A method for detecting a tripping of a disconnection unit in an electrical enclosure, the electrical enclosure comprising at least one disconnection unit for disconnecting electric current and at least one switching unit for switching the electric current, the method comprising a learning phase configured to generate decision-making categories associated with an acoustic signature of the tripping of said at least one disconnection unit, and a phase of detecting the tripping of the disconnection unit, comprising:
- acquiring a unit noise signal generated by at least one of said disconnection units for disconnecting the electric current or at least one switching unit for switching the electric current, and
- comparing said unit noise signal with said decision-making categories in order to detect whether the unit noise signal corresponds to the tripping of the disconnection unit,
- and wherein the learning phase comprises:
- injecting a predetermined disconnection current into the disconnection unit and recording the signal of the disconnection noise generated,
- sending the signal of the disconnection noise to a machine learning system configured to recognize the acoustic signature of the disconnection unit and providing discrete disconnection values as output,
- grouping the discrete disconnection values together so as to create the decision-making categories,
- injecting a predetermined current into the switching unit and recording the signal of the switching noise generated,
- sending the signal of the switching noise to the machine learning system and obtaining discrete switching values as output, and
- distinguishing the discrete switching values from the discrete disconnection values by modifying the decision-making categories so as to exclude discrete switching values which come within these decision-making categories.