Comparative performance of the proposed CNN–LSTM digital twin against representative AE-based prognostic frameworks reported in the literature. The proposed model achieves higher classification accuracy and lower RUL prediction error while eliminating the need for experimental or finite element datasets
| Study (Year, Journal) | Approach | Dataset type | Classification accuracy (%) | RMSE | Remarks |
|---|---|---|---|---|---|
| Du et al. (2024), Eng. Fract. Mech. | CNN–LSTM | Experimental AE (composite) | 95.2 | 6.8 | Single-task classification only |
| Ai et al. (2023), Mech. Syst. Signal Process. | LSTM + Transfer Learning | FEA AE | 96.4 | 5.9 | Domain-specific model |
| Karvelis et al. (2021), Ships and Offshore Struct. | SVM (handcrafted features) | Experimental AE | 93.0 | 7.2 | Limited generalization |
| Kim et al. (2022), J. Intell. Manuf. | Multi-task CNN (HS + RUL) | C-MAPSS (aero-engine prognostics) | – (dataset-specific) | – | Multi-task model demonstrating joint health-state and RUL prediction; strong benchmark for dual-task design |
| Li et al. (2015), Int. J. Adv. Manuf. Technol. | Hybrid CNN–LSTM + Transfer Learning | Experimental AE (cutting-tool wear) | – (dataset-specific) | – | Achieved improved RUL estimation and domain adaptation across tool datasets |
| Present Study (Proposed) | Multi-task CNN–LSTM Digital Twin | Synthetic AE + NASA PCoE Validation | 99.0 (Synthetic)/89.3 (NASA) | 4.8/7.5 | Unified, interpretable, physics-informed, data-efficient digital twin |
| Study (Year, Journal) | Approach | Dataset type | Classification accuracy (%) | RMSE | Remarks |
|---|---|---|---|---|---|
| CNN–LSTM | Experimental AE (composite) | 95.2 | 6.8 | Single-task classification only | |
| LSTM + Transfer Learning | FEA AE | 96.4 | 5.9 | Domain-specific model | |
| SVM (handcrafted features) | Experimental AE | 93.0 | 7.2 | Limited generalization | |
| Multi-task CNN (HS + RUL) | C-MAPSS (aero-engine prognostics) | – (dataset-specific) | – | Multi-task model demonstrating joint health-state and RUL prediction; strong benchmark for dual-task design | |
| Hybrid CNN–LSTM + Transfer Learning | Experimental AE (cutting-tool wear) | – (dataset-specific) | – | Achieved improved RUL estimation and domain adaptation across tool datasets | |
| Present Study (Proposed) | Multi-task CNN–LSTM Digital Twin | Synthetic AE + NASA PCoE Validation | 99.0 (Synthetic)/89.3 (NASA) | 4.8/7.5 | Unified, interpretable, physics-informed, data-efficient digital twin |
Note(s): Reported metrics are dataset-specific and not directly comparable across studies. The purpose of this table is to highlight methodological trends and demonstrate that the proposed multi-task, physics-informed framework achieves state-of-the-art performance with superior interpretability and data efficiency
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.