Automatic normality estimation of 3D printers using peeling sounds

Liquid crystal display type of three-dimensional (3D) printers generates a small peeling sound when creating 3D objects. The sound is generated when the 3D printer peels off the curable resin from a fluorinated ethylene propylene film. Moreover, this peeling sound occurs continuously when the 3D objects are properly created. Thus, if we can properly distinguish this peeling sound from the mechanical noise of the 3D printer, we can prevent the loss of time and increase in material cost due to modeling failure. Thus, in this paper, we propose a method to automatically identify the normal operation of a 3D printer using the peeling sound in the operating noise. To identify the normality of the 3D printer, we convert the obtained machine noise from the 3D printer into short-time spectrogram images and build a machine learning model using the created spectrogram images. Moreover, we evaluate the normality of the 3D printer from its machine noises using the learning model.