铁道科学与工程学报2026，Vol.23Issue(8)：4055-4065,11.DOI:10.19713/j.cnki.43-1423/u.T20251621
基于改进两阶段卡尔曼滤波的时变参数及未知激励识别
Identification of time-varying parameters and unknown inputs based on an improved two-stage Kalman filter method
摘要
Abstract
The Kalman filter(KF)serves as an efficient tool for reconstructing structural states,yet its implementation requires precise knowledge of both structural parameters and external excitations.More critically,classical KF exhibits inherent limitations in addressing the issues of time-varying parameters caused by material aging or damage.To achieve real-time identification of structural time-varying parameters under unknown excitation conditions,this study proposed a novel online two-stage Kalman filter with unknown inputs(OTKF-UI)that incorporates a real-time tracking matrix.In stage 1,by fusing acceleration and displacement measurements,structural states and excitations were reconstructed using the method previously developed by the authors.Then,the structural parameter identification scheme combined with the sensitivity matrix was derived within the Kalman filtering framework in stage 2.To tackle the limitations of the above-mentioned approach in identifying time-varying parameters,a real-time tracking matrix was introduced and obtained through an optimization-based algorithm.Numerical cases,including scenarios of gradual deterioration and abrupt damage,assessed the feasibility of the proposed method,while experimental tests were conducted to evaluate its practical effectiveness.The identified results indicate that the maximum error of time-varying parameters remained below 4%in numerical simulations across both pre-and post-damage states,and the maximum error of the experimental validation is under 5%for the damaged story.All identified time-invariant parameters consistently fell within acceptable bounds.Moreover,unknown external excitations can be successfully identified concurrently in both numerical cases and experimental tests.The proposed approach accurately identifies time-varying parameters and unknown inputs with only partial measurements across diverse damage scenarios.Featuring a straightforward derivation process and yielding precise identification results,this approach demonstrates engineering value for providing a theoretical foundation for health-state assessment and maintenance planning during structural service life.关键词
时变参数/未知激励/灵敏度矩阵/实时追踪矩阵/两阶段卡尔曼滤波Key words
time-varying parameters/unknown inputs/sensitive matrix/real-time tracking matrix/two-stage Kalman filter分类
建筑与水利引用本文复制引用
张肖雄,贺佳,沈炼..基于改进两阶段卡尔曼滤波的时变参数及未知激励识别[J].铁道科学与工程学报,2026,23(8):4055-4065,11.基金项目
国家自然科学基金资助项目(52278305) （52278305）
