The use of additive manufacturing is growing, especially in the small and medium sized enterprise sector. Still, the print process and its quality is prone to errors. Though there exist a variety of visual detection methods for additive manufacturing, acoustic ones are rare to find. This approach will serve as a method to detect acoustic cues and errors of a Fused Deposition Modeling printer. We propose a machine learning system detecting flaws and errors of a printer with varying difficulty. Regarding the first challenge, which is recording audio data, a microphone is attached close to the extruder of a printer. Since there is no public available data samples are recorded and annotated. To guarantee variety of the samples and more data different methods of data augmentation are applied. Mel-frequency cepstral coefficients and Mel filterbank energies are extracted from the recorded and augmented data to be used as features. A Long Short- Term Memory model was trained and validated with multiple classes of relevant sounds during 3d printing.