LAST-RAG: Literature-Anchored Retrieval for Degradation Model Selection

A new method called Literature-Anchored Stochastic Trajectory Retrieval-Augmented Generation (LAST-RAG) is proposed to improve stochastic process selection for remaining useful life (RUL) estimation. Existing model selection methods rely on statistical fit of observed health indicator (HI) trajectories, which can be inconsistent with underlying degradation mechanisms under short observation windows or high noise. LAST-RAG addresses this by combining observed HI trajectories with domain-specific context from a local evidence bank, hierarchically conditioning candidate degradation models based on theoretical and mechanical evidence. The method also incorporates Rule-based Confidence Reasoning with Uncertain State (RCRUS) to handle uncertainty. The paper is available on arXiv under identifier 2605.17902.