Open Access
Review
Table 9
Comprehensive performance analysis of soft sensing models in terms of accuracy, robustness, real-time capability, and applicability.
| Model category | Accuracy range (%) | Robustness/operating condition disturbance (%) | Real-time performance (ms) | Sample requirement | Applicable scenarios |
|---|---|---|---|---|---|
| Traditional ML | 87–97 | Poor (Δ-15-25) | Excellent (5–50) | Low (50–500) | Real-time edge node warning [15] |
| Basic Deep Learning | 87–94 | Moderate (Δ-8-15) | Moderate (50–100) | Medium (1,000–3,000) | Lab-condition diagnostics [28] |
| Advanced Deep Learning | 96–99.5 | Good (Δ-2-8) | Medium–high (80–150) | High (3,000–5,000) | Complex fault identification [29] |
| Transfer Learning | 92–98 | Excellent (Δ-1-5) | Moderate (60–120) | Low–medium (100–1,000) | Cross-condition adaptation [12] |
| Lightweight Models | 94–98 | Good (Δ-4-10) | Good (20–60) | Medium (500–2,000) | Edge-cloud collaborative deployment [37] |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.