航空学报 > 2026, Vol. 47 Issue (7): 432458-432458   doi: 10.7527/S1000-6893.2025.32458

融合多源信息的电液伺服机构健康评估

李文博, 周志杰(), 王兆强, 冯志超, 孙一杰, 张心怡   

  1. 火箭军工程大学 导弹工程学院,西安 710038
  • 收稿日期:2025-06-20 修回日期:2025-07-16 接受日期:2025-10-11 出版日期:2025-12-01 发布日期:2025-11-28
  • 通讯作者: 周志杰
  • 基金资助:
    国家自然科学基金(62573349); 国家自然科学基金(62203365); 国家自然科学基金(62203461); 航空科学基金(2023Z034053004); 陕西省科学技术协会青年人才培养计划(20230125); 中国博士后科学基金(2023M742843); 陕西省重点研发计划(2025CY-YBXM-102); 陕西省自然科学基础研究计划重点项目(2025JC-QYXQ-038); 流体动力基础件与机电系统全国重点实验室开放基金(GZKF-202430)

Health assessment for electro-hydraulic servomechanism with multi-source information fusion

Wenbo LI, Zhijie ZHOU(), Zhaoqiang WANG, Zhichao FENG, Yijie SUN, Xinyi ZHANG   

  1. College of Missile Engineering,Rocket Force University of Engineering,Xi’an 710038,China
  • Received:2025-06-20 Revised:2025-07-16 Accepted:2025-10-11 Online:2025-12-01 Published:2025-11-28
  • Contact: Zhijie ZHOU
  • Supported by:
    National Natural Science Foundation of China(62573349); Aeronautical Science Foundation of China(2023Z034053004); Young Talent Promotion Program of Shaanxi Association for Science and Technology(20230125); China Postdoctoral Science Foundation(2023M742843); Key Research and Development Program of Shaanxi(2025CY-YBXM-102); Key Projects of the Shaanxi Province Natural Science Foundation(2025JC-QYXQ-038); Open Foundation of the State Key Laboratory of Fluid Power and Mechatronic Systems(GZKF-202430)

摘要:

电液伺服机构(EHS)是航天设备中重要的执行机构,其健康状态的保持对于航天设备安全稳定飞行具有至关重要的作用。然而,现有的健康评估模型面临两大问题:一是电液伺服机构在运行过程中产生的大量履历文本信息难以利用,二是缺乏有效的履历文本信息与测试数据融合手段。为解决上述问题,提出一种基于自然语言处理证据推理规则的健康状态评估模型。首先,基于大规模自然语言处理(BERT)设计了履历文本信息提取策略;然后,分析EHS运行机理,构建EHS健康评估指标体系,提出维修等级和运行时间影响因子计算方法,进而构建融合多源信息的健康状态评估模型(ER-MIF)。为克服专家知识不确定性影响,构建健康评估模型参数优化模型,对模型参数进行优化确定最优模型参数,提高评估精度。最后,通过某型电液伺服机构评估案例和对比,验证了所提模型有效性。

关键词: 证据推理规则, 健康评估, 自然语言处理, 电液伺服机构, 诊断推理

Abstract:

Electro-Hydraulic Servomechanism (EHS) serves as critical actuators in aerospace equipment, and maintaining its health status is vital for ensuring the safe and stable operation of aerospace systems. However, existing health assessment models face two major challenges: first, the vast amount of historical text data generated during EHS operation is difficult to utilize effectively; second, there is a lack of efficient methods for integrating historical text data with test data. To address these issues, a health status assessment model is proposed based on natural language processing evidence inference rules. First, a historical text information extraction strategy is designed based on Bidirectional Encoder Representations from Transformers (BERT). Subsequently, by analyzing the operational mechanisms of EHS, a health assessment indicator system is constructed, and methods for calculating maintenance levels and operational time influence factors are proposed. On this basis, an Evidential Reasoning rule with Multi-source Information Fusion health status assessment model (ER-MIF) is developed. To mitigate the impact of expert knowledge uncertainty, a health assessment model parameter optimization model is established to determine optimal parameters and enhance assessment accuracy. Finally, the effectiveness of the proposed model was validated through an evaluation case study and comparative analysis of a specific electro-hydraulic servo mechanism.

Key words: evidential reasoning rule, health assessment, natural language processing, electro-hydraulic servomechanism, diagnostic reasoning

中图分类号: