AI 中文总结
本文针对LED照明系统，提出两阶段监测设计与预算型CBM框架，采用EnKF结合伽马过程预测更新退化状态，通过案例验证了策略优化的有效性。
AI 中文摘要
发光二极管（LED）灯具在运行期间会逐渐退化，导致工作面（WP）照度降低，增加室内照明不足的风险。由于灯具安装后难以直接测量单个灯具的退化状态，本文从稀疏的原位WP照度测量值中估计潜在的灯具退化状态。本文提出了一种两阶段监测设计与预算型基于状态的维护（CBM）优化框架。在设计阶段，采用Radiance模拟构建线性高斯观测模型，将潜在的灯具退化状态映射到WP照度测量值；然后采用可识别性约束的D-最优设计，为单个时刻的逆状态估计选择信息丰富的WP测量点参考布局。在运行阶段，测量设计确定访问计划和每次访问的活动参考测量点，同时集合卡尔曼滤波（EnKF）将这些稀疏测量值与非齐次伽马过程退化预测相结合，以更新潜在的灯具退化状态估计值。后验预测失效概率被用作预防性更换的风险度量。在监测预算约束下，所得的运行监测与CBM策略联合选择访问计划、活动参考测量点和风险阈值，以最小化预期总停机时间和更换成本。针对办公照明区域的案例研究验证了设计阶段测量布局选择、运行阶段测量设计以及预算型CBM策略优化的有效性。
英文摘要
Light-emitting diode (LED) luminaires degrade gradually during operation, reducing working plane (WP) illuminance and increasing the risk of inadequate indoor lighting. Because individual luminaire degradation states are difficult to measure directly after installation, this paper estimates latent luminaire degradation states from sparse in situ WP illuminance measurements. A two-stage monitoring design and budgeted condition-based maintenance (CBM) optimization framework is proposed. At the design stage, Radiance simulations are used to construct a linear-Gaussian observation model that maps latent luminaire degradation states to WP illuminance measurements. An identifiability-constrained D-optimal design then selects an informative reference layout of WP measurement points for inverse state estimation at a single epoch. At the runtime stage, the measurement design determines the visit schedule and the active reference measurement points for each visit, while an ensemble Kalman filter (EnKF) combines these sparse measurements with nonhomogeneous gamma process degradation predictions to update latent luminaire degradation state estimates. Posterior predictive failure probabilities are used as the risk metric for preventive replacement. Under a monitoring budget, the resulting runtime monitoring and CBM policy jointly selects the visit schedule, active reference measurement points, and risk threshold to minimize expected total downtime and replacement cost. A case study on an office lighting zone demonstrates design-stage measurement layout selection, runtime-stage measurement design, and budgeted CBM policy optimization.
Comments33 pages, 11 figures
