Table 7.
Results of neural network estimation from different features in [52].
| Model | Features | ANN Architecture | RMSE for Validation (ppm) |
|---|---|---|---|
| 1 | Final response | 1-2-2 | 13.4 |
| 2 | Final response | 3-3-2 | 2.2 |
| On/off derivative | |||
| 3 | Final response | 3-3-2 | 2.6 |
| Min/max derivative | |||
| 4 | Final response | 3-3-2 | 3.6 |
| Short on/off integral | |||
| 5 | Final response | 3-3-2 | 4.5 |
| 30 s on response | |||
| 90 s on response | |||
| 6 | Final response | 3-3-2 | 2.4 |
| 30 s off response | |||
| 90 s off response | |||
| 7 | Final response | 5-4-2 | 1.6 |
| 30 s on/off response | |||
| 90 s on/off response | |||
| 8 | Final response | 5-4-2 | 1.7 |
| On/off derivative | |||
| on/off integral | |||
| 9 | Final response | 6-5-2 | 1.7 |
| On/off derivative | |||
| Plateau derivative | |||
| Response/on integral | |||
| 10 | Polynomial on/off | 8-6-2 | 1.6 |
| 11 | 1. Exponential on/off | 4-4-2 | 2.0 |
| 12 | 2. Exponential on/off | 10-6-2 | 2.1 |
| 13 | ARX on/off | 6-5-2 | 2.3 |
The ANN architecture column presents the number of inputs, number of neurons in the hidden layer and number of outputs.