Improvement of part strength prediction modelling by artificial neural networks for filament and pellet based additively manufactured parts

Abstract

ABSTRACT Nowadays, extrusion-based additive manufacturing techniques have been utilised from domestic to industrial end-product fabrications. However, parts fabricated through these techniques are facing poorer part strength issues due to its high nature of anisotropic fabrication strategies. Moreover, it may be improved if it can be predicted in the initial stage of fabrication through part strength modelling technique. In this direction, many researchers have attempted to model part strength estimation models that could be utilised for prediction purposes and hence controlled and improved the part strength. However, due to the anisotropic nature of 3D printed parts, models developed in the existing literature are showing more prediction errors, which can affect the final part strength. To overcome this issue, a new part strength modelling technique based on ANN technique has been defined in this study, which reduces the prediction error significantly. For the validation of the developed model, a comparative study has been performed by considering previously developed models and corresponding data available in the existing literature. Results obtained from the proposed ANN model show that prediction errors reduced significantly and it can be one of the possible alternatives to predict part strength of anisotropic behaviour of 3D printed parts.