Acoustic Anomaly Detection in Additive Manufacturing with Long Short-Term Memory Neural Networks

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.

[1]  Alain Rakotomamonjy,et al.  Histogram of gradients of Time-Frequency Representations for Audio scene detection , 2015, ArXiv.

[2]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[3]  Lars Lundberg,et al.  Classifying environmental sounds using image recognition networks , 2017, KES.

[4]  Karol J. Piczak ESC: Dataset for Environmental Sound Classification , 2015, ACM Multimedia.

[5]  Tuomas Virtanen,et al.  Sound Event Detection in Multichannel Audio Using Spatial and Harmonic Features , 2017, DCASE.

[6]  Jack Beuth,et al.  Anomaly Detection and Classification in a Laser Powder Bed Additive Manufacturing Process using a Trained Computer Vision Algorithm , 2018 .

[7]  Mathieu Lagrange,et al.  Detection and Classification of Acoustic Scenes and Events: Outcome of the DCASE 2016 Challenge , 2018, IEEE/ACM Transactions on Audio, Speech, and Language Processing.

[8]  Liang Hou,et al.  Additive manufacturing and its societal impact: a literature review , 2013 .

[9]  Daniel Povey,et al.  Revisiting Recurrent Neural Networks for robust ASR , 2012, 2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[10]  Chih-Jen Lin,et al.  A comparison of methods for multiclass support vector machines , 2002, IEEE Trans. Neural Networks.

[11]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[12]  Justin Salamon,et al.  A Dataset and Taxonomy for Urban Sound Research , 2014, ACM Multimedia.

[13]  Xiaoli Li,et al.  A brief review: acoustic emission method for tool wear monitoring during turning , 2002 .

[14]  Yan Wang,et al.  Real-time FDM machine condition monitoring and diagnosis based on acoustic emission and hidden semi-Markov model , 2017 .

[15]  T. Dvir,et al.  3D Printing of Personalized Thick and Perfusable Cardiac Patches and Hearts , 2019, Advanced science.

[16]  Chloé Clavel,et al.  Events Detection for an Audio-Based Surveillance System , 2005, 2005 IEEE International Conference on Multimedia and Expo.

[17]  Annamaria Mesaros,et al.  Metrics for Polyphonic Sound Event Detection , 2016 .

[18]  Thomas S. Huang,et al.  Real-world acoustic event detection , 2010, Pattern Recognit. Lett..