Table 15.
Machine learning models in studies.
| ML Models | Number of Studies | Studies | Type of Faults | Importance of Faults | Environment of the Tested Result | Efficiency of the Method |
|---|---|---|---|---|---|---|
| Random forests | 3 | [48,50,59] | Machine faults, including vibration, rotation, noise, sealing, oxidation, and elongation. Four types of conditions: no-fault, ball bearing fault, main shaft fault, and combined faults. | Important. | Laboratory and applied in on mining industry. | Level 5 |
| Support vector machine | 4 | [4,23,40,47] | Damage to idler bearing and roller, off-center rotation, drum impact. | Important. | Laboratory and applied in on mining industry. | Level 5 |
| Decision tree | 1 | [58] | Faulty bearings and shafts. | Less important. | Laboratory. | Level 1 |
| Gradient boosting | 1 | [8] | Artificially defected bearings | Less important. | Laboratory. | Level 1 |
| KNN | 1 | [40] | Broken roller and off-center rotation causing drum collision. | Important. | Laboratory and applied in on mining industry. | Level 1 |
| K star | 1 | [45] | Faulty bearings and shafts. | Less important. | Laboratory. | Level 1 |
| Isolation forest | 1 | [12] | Bearing fault, thermo fault, and shell collapse. | Important. | Validated in the real condition in Western Australia for 10 months Laboratory. |
Level 3 |
| Naïve Baise | 1 | [46] | Damage to bearings and shafts. | Less important. | Laboratory. | Level 1 |
| Multilayer perceptron | 1 | [50] | Abnormal movement of a roller, off-center rotation, excessive noise, inadequate seals, damage from oxidation, and elongation of rollers. | Important. | Validated in real condition. | Level 3 |
| Artificial neural network | 4 | [23,40,46,60] | Damage to idler bearings, main shaft faults, broken roller, off-center roller rotation, and tire wear. | Important. | Laboratory and validated in real conditions. | Level 5 |
| Convolutional Neural network | 2 | [24,59] | Stuck and fracture roller. | Important. | Laboratory and validated in real conditions. | Level 5 |
| LSTM autoencoder | 1 | [44] | Surface of roller tubes, roller unbalance, and radial offset. | Important. | Laboratory and validated in real conditions. | Level 3 |
Note: Important can lead to device fault; Less-important can lead to incipient fault. Level 5 is the most efficient approach for the efficiency of the method, and level 1 is the least efficient approach.