A comparative study between backpropagation and general regression neural networks to predict tensile properties of fused filament fabricated parts

Abstract

Abstract The Fused Filament Fabrication (FFF) process comprises a large number of process parameters that affect the resultant mechanical properties of the parts, and that generates uncertainties in the design for Additive Manufacturing process. The use of Artificial Neural Networks (ANN) based on classification machine learning techniques such as Backpropagation Neural Networks (BPNN) have been proposed in the literature to evaluate the dimensional accuracy, surface roughness, compressive, flexural and tensile strength of FFF parts. As an alternative, in this paper a new General Regression Neural Networks (GRNN) approach, based on a regression machine learning technique, is proposed and compared with the performance of a BPNN to estimate the tensile structural properties of PLA-FFF parts using variable process parameters. The comparison and evaluation are based on their capability to accurately predict the experimental Ultimate Tensile Stress (UTS) and the Elastic Modulus (E). The results have shown that although the BPNN and the GRNN are able to estimate with high accuracy the structural behaviour of FFF parts, the new proposed GRNN better fits the experimental results and the current needs of Design for Additive Manufacturing (DfAM).