Data-driven methodologies to estimate process parameters, design parameters and mechanical properties of fused deposition modelling polylactide components

Due to the high complexity of relationships between the various process parameters and the mechanical properties of the polylactide (PLA) components produced by the fused deposition modelling (FDM), predictions of their mechanical behaviour could be a complicated task to complete using the conventional approaches. Hence, alternative data-driven methodologies are adopted in the present investigation to serve this goal.
The present research has included the development of two alternative frameworks: dependent on the user needs, “direct” and “inverse” schemes. The former could be used to estimate mechanical properties with given process parameters, whereas the latter could identify the optimal combination of the process parameter ensuring the required mechanical behaviour. The process parameters estimated from the inverse framework can be adjusted with respect to the specifications of the printer and the software.
Three various data-driven methodologies were adopted in the present investigations, including the fuzzy inference system (FIS), artificial neural network (NN) and adaptive neural fuzzy inference system (ANFIS). The research has confirmed that with the priority being accuracy, the ANFIS is seen to be the most accurate approach, which requires particular computing power. However, FIS is reported to be the most efficient approach, which has a similar level of accuracy to the ANFIS approach.
The intrinsic versatility of the analysed data-driven methodologies has proven that these approaches could be adopted not only for process and geometrical design parameters, but also they are successful in analysing cost-relevant parameters such as printing time and material consumption. It has been shown that data-driven methodologies could be an effective and robust decision-making tool in design and cost management problems.