the energy consumption in the Fused Deposition Modeling manufacturing process (AM-FDM). The ANN and GPR models are compared based on a numerical general factorial design accounting for six graded-level parameters, i.e., A: learning rate, B: momentum. C: number of hidden nodes, D: proportions of the training data, E: proportion of test data, and F: the activation function. The sensitivity of the ANN and GPR models for the training subsets is further examined using different percentages of the original dataset—25%, 50%, and 75%—that are sampled using a modified hypercube sampling technique. Under all circumstances, the GPR model has performed better than the ANN.