Deep Learning Enabled Fault Diagnosis Using Time-Frequency Image Analysis of Rolling Element Bearings

Table 6

Prediction accuracies for 32 32-pixel image inputs.

Model	Spectrogram	Scalogram	HHT
MLP flat	70.3%	94.0%	49.2%
LSVM flat	63.6%	91.8%	50.0%
SVM flat	73.9%	92.7%	58.5%
MLP PCA	62.3%	95.3%	56.7%
LSVM PCA	48.8%	89.9%	45.8%
SVM PCA	51.3%	92.5%	56.4%
Architecture 2	77.3%	92.4%	68.9%
Architecture 1	80.6%	99.8%	74.5%
Proposed CNN architecture	81.4%	99.7%	75.7%