Modelling and prediction of mechanical properties of FFF-printed polycarbonate parts using ML and DA hybrid approach

Abstract

Fused filament fabrication (FFF) is a rapidly growing additive manufacturing technique. It is widely used in various industrial applications due to its ability to efficiently produce functional parts with complex geometrical features. Estimating the mechanical properties and dimensional accuracy is essential for the functional testing of objects fabricated using the FFF process. Several process variables influence the mechanical qualities and dimensional accuracy of objects manufactured using FFF technology. Selecting the optimal set of parameters is crucial for achieving the desired properties in the final parts. This research investigated the influence of four crucial process variables, layer thickness, extrusion temperature, printing speed, and extrusion width, on the impact resistance and shear strength of polycarbonate parts printed using the fused filament fabrication (FFF) technique. A hybrid modelling approach involving dimensional analysis (DA)–based mathematical modelling and regression-based machine learning (ML) modelling was adopted to predict the two output responses and determine the correlation between the process parameters and mechanical properties. A comparison based on various error metrics and the performance of the models suggested that ML models have higher prediction performance and accuracy than DA models. The developed prediction models exhibited significant agreement with the observed values and may be used to forecast the mechanical characteristics of FFF components while manipulating the input parameters. The findings revealed that a maximum impact strength of 66.37 J/m and shear strength of 50.43 MPa were obtained when the layer height, extrusion temperature, printing speed, and extrusion width were 320 µm, 280 °C, 20 mm/s, and 0.56 mm, respectively.

Graphical abstract