Pre-process Optimisation of Filament Feed Rate in Fused Filament Fabrication by Using Digital Twins and Machine Learning

One of the limitations of fused filament fabrication (FFF) in mass customisation is the long trial and error process required to optimise process parameters under frequent changes of geometries, materials and structural/mechanical requirements. Extrusion parameters may also need to be changed in-process, for example to address different requirements of skin and internal regions within the same part. This work explores the possibility of making a FFF machine capable of autonomous optimisation of extrusion parameters, currently for use in pre-process optimisation, but in future also applicable to in-process adaptive optimisation and control. Through a combination of machine learning and digital twinning, the proposed solution is able to automatically modify a part program optimising extrusion parameters to improve uniformity of widths of the extruded strands. The solution learns how to modify the part program using data from example depositions (tests runs) and simulation models. The proposed approach is demonstrated through the application to a test case.