11 Improving Bearing Diagnostic Performance …
123
Fig. 11.4 Predicted classification accuracies of the k-NN-based classifier using 30 discriminatory
feature subset candidates
Table 11.2 Summary of
feature subsets determined by
the proposed faul-signature
analysis
Dataset
The most discriminatory feature subset
Dataset 1
{f 5 , f 7 , f 10 }
Dataset 2
{f 2 , f 10 , f 13 }
Based on the above analyses and the decision rule, the two most discriminatory
feature sets, associate with two datasets, are determined as in Table 11.2.
11.3.3 Efficacy of the New GA-Based Feature-Selection
Scheme
This section proves the outstanding quality of a new feature-evaluation scheme by
comparing the predicted accuracy of the classifier utilizing the new-feature scheme
and the figures of the classifier as applying two other state-of-the-art counterparts. These rival methods are principal component analysis (PCA) and independent
component analysis (ICA). Such methods are a kind of the component analysis in