Predictive maintenance bearing failure using sound

Creation for a company in the railway sector of a model for the detection of rolling stock bearing problems based on recordings taken from the running train. The possible alternatives were analyzed given the limitation of the recordings and it was decided to use a logistic regression model based on the frequency spectrum.

• Debugging of data sources: Recordings of predicted failures and non-anomalous situations
  
• Analysis of alternatives
  
• Model creation and training
  
• Testing and deployment