In this paper, a method for predicting failure due to malfunction of a Fused Deposition Modelling (FDM) 3D printer is presented. The method is centered around single-element piezo pads as low-power, unobtrusive sensors able to characterize the acoustic emissions of the stepper motors in an FDM printer, and from those characterizations, predict the state of the stepper motor. Acoustic signatures from the actuators in the machine are characterized for a filament feeder stepper motor. The raw signal data from the acoustic sensors is recorded during different activity from the stepper motor. The raw signal data is read in over a data acquisition device (DAQ) and converted into a readable format for a Machine Learning Algorithm (MLA). The MLA works with frequency bins averaged over a characteristic time window length, and is trained for the specific stepper states that are desired to be observed by recording large amounts of data from the acoustic sensors during known states of the stepper motor. Using the labels in the training set, new data obtained during the printing of a part is analyzed for each characteristic time window. The MLA predicts and assigns a label for that data pertaining to the stepper states that are desired to be observed. This sensor data is compared to a ground truth which is automatically generated from G-code. This can be extended out to all the stepper motors comprising an FDM machine to predict failure during a print. This work tests the above hypothesis on the filament feeder stepper motor of an FDM printer, which is responsible for pushing the filament into the heating chamber and through the extruder. The state of the feeder stepper motor is predicted for multiple tests. The first test is to predict the state of the feeder stepper motor when it is extruding filament, running empty (out of filament), or jammed. Once a ground truth is established from the G-code file of a part, another test is conducted to see if the extrusion state (extruding versus travelling) state of the stepper can be predicted over the duration of a print.The experiments result in a clear method for a universal classifier for assessing the proper motion of moving parts in an FDM machine, in this case the feeder stepper, with up to 90% accuracy in prediction. This methodology can be extended to actuators in any FDM machine due to the non-invasive, low-powered acoustic sensors chosen. The sensor array can be attached to a machine with the adhesive on the back of the piezo pads, and does not require any integration with the printer itself, allowing for a retrofit of closed loop control onto open loop FDM machines. The methods are assessed by the classifier algorithm’s accuracy in correctly predicting the current state of the stepper, ie how closely it follows the ground truth.&nbsp;&nbsp;&nbsp;

Detection of Jamming and Filament Breakage in Fdm Using Vibration of Feeder Stepper

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.