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

Table 18

Prediction accuracies for 32 32 image inputs.

Model	Spectrogram	Scalogram	HHT
MLP flat	92.7%	83.6%	59.6%
LSVM flat	88.6%	80.8%	59.7%
SVM flat	97.3%	89.3%	72.5%
MLP PCA	89.4%	94.7%	76.0%
LSVM PCA	77.9%	69.3%	59.7%
SVM PCA	74.4%	90.0%	80.0%
Architecture 2	95.9%	92.6%	78.0%
Architecture 1	98.4%	99.2%	88.9%
Proposed CNN architecture	98.1%	98.8%	86.5%