【摘要】 针对航天器热控单相流体回路直接采集的数据白噪声多、波动严重、难以直接提取有效信息的问题，文章提出赫斯特(Hurst)指数驱动的时序分解-滤波-预测三级在轨数据处理方法。通过Hurst指数量化数据特性，结合时间序列分解分离趋势项、周期项与残差项，并应用傅里叶变换滤除白噪声与离散突变，构建调节分数积分模型进行未来趋势预测。经在轨数据分析，表明：航天器热控在轨数据会因为轨道运动引发的温度变化呈现周期性波动；泵转速呈现强回归性，液位呈现线性泄漏趋势，年泄漏量约0.6 mm;压力与液位显著线性相关。该方法可为航天器热控系统在轨安全运行提供支持。更多还原

【Abstract】 To address the problem that data directly acquired from single-phase fluid loops in spacecraft thermal control systems contain substantial white noise and exhibit severe fluctuations, which makes it difficult to directly extract effective information, a three-stage on-orbit data processing method based on Hurst-exponent-driven time-series decomposition, filtering and prediction is proposed in this paper. Hurst exponents are used to quantify the characteristics of the data. Time-series decomposition is the performed to separate the trend, periodic, and residual components. Fourier transform-based filtering is applied to attenuate white noise and discrete abrupt variations. An adjustable fractional-order integration model is then constructed fpr future trend prediction. Analysis of on-orbit data indicates that spacecraft thermal control data exhibits periodic fluctuations due to temperature variation cycles induced by orbital motion. The pump rotational speed demonstrates strong mean-reverting characteristics, while the liquid level shows a linear leakage trend, with an annual leakage rate of approximately 0.6mm. In addition, a high-confidence linear relationship is observed between pressure and liquid level. This method provides support for the safe on-orbit operation of spacecraft thermal control systems.更多还原