考虑非齐次退化与资金折现的桥梁维修决策方法

李光涛; 刘鲁齐; 时鹏; 兰成明

doi:10.3724/j.gyjzG26031401

考虑非齐次退化与资金折现的桥梁维修决策方法

doi: 10.3724/j.gyjzG26031401

1. 北京科技大学未来城市学院城镇化与城市安全研究院, 北京 100083;
2. 北京科技大学国家材料服役安全科学中心, 北京 102206

基金项目:

国家重点研发计划（2023YFC3805700）。

详细信息

作者简介:
李光涛,博士研究生,主要从事桥梁智慧运维研究。

通讯作者:
兰成明,教授,主要从事桥梁智慧运维研究,lanchengming@ustb.edu.cn。

计量
- 文章访问数: 4
- HTML全文浏览量: 1
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2026-03-14
- 网络出版日期: 2026-06-06

A Bridge Maintenance Decision-Making Method Considering Non-Homogeneous Deterioration and Cost Discounting

1. Research Institute of Urbanization and Urban Safety, School of Future Cities, University of Science and Technology Beijing, Beijing 100083, China;
2. National Center for Materials Service Safety, University of Science and Technology Beijing, Beijing 102206, China

摘要

摘要: 面向桥梁智慧运维中的长期维修决策优化需求，针对有限时域条件下桥梁状态退化具有时变性、维修成本存在资金折现且长期回报传递困难等特点，构建了一种考虑非齐次退化与折现效应的全寿命周期维修决策模型。桥梁退化过程采用非齐次马尔科夫状态转移进行表征，以离散健康状态和维修动作为基础，将维护成本与风险成本统一纳入成本函数，并在决策过程中引入现金流折现。为克服常规强化学习方法难以有效处理有限时域阶段性决策需求、非齐次状态转移以及训练不稳定等问题，提出一种结合状态增广与反向课程学习的强化学习求解框架。其中，通过扩展状态表征增强策略对有限时域特征的识别能力；通过设置反向课程逐步扩展训练区间，提高策略学习的稳定性与收敛效率。算例结果表明，所提出方法能够有效适应桥梁非齐次退化下的有限时域维修决策问题，在策略性能及训练稳定性方面均表现出较好效果，可为桥梁全寿命周期运维管理中的维修方案制定提供方法支撑。
- 桥梁智慧运维 /
- 维修决策 /
- 非齐次马尔科夫过程 /
- 强化学习 /
- 反向课程学习
Abstract: To meet the demand for optimizing long-term maintenance decisions in intelligent bridge operation and maintenance， considering the characteristics of time-varying bridge deterioration under a finite horizon， financial discounting of maintenance costs， and difficulty in long-term reward propagation， a life-cycle maintenance decision model incorporating non-homogeneous deterioration and discounting effects was developed. The bridge deterioration process was characterized by non-homogeneous Markov state transitions. Based on discrete health states and maintenance actions， maintenance costs and risk costs were integrated into a unified cost function， while cash-flow discounting was introduced into the decision-making process. To address the limitations of conventional reinforcement learning methods in handling finite-horizon stage-wise decision tasks， non-homogeneous state transitions， and unstable training， a reinforcement learning framework combining state augmentation and backward curriculum learning was proposed. Expanded state representation enhanced the policy’s capability to capture finite-horizon characteristics， while backward curriculum learning gradually extended the training interval to improve learning stability and convergence efficiency. Numerical results demonstrated that the proposed method effectively adapted to finite-horizon maintenance decision problems under non-homogeneous bridge deterioration and achieved favorable performance in both policy quality and training stability， thereby providing methodological support for maintenance planning in life-cycle bridge operation and maintenance management.
- intelligent bridge operation and maintenance /
- maintenance decision-making /
- non-homogeneous Markov process /
- reinforcement learning /
- backward curriculum learning

HTML全文

参考文献(22)

[1]	秦权. 桥梁结构的健康监测［J］. 中国公路学报，2000，13（2）：38- 43.
[2]	中华人民共和国交通运输部. 公路桥涵养护规范：JTG 5120—2021［S］. 北京：人民交通出版社股份有限公司，2021.
[3]	周绪红，张喜刚. 关于中国桥梁技术发展的思考［J］. Engineering，2019，5（6）：1120- 1130.
[4]	贺桥，雷家艳，高婧，等. 可持续发展的项目级桥梁多目标维修决策优化［J］. 厦门大学学报（自然科学版），2025，64（6）：1026- 1033.
[5]	DABOUS S ABU，ALKASS S. Managing bridge infrastructure under budget constraints:a decision support methodology［J］. Canadian Journal of Civil Engineering，2011，38（11）：1227- 1237.
[6]	刘纲，孙瑞卿，李琦，等. 基于序贯决策的桥梁多阶段维修加固策略优化方法［J］. 重庆大学学报，2026，49（1）：60- 69.
[7]	DU A，GHAVIDEL A. Parameterized deep reinforcement learning-enabled maintenance decision-support and life-cycle risk assessment for highway bridge portfolios［J］. Structural Safety，2022，97：102221.
[8]	GENG Z，ZHANG C，JIANG Y，et al. Integrating multi-source data for life-cycle risk assessment of bridge networks:a system digital twin framework［J］. Journal of Infrastructure Preservation and Resilience，2025，6（1）：1- 20.
[9]	戴理朝，康哲，陈瑞，等. 基于NSGA-Ⅲ的桥梁网络多目标维修决策优化研究［J］. 土木工程学报，2024，57（5）：41- 52.
[10]	SUTTON R S，BARTO A G. Reinforcement Learning：An Introduction［M］. Massachusetts:The MIT Press，2018.
[11]	吕颖钊，贺拴海. 在役桥梁承载力模糊可靠性的马尔科夫预测［J］. 长安大学学报（自然科学版），2005（4）：39- 43.
[12]	TAO Z，COROTIS R B，ELLIS J H. Reliability-based bridge design and life cycle management with markov decision processes［J］. Structural Safety，1994，16（1/2）：111- 132.
[13]	刘渝. 基于部分可观测马尔可夫决策过程的钢桥腐蚀维护策略优化研究［D］. 重庆：重庆大学，2025.
[14]	MORCOUS G. Performance prediction of bridge deck systems using markov chains［J］. Journal of Performance of Constructed Facilities，2006，20（2）：146- 155.
[15]	WEI S，BAO Y，LI H. Optimal policy for structure maintenance：A deep reinforcement learning framework［J］. Structural Safety，2020，83：101906.
[16]	王超. 马尔科夫链在桥梁状态预测中的研究与应用［D］. 北京：北京交通大学，2009.
[17]	SRIKANTH I，AROCKIASAMY M. Deterioration models for prediction of remaining useful life of timber and concrete bridges:a review［J］. Journal of Traffic and Transportation Engineering（English Edition），2020，7（2）：152- 173.
[18]	BOLUKBASI M，MOHAMMADI J，ARDITI D. Estimating the future condition of highway bridge components using national bridge inventory data［J］. Practice Periodical on Structural Design and Construction，2004，9（1）：16- 25.
[19]	TOLLIVER D，LU P. Analysis of bridge deterioration rates:a case study of the northern plains region［J］. Journal of the Transportation Research Forum，2012，50（2）：87- 100.
[20]	AGRAWAL A K，KAWAGUCHI A. Bridge element deterioration rates:project：C-01-51［R］. New York：Transportation Infrastructure Research Consortium New York State，Department of Transportation，2009.
[21]	ARROW K，CROPPER M，Gollier C，et al. Determining benefits and costs for future generations［J］. Science，2013，341：349- 350.
[22]	TREASURY H M. The green book：central government guidance on appraisal and evaluation［M］. Updated ed. London：OGL Press，2022.