Few-Shot Rotating Machinery RUL Prediction Based on Reptile Framework and Transformer

The prediction of the remaining useful life (RUL) of rotating machinery is important in Prognostics and Health Management (PHM) of industrial equipment. Traditional RUL prediction methods rely on large amounts of labeled data, making it challenging to meet the prediction needs in data-scarce situations. This paper proposes a Transformer-based meta-learning algorithm for few-shot RUL prediction of rotating machinery. The method constructs a meta-learner based on Transformer and conducts meta-training under the Reptile framework, enabling the model to learn general knowledge across multiple tasks and rapidly adapt to new tasks with limited samples. Experiments on the CMAPSS dataset demonstrate that this method performs well under few-shot conditions.