Research Article
Feed-Forward Neural Network Soft-Sensor Modeling of Flotation Process Based on Particle Swarm Optimization and Gravitational Search Algorithm
Table 2
Predictive performance comparison of soft-sensor models.
| Predictive object | Predictive method | MSE | RMSE | NRMSE | SSE | MAPE |
| Concentrate grade (%) | FNN | 0.2416 | 0.4915 | 0.0223 | 12.081 | 0.5564 | PSO-FNN | 0.1528 | 0.3909 | 0.0008 | 7.6402 | 0.4471 | GSA-FNN | 0.0764 | 0.2764 | 0.0006 | 3.8209 | 0.2649 | PSOGSA-FNN | 0.0300 | 0.1733 | 0.0004 | 1.5009 | 0.1890 |
| Flotation recovery rate (%) | FNN | 0.1751 | 0.4184 | 0.0163 | 8.7535 | 0.3089 | PSO-FNN | 0.0627 | 0.2505 | 0.0004 | 3.1372 | 0.1751 | GSA-FNN | 0.0447 | 0.2114 | 0.0003 | 2.2341 | 0.1422 | PSOGSA-FNN | 0.0194 | 0.1393 | 0.0002 | 0.9707 | 0.1054 |
|
|