Fault Detection in a Blower by Machine Learning-Based Vibrational Analysis

This work presents a system for Fault Detection in a Blower by Machine Learning-based Vibrational Analysis. Fault Detection System is composed of two stages. The former carries out the wavelet decomposition of the vibration signal and represents the vibration signal by the projection onto the principal components retaining 99% of the available information. The latter performs the classification by a Linear Support Vection Machine. To validate the system an experimental laboratory, where it is possible to reproduce various faults, different in intensity and in type, has been properly built. Preliminary results, even obtained on a test of limited size, are quite encouraging.