In the evolving field of maintenance and reliability engineering, the organization of equipment into hierarchical structures presents both a challenge and a necessity, directly impacting the operational integrity of industrial facilities. This paper introduces an innovative approach employing machine learning, specifically Long Short-Term Memory (LSTM) models, to automate and enhance the creation and management of these hierarchies. By adapting techniques commonly used in natural language processing, the study explores the potential of LSTM models to interpret and predict relationships within equipment tags, offering a novel perspective on understanding facility design. This methodology involved character-wise tokenization of asset tags from approximately 29,000 entries across 50 upstream oil and gas facilities, followed by modeling these sequences using an LSTM-based recurrent neural network. The model's architecture capitalizes on LSTM's ability to learn long-term dependencies, facilitating the prediction of hierarchical relationships and contextual understanding of equipment tags.

翻译：在维护与可靠性工程这一不断发展领域，将设备组织成层级结构既是一项挑战，也是一种必要，直接影响工业设施的运营完整性。本文提出了一种创新方法，采用机器学习（特别是长短期记忆模型，LSTM）来自动化和增强这些层级结构的创建与管理。通过借鉴自然语言处理中常用的技术，本研究探索了LSTM模型解释和预测设备标签间关系的潜力，为理解设施设计提供了全新视角。该方法对来自50个上游油气设施的约29,000条资产标签进行字符级分词处理，随后基于LSTM递归神经网络对这些序列建模。该模型架构利用LSTM学习长期依赖关系的能力，促进了层级关系的预测与设备标签的上下文理解。