As presented by [29,30], the traditional classification for mul-tiple faults in TIMs is essentially composed of the existence ofthe combination module, which requires all the outputs of theclassifiers in order to generate the final output. Accordingly, amulti-classifying system can become an inflexible system, as thedecision making becomes a centralized process or even a modelcreated from training a dataset with multiple faults, which cangenerate too many rules on a centralized basis.