Internship | Machine learning for acoustic quality control
End-Of-Line & Quality Control
Traditional End-Of-Line (EOL) solutions often encounter difficulties adapting from controlled environments to industrial production lines due the presence of high levels of noise and vibrations generated by the surrounding machinery. In contrast, particle velocity measurements performed near a rigid radiating surface are less affected by background noise and they can potentially be used to address noise problems even in such conditions. The vector nature of particle velocity, an intrinsic dependency upon surface displacement and sensor directivity are the main advantages over conventional solutions. As a result, quantitative measurements describing the vibro-acoustic behavior of a device can be performed at the final stage of the manufacturing process.
In this project machine learning algorithms to automatically detect vibro-acoustic anomalies should be developed. The work will be focused on implementing unsupervised algorithms for feature extraction and classification. Computer simulations and experimental measurements should be carried out to validate the proposed algorithms and evaluate the limitations of the implemented tools. If the work and findings are of good quality, preparing a publication for an international congress or journal will be encouraged. In summary, the work should include a good description of audio signal processing, computer simulations, test measurements and assessment of experimental results.
The candidate should:
- be proficient programming in MatLab
- be proficient in signal processing
- have prior knowledge in machine learning, audio AI and/or classification techniques