PEFT-MuTS: Few-Shot RUL Prediction via Cross-Domain Pre-Trained Models
A new framework, PEFT-MuTS, aims to improve remaining useful life (RUL) prediction with limited data by leveraging cross-domain pre-trained time-series representation models. The approach challenges the conventional belief that knowledge transfer in RUL prediction is only possible among similar devices, showing that pre-training on large-scale cross-domain time series datasets can yield substantial benefits. The framework includes an independent feature tuning network and a meta-variable-based low-rank multivariate fusion mechanism. The paper is available on arXiv (2601.22631) and was announced as a replace-cross type, indicating a revision. This research addresses the long-standing constraint of requiring large amounts of degradation data for data-driven RUL prediction, offering a parameter-efficient solution for few-shot scenarios.
Key facts
- PEFT-MuTS is a Parameter-Efficient Fine-Tuning framework for few-shot RUL prediction.
- It is built on cross-domain pre-trained time-series representation models.
- The framework demonstrates that knowledge transfer can occur across different devices, not just similar ones.
- It uses an independent feature tuning network and a meta-variable-based low-rank multivariate fusion mechanism.
- The paper is available on arXiv with ID 2601.22631.
- The announcement type is replace-cross, indicating a revised version.
- The approach aims to overcome the need for large amounts of historical degradation data.
- The research is relevant to predictive maintenance and industrial applications.
Entities
Institutions
- arXiv