Smart machines for fused filament fabrication based on multi-sensor data fusion, digital twins and machine learning

Among additive manufacturing technologies, fused filament fabrication (FFF) is becoming increasingly important for highperformance applications, e. g. in the biomedical and pharmaceutical fields, which require products to conform to strict functional and geometric specifications. At the state-of-the-art, in-process monitoring is being actively investigated to improve FFF: monitoring the machine and the part during the fabrication provides opportunities for keeping quality under constant control, allowing for early process termination or for taking corrective actions in case issues are detected. In this work, ongoing research towards the implementation of a “smart” FFF machine is illustrated, where sensing and machine learning are combined to achieve real-time process monitoring and capability for self-adjustment. Through sensors, a smart FFF machine can monitor extrusion rate, temperatures and pressure. Machine vision can be used to monitor the geometry and topography of the current layer, detecting both topographic defects and part shape errors as they appear. A fundamental role is played by the presence of digital twins, i.e. computer simulations of the part being fabricated and of the FFF system, which are used by the machine AI as an aid to the decisional process, and are continuously updated through sensor data to reflect the current state of fabrication. The current opportunities and open challenges of developing a smart FFF machine are highlighted through the illustration of an open, modular architecture which we have been developing as a testbed for multisensing and AI in FFF. Issues are discussed through the application to a selected set of test cases