This library contains a collection of unsupervised domain adaption algorithms for RUL estimation. They are provided as LightningModules to be used in PyTorch Lightning.

Currently, five approaches are implemented, including their original hyperparameters:

• LSTM-DANN by Da Costa et al. (2020)
• ADARUL by Ragab et al. (2020)
• LatentAlign by Zhang et al. (2021)
• TBiGRU by Cao et al. (2021)
• Consistency-DANN by Siahpour et al. (2022)

Three approaches are implemented without their original hyperparameters:

• ConditionalDANN by Cheng et al. (2021)
• ConditionalMMD by Cheng et al. (2021)
• PseudoLabels as used by Wang et al. (2022)

This includes the following general approaches adapted for RUL estimation:

• Domain Adaption Neural Networks (DANN) by Ganin et al. (2016)
• Multi-Kernel Maximum Mean Discrepancy (MMD) by Long et al. (2015)

Each approach has an example notebook which can be found in the examples folder.

This library is pip-installable. Simply type:

pip install rul-adapt

Contributions are always welcome. Whether you want to fix a bug, add a feature or a new approach, just open an issue and a PR.