Remaining Useful Life prediction on C-MAPSS FD004
12.88RMSEHSMGNN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HSMGNN2025.12 | 12.88 | 8.55 | 165.73 | — | |
| FCSTGNN2025.12 | 13.62 | 9.92 | 185.5 | — | |
| Bi_cLSTM (Ours, 4 blocks)Year=2025, Pre-processing steps=OS norm; z-score by mode; RUL cap=1252026.02 | 14.25 | — | — | — | |
| MAGNN2025.12 | 14.3 | 14.13 | 204.49 | — | |
| DFormerYear=2024, Pre-processing steps=Transformer variant2026.02 | 14.55 | — | — | — | |
| MsFormer2026.03 | 14.81 | — | — | 961.42 | |
| Neural ODEYear=2023, Pre-processing steps=ODE layer; norm2026.02 | 15.06 | — | — | — | |
| MSAN2026.03 | 15.13 | — | — | 1,179 | |
| Crossformer2025.12 | 15.39 | 10.66 | 237.32 | — | |
| PE-Net2026.03 | 15.4 | — | — | 1,103.18 | |
| DVGTformer2026.03 | 15.5 | — | — | 1,107.5 | |
| SAGDFN2025.12 | 15.77 | 10.52 | 248.75 | — | |
| TS-MLLM2026.03 | 15.94 | — | — | 1,715.11 | |
| KDnet2026.03 | 15.96 | — | — | 1,303.19 | |
| AMR-Net2026.03 | 16.3 | — | — | 998.57 | |
| Variational EncodingYear=2022, Pre-processing steps=Variational reg.; norm2026.02 | 16.37 | — | — | — | |
| DeeBERT2026.03 | 16.53 | — | — | 1,671.03 | |
| NSD-TGTN2026.03 | 16.64 | — | — | 1,493 | |
| GAT-DAT2026.03 | 16.8 | — | — | 1,928.6 | |
| MR-LSTM2026.03 | 16.81 | — | — | 1,785.33 | |
| MLEAN2026.03 | 16.89 | — | — | 1,370 | |
| TF-SCN2026.03 | 16.95 | — | — | 1,367 | |
| Bi-LSTM based Attention methodYear=2022, Pre-processing steps=RUL target function2026.02 | 16.96 | — | — | — | |
| LSTM + multi-layer self-attnYear=2021, Pre-processing steps=Windows; norm; self-attn2026.02 | 17.21 | — | — | — | |
| MGLSNYear=2023, Pre-processing steps=Multi-granularity; norm2026.02 | 17.26 | — | — | — | |
| MSTSDNYear=2024, Pre-processing steps=Multi-scale two-stream; norm2026.02 | 17.33 | — | — | — | |
| DTW-GPR2026.03 | 17.56 | — | — | 2,509.3 | |
| LSTM (with auto piece-wise)Year=2022, Pre-processing steps=Corr. analysis; median filter; norm; auto piece-wise RUL2026.02 | 17.63 | — | — | — | |
| DMHA-ATCNYear=2024, Pre-processing steps=Hybrid CNN2026.02 | 17.76 | — | — | — | |
| BLS+TCNYear=2022, Pre-processing steps=Feature sel.; norm; piece-wise RUL2026.02 | 18.12 | — | — | — | |
| EAPNYear=2023, Pre-processing steps=Embedded attention; norm2026.02 | 18.12 | — | — | — | |
| EAPN2026.03 | 18.12 | — | — | 2,050.72 | |
| MegaCRN2025.12 | 18.37 | 13.17 | 337.77 | — | |
| DiffRULYear=2024, Pre-processing steps=Diffusion-based RUL2026.02 | 18.43 | — | — | — | |
| One Fits AllBackbone=GPT-22026.03 | 18.5 | — | — | 2,088.39 | |
| TATFA-TransformerYear=2024, Pre-processing steps=Transformer + attention2026.02 | 18.81 | — | — | — | |
| Hybrid DL prognosticsYear=2021, Pre-processing steps=Fusion; norm2026.02 | 19.15 | — | — | — | |
| BiGRU2026.03 | 19.23 | — | — | 1,605.18 | |
| CNN-SSEYear=2024, Pre-processing steps=CNN + squeeze-excite2026.02 | 19.39 | — | — | — | |
| CDSGYear=2023, Pre-processing steps=Graph structure; norm2026.02 | 19.73 | — | — | — | |
| GA–RNN/LSTMYear=2021, Pre-processing steps=GA tuning; norm2026.02 | 19.74 | — | — | — | |
| C-Transformer2026.03 | 19.77 | — | — | 3,237.37 | |
| IMDSSN2026.03 | 19.78 | — | — | 2,852.81 | |
| Attention-DCNNYear=2021, Pre-processing steps=Conv features; attention; norm2026.02 | 19.88 | — | — | — | |
| GA+PredictorYear=2023, Pre-processing steps=GA tuning; norm2026.02 | 20.15 | — | — | — | |
| BiGRU + Temporal AttnYear=2022, Pre-processing steps=Norm; temporal attn2026.02 | 20.47 | — | — | — | |
| One Fits AllBackbone=Qwen3-0.6B2026.03 | 20.69 | — | — | 8,543.54 | |
| MATT2025.12 | 20.9 | 16.13 | 436.48 | — | |
| AGCNNYear=2020, Pre-processing steps=Feature sel.; norm; RUL target2026.02 | 21.5 | — | — | — | |
| Multi-Scale CNNYear=2020, Pre-processing steps=Multi-scale conv; norm2026.02 | 22.22 | — | — | — | |
| Hybrid modelYear=2021, Pre-processing steps=Feature sel.; norm; piece-wise RUL2026.02 | 22.32 | — | — | — | |
| Multi-head CNN+LSTMYear=2020, Pre-processing steps=Feature sel.; RUL target2026.02 | 22.89 | — | — | — | |
| CNN+LSTMYear=2019, Pre-processing steps=Var. threshold; norm; HI2026.02 | 23.25 | — | — | — | |
| Bi-level LSTMYear=2022, Pre-processing steps=Hierarchical LSTM2026.02 | 23.38 | — | — | — | |
| BiLSTM2026.03 | 23.76 | — | — | 4,349.31 | |
| TFSCL2026.03 | 24.56 | — | — | 13,899.48 | |
| BiLSTMYear=2018, Pre-processing steps=Feature sel.; norm; RUL target2026.02 | 24.86 | — | — | — | |
| BiLSTM2026.03 | 24.86 | — | — | 5,430 | |
| FedLSTMYear=2022, Pre-processing steps=Federated learning; norm2026.02 | 25.1 | — | — | — | |
| LSTMYear=2017, Pre-processing steps=Data normalization; RUL target2026.02 | 28.17 | — | — | — | |
| CNNYear=2016, Pre-processing steps=Data normalization; RUL target2026.02 | 29.15 | — | — | — | |
| QAQLstatus=proposed2026.06 | 29.43 | — | — | — | |
| Quantum VQL2026.06 | 38.76 | — | — | — | |
| Quantum DQN2026.06 | 39.81 | — | — | — | |
| Quantum AOA2026.06 | 43.2 | — | — | — | |
| Transformer2026.06 | 46.41 | — | — | — | |
| Quantum Eigensolver2026.06 | 47.07 | — | — | — | |
| PPO2026.06 | 47.22 | — | — | — | |
| Deep Q-Network2026.06 | 47.4 | — | — | — | |
| Quantum SGD2026.06 | 47.97 | — | — | — | |
| SARSA2026.06 | 49.33 | — | — | — | |
| SGD2026.06 | 51.23 | — | — | — | |
| Quantum Decision Tree2026.06 | 51.25 | — | — | — | |
| Quantum LSTM2026.06 | 51.5 | — | — | — | |
| GRU2026.06 | 56.94 | — | — | — | |
| Bi-LSTM2026.06 | 61.78 | — | — | — |