发表机构
Comtrade 360 d.o.o.; Comtrade AI GmbH; LE-Tehnika d.o.o.; Alma Mater Studiorum – Università di Bologna（康创360有限责任公司; 康创人工智能有限公司; LE技术有限责任公司; 博洛尼亚大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对工业科学领域小数据场景，提出FSD-RM范式结合NAS优化模型容量与输入维度，用于制冷机寿命预测，可在降本减复杂度的同时实现有竞争力的预测性能。
AI 中文摘要
大规模预训练时间序列模型通过大规模预训练和任务无关的表示学习取得了优异性能，但它们依赖于工业和科学领域往往缺乏的丰富多样的数据。因此，我们提出FSD-RM（小数据表示模型家族）范式，作为有限领域遥测数据的实用替代方案。不依赖大规模预训练，我们聚焦于使用成熟的编码器架构（CNN1D、LSTM、GRU、Transformer）进行容量可控的表示学习，这些架构因适用于小数据场景且可解释而被选用。这些编码器在多变量遥测数据上进行无监督训练，并被整合到用于下游寿命预测的两阶段流程中。为系统研究数据约束下的架构权衡，我们采用维度感知神经架构搜索（NAS），以联合优化模型容量和输入维度。在制冷机遥测数据上的实验表明，所提方法在实现有竞争力的预测性能的同时，降低了训练成本和模型复杂度。本研究的贡献在于，将成熟的表示学习技术整合到一个连贯的、由NAS驱动的框架中，该框架针对小数据场景量身定制，具有明确定义的参数设置和设计选择。结果表明，当应用适当的归纳偏置和容量控制时，无需大规模预训练即可实现有效的表示学习。
英文摘要
Large-scale pretrained time-series models achieve strong results through large-scale pretraining and task-agnostic representation learning, but they rely on abundant, diverse data that industrial and scientific domains often lack. We therefore propose the FSD-RM (Family of Small-Data Representation Models) paradigm as a practical alternative for limited, domain-specific telemetry. Rather than relying on large-scale pretraining, we focus on capacity-controlled representation learning using established encoder architectures (CNN1D, LSTM, GRU, Transformer), selected for their suitability in small-data settings and interpretability. These encoders are trained unsupervised on multivariate telemetry data and integrated into a two-stage pipeline for downstream lifetime prediction. To systematically examine architectural trade-offs under data constraints, we employ \textbf{dimension-aware neural architecture search (NAS)} to jointly optimize model capacity and input dimensionality. Experiments on cryocooler telemetry show that the proposed approach achieves competitive predictive performance while reducing training cost and model complexity. The contribution lies in combining established representation learning techniques within a coherent, NAS-driven framework tailored to small-data regimes, with explicitly defined parameter settings and design choices. The results indicate that effective representation learning can be achieved without large-scale pretraining when appropriate inductive bias and capacity control are applied.
Comments48 pages
Journal refReliability Engineering and System Safety 277 (2027) 113105
DOI：10.1016/j.ress.2026.113105
