基于精准病害数据与多目标优化的路面养护决策一体化方案研究
摘 要:病害测量不准、养护等级主要依靠经验判定,以及预算分配缺少量化依据,是传统路面养护决策中的几项具体问题。本文以YOLO系列模型输出的病害数据为输入,梳理已有预测与决策方法,将数据预处理、路面性能预测和养护方案优化衔接起来。病害参数与环境、交通等信息经标准化后进入随机森林预测模型,所得性能变化趋势再用于多目标养护方案筛选,由此形成“数据-预测-决策”流程。在某省国省道工程中,应用该方案后,单位里程养护成本由27元/米降至24元/米,按这两个数值计算降幅约为11.11%;养护决策方案制定效率提高60%,性能较差路段的占比从4.2%降至1.8%。这一案例说明,将病害量化数据与性能预测结果共同用于方案选择,有助于在预算约束下改善养护安排。
关键词:机器学习;路面病害;养护决策;性能预测;多目标优化
正文
可下载并阅读全文PDF,请按照本文版权许可使用。
Download the full text PDF for viewing and using it according to the license of this paper.
Research on an Integrated Decision-Making Scheme for Pavement Maintenance Based on Accurate Distress Data and Multi-Objective Optimization
Abstract: Pavement maintenance planning depends on how measured distress is translated into treatment choices and budget allocations. Conventional practice has difficulties with measurement accuracy, experience-based grading, and the quantitative basis for allocating funds. Following a review of prediction and decision methods, this paper brings these tasks into a data-prediction-decision sequence. YOLO-series models supply the distress measurements. These are standardized with environmental, traffic, and other relevant data for an optimized random forest, whose performance forecasts are used in multi-objective maintenance optimization. The sequence was applied in a provincial project covering national and provincial highways. After implementation, maintenance cost was CNY 24/m, compared with CNY 27/m previously; this corresponds to a reduction of approximately 11.11%. Maintenance plan preparation became 60% more efficient, while the share of poorly performing road segments fell from 4.2% to 1.8%. Within this case, linking distress measurements and performance forecasts provided a workable basis for maintenance planning with a limited budget.
Keywords: Machine Learning; Pavement Distress; Maintenance Decision-Making; Performance Prediction; Multi-Objective Optimization
参考文献
- 交通运输部. 2023 年交通运输行业发展统计公报 [R]. 北京:交通运输部,2024.
- 李邵华。基于图像的沥青公路路面养护决策研究 [D]. 武汉:武汉理工大学,2023.
- 徐鹏,祝轩,姚丁,等。沥青路面养护智能检测与决策综述 [J]. 中南大学学报 (自然科学版),2021,52 (7):2099-2117.
- 周乐东,王骄。基于路面病害修补还原换算的养护决策分析 [J]. 公路与汽运,2023,(2):135-138.
- 汪凡,杨建喜。基于 BP 神经网络的路面使用性能预测模型 [J]. 中外公路,2009,29 (1):86-89.
- 董元帅,周绪利,侯芸,等。基于寿命周期的沥青路面预养护时机决策优化 [J]. 公路,2020,65 (4):325-331.
- 周红宇,冯晓,周卓,等。基于 WebGIS 的道路养护可视化系统的设计与开发 [J]. 重庆交通大学学报 (自然科学版),2012,31 (2):274-278.
- 交通运输部公路科学研究院。公路技术状况评定标准:JTG 5210-2018 [S]. 北京:人民交通出版社股份有限公司,2018.
- 李昊昌.考虑多目标需求的西北地区沥青路面养护决策理论和方法研究[D].兰州交通大学,2024.DOI:10.27205/d.cnki.gltec.2024.000945.
- 张高伟.公路沥青路面病害识别与养护决策研究[J].产业创新研究,2024,(04):132-134.
- 黄九达,陈涛,洪盛祥,等.高速公路沥青路面养护现状与决策分析[J].河南科技,2021,40(36):73-76.

