Research Article
Evaluation of Effectiveness of Wavelet Based Denoising Schemes Using ANN and SVM for Bearing Condition Classification
Table 5
Performance of the MLPNN classifier.
|
Scheme |
Number of neurons in hidden layer | All the 30 features as input | Features selected by FC as inputs | Epochs | Training accuracy (%) | Test accuracy (%) | Epochs | Training accuracy (%) | Test accuracy (%) |
| s1 | 5 | 30 | 100 | 95.67 | 82 | 100.00 | 88.75 | 10 | 35 | 100 | 92.67 | 78 | 100.00 | 89.75 |
| s2 | 5 | 36 | 100 | 93.67 | 200 | 98.83 | 92.25 | 10 | 47 | 100 | 94.67 | 63 | 100.00 | 89.25 |
| s3 | 5 | 41 | 100 | 94.33 | 38 | 100.00 | 95.25 | 10 | 61 | 100 | 90.33 | 163 | 100.00 | 90.50 |
| s4 | 5 | 15 | 100 | 96.00 | 153 | 99.92 | 91.75 | 10 | 42 | 100 | 94.00 | 102 | 100.00 | 86.50 |
| s5 | 5 | 61 | 100 | 93.67 | 157 | 99.92 | 92.50 | 10 | 182 | 100 | 86.00 | 92 | 100.00 | 88.00 |
| s6 | 5 | 207 | 99.11 | 90.00 | 152 | 99.92 | 92.25 | 10 | 80 | 100 | 87.33 | 143 | 100.00 | 89.25 |
| s7 | 5 | 22 | 100 | 97.67 | 18 | 100.00 | 95.50 | 10 | 52 | 100 | 92.33 | 124 | 100.00 | 91.25 |
|
|