Multiclass Incremental Learning for Fault Diagnosis in Induction Motors Using Fine-Tuning with a Memory of Exemplars and Nearest Centroid Classifier

Table 6

Average accuracy and standard deviation (SD) for MCIL methods over CWRU, using a different number of incremental phases of exemplars.

1 phase	4 phases	8 phases
FT + R	91.65  03.40	83.90  03.44	81.22  04.22	98.65  01.00	98.86  00.96
FT + NCC + H	97.92  01.18	94.77  00.88	93.36  01.96	99.13  00.54	99.25  00.39
Full	100.0  0.00

Best results are boldfaced for each setting.