Detection and Classification of Defects in 3D Printing using a Novel Skewness and Kurtosis-based Parameter of Sound Signals and Machine Learning

This work proposes a monitoring strategy based on kurtosis and skewness of sound signals to detect and classify the machine conditions in fused deposition modeling (FDM). The methodology consisted in experimental tests conducted in a 3D printer in which an electret microphone was attached to the extruder support. The signals were acquired by an oscilloscope at 200 kHz, and then digitally processed in MATLAB. The results showed that the proposed parameter along with machine learning models produced a significant improvement when compared to the use of the skewness and kurtosis alone.