Research Article
A Damage Classification Approach for Structural Health Monitoring Using Machine Learning
Table 2
Behavior of machines with two scores per sensor (specimen 1, four sensors).
| Machine type | UND | DMG1 | DMG2 | DMG3 | DMG4 | DMG5 | DMG6 |
| Complex Tree | 90% | 99% | 13% | 92% | 100% | 90% | 100% |
| Medium Tree | 90% | 88% | 13% | 92% | 100% | 90% | 100% |
| Simple Tree | 90% | 99% | 0% | 0% | 100% | 90% | 100% |
| Linear SVM | 96% | 98% | 81% | 95% | 99% | 99% | 100% |
| Quadratic SVM | 96% | 98% | 96% | 95% | 99% | 99% | 100% |
| Cubic SVM | 96% | 99% | 98% | 95% | 99% | 99% | 100% |
| Fine Gaussian SVM | 68% | 100% | 57% | 87% | 79% | 78% | 99% |
| Medium Gaussian SVM | 97% | 100% | 76% | 100% | 97% | 98% | 100% |
| Coarse Gaussian SVM | 95% | 98% | 94% | 96% | 99% | 99% | 100% |
| Fine KNN | 97% | 100% | 96% | 98% | 99% | 100% | 100% |
| Medium KNN | 95% | 100% | 93% | 94% | 99% | 100% | 100% |
| Coarse KNN | 91% | 100% | 85% | 80% | 99% | 100% | 94% |
| Cosine KNN | 95% | 100% | 74% | 89% | 99% | 100% | 100% |
| Cubic KNN | 95% | 99% | 89% | 93% | 99% | 99% | 100% |
| Weighted KNN | 95% | 100% | 95% | 97% | 99% | 100% | 100% |
| Boosted Trees | 90% | 100% | 20% | 1% | 100% | 98% | 100% |
| Bagged Trees | 99% | 100% | 71% | 95% | 100% | 100% | 100% |
| Subspace Discriminant | 97% | 100% | 64% | 97% | 100% | 100% | 100% |
| Subspace KNN | 97% | 100% | 82% | 98% | 100% | 100% | 100% |
| Rusboosted Trees | 90% | 100% | 0% | 0% | 0% | 0% | 0% |
|
|