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

Table 1

Overview of CNN architectures used for fault diagnosis.

Proposed model	CNN architecture
Architecture 1 [2]	Input32 × 32−64C3 × 3−64P2 × 2−64C4 × 4−64P2 × 2−128C3 × 3−128P2 × 2−FC512
Architecture 2, Chen et al. [16]	Input32 × 32-16C3 × 3-16P2 × 2-FC10
Proposed architecture	Input32 × 32-32C3 × 3-32C3 × 3-32P2 × 2-64C3 × 3-64C3 × 3-64P2 × 2-128C3 × 3-128C3 × 3-128P2 × 2-FC100-FC100
Proposed architecture	Input96 × 96- 32C3 × 3-32C3 × 3- 32P2 × 2-64C3 × 3-64C3 × 3- 64P2 × 2-128C3 × 3-128C3 × 3-128P2 × 2-FC100-FC100
Guo et al. [17, 18]	Input32 × 32−5C5 × 5−5P2 × 2−10C5 × 5−10P2 × 2−10C2 × 2−10P2 × 2−FC100−FC50
Abdeljaber et al. [19]	Input128−64C41−64P2−32C41−32P2−FC10−10