Fault Diagnosis Method for Rolling Bearing Based on Sparse Principal Subspace Discriminant Analysis

Table 3

The recognition accuracies of different methods to the bearing sets with the SVM classifier.

Methods	Normal (%)	Outer race fault (%)	Inner race fault (%)	Ball fault (%)	Average accuracy
Original	77.8	76.8	82.4	76.6	78.4
PCA	81.6	78.6	90.4	84.2	83.7
FDA	94.2	94	95.4	94.4	94.5
PSDA	94.4	94.2	100	93.8	95.6
MFA	88.4	91.8	98.2	95.8	93.55
SPSDA	96.4	94.2	100	100	97.65