HyperLLM: Hypergraph-Enhanced Large Language Model for Remaining Useful Life Prediction

Accurately predicting the remaining useful life (RUL) of aircraft engines is crucial for ensuring flight safety. However, the degradation process of aircraft engines is highly complex and difficult to model accurately. In addition, multivariate sensor signals exhibit strong coupling relationships, which hinder the effective extraction and utilization of degradation information.In recent years, large language models (LLMs) have achieved remarkable performance in numerous tasks owing to their powerful understanding and reasoning capabilities, thus offering a new approach for modeling complex degradation processes.Therefore, this article proposes a hypergraph-enhanced LLM method for RUL prediction. The method first constructs an adaptive hypergraph to fuse multisource degradation information and extracts topological spatial degradation features via hypergraph convolution. Simultaneously, a degradation-aware temporal embedding module is introduced to capture temporal degradation features, and a lightweight adapter is employed to achieve multidimensional feature fusion. On this basis, task-embedding prompts are constructed to guide efficient adaptation with the LLM parameters frozen, thereby preserving the prior knowledge of the LLMs and enhancing prediction performance. Experimental results on the NASA commercial modular aero-propulsion system simulation (C-MAPSS) and new commercial modular aero-propulsion system simulation (N-CMAPSS) datasets demonstrate that the proposed method exhibits strong prediction accuracy and generalization capability.