摘要
为了提高路面性能预测的精度及优化路面养护方案,针对路面性能预测中不同数学模型特点和使用现状,考虑影响模型预测精度的主要因素,对模型预测精度进行对比分析研究。采用灰关联和BP神经网络相组合的数学模型,对路面状况性能指数PCI进行预测研究,首先,基于灰关联理论分析PCI的相关影响因素;其次,选取关联度大的影响因素作为路面预测模型的属性数据,基于BP神经网络进行性能预测;最后,将组合模型应用于某水泥路面的性能预测实践,对比分析组合模型与灰色预测模型GM(1,1)的预测结果。结果表明:基于灰关联度和BP神经网络组合的预测模型具有更高的精度和准确性,更广的适用性和可行性,是为长寿命期内的农村公路预防性养护决策提供模型参考和决策依据的有效手段。
To improve the accuracy of pavement performance prediction and optimize pavement maintenance schemes,this study conducted a comparative analysis of model prediction accuracy by considering the characteristics and application status of different mathematical models in pavement performance prediction,as well as the main factors affecting prediction accuracy.A combined mathematical model of grey correlation and BP neural network was adopted to predict the Pavement Condition Index(PCI).First,the influencing factors of PCI were analyzed based on grey correlation theory.Second,the influencing factors with high correlation degrees were selected as attribute data for the pavement prediction model,and performance prediction was carried out using a BP neural network.Finally,the combined model was applied to the performance prediction of a cement pavement,and the prediction results of the combined model were compared with those of the grey prediction model GM(1,1).The results show that the prediction model based on the combination of grey correlation degree and BP neural network has higher accuracy and precision,broader applicability and feasibility,and provides an effective model reference and decision-making basis for preventive maintenance decisions of rural roads with long service life.
作者
杨娥
谭羽彤
赵梓雯
邱逸
曾惠超
YANG E;TAN Yu-tong;ZHAO Zi-wen;QIU Yi;ZENG Hui-chao(School of Future Transportation,Guangzhou Maritime University,Guangzhou Guangdong 510700,China)
出处
《广州航海学院学报》
2025年第3期52-56,共5页
Journal of Guangzhou Maritime University
基金
大学生创新创业训练计划项目(S202501008)
广东省教育厅项目(2023KTSCX111)
广州市教育局项目(J202503038)。
关键词
道路工程
路面性能
灰关联
BP神经网络
预测
road engineering
pavement performance
grey correlation
BP neural network
prediction
