In this paper, a new bearing defect diagnostic and classification method is
proposed based on pattern recognition of statistical parameters. Such a pa
ttern recognition problem can be described as transformation from the patte
rn space to the feature space and then to the classification space. Based o
n trend analysis of six commonly used statistical parameters, four paramete
rs, namely, RMS, Kurtosis, Crest Factor, and Impulse Factor, are selected t
o form a pattern space. A 2-D feature space is formulated by a nonlinear tr
ansformation. An intraclass transformation is used to cluster the data of d
ifferent bearing defects into different regions in the feature space. The c
lassification space is constructed by piecewise linear discriminant functio
ns. Training the classification space is performed, in this paper, by using
data of bearings with seeded defects. Diagnosis of the defected bearings i
n the classification space then becomes straightforward. Numerical experime
nts show that the proposed method is effective in indicating both the locat
ion and the severity of bearing defects.