ABSTRACT

Nowadays, probabilistic prediction models commonly represented by Markov Chain Process (MCP) attract more attention in pavement management systems. Additionally, roughness condition indices provide a broad judgment. One element that contributes the most to the roughness progression is the initial roughness of pavements which has not been considered in the prediction models developed by MCP. On the other hand, the prediction results of MCP, which address the whole pavement network as an average, are inconvenient to be deployed in decision-making programmes. This paper utilised MCP to forecast pavement roughness regarding its initial value. International Roughness Index (IRI), extracted from the long-term pavement performance (LTPP) database, was selected to be analyzed. Based on M&R history, four major families, including 1770 pavement sections in total, were introduced. The prediction process of MCP was modified led to the direct prediction of IRI values instead of the deteriorated portion of the pavement sections. A framework, which had derived from the concept of Master Curves, is proposed to pave the way for the optimisation programmes to be applied to the MCP results. The Root Mean Squared Errors (RMSE) of the composed master curves were significantly low; the average RMSE was 0.00675.

摘要

如今，以马尔可夫链过程（MCP）为代表的概率预测模型在路面管理系统中越来越受到关注。此外，粗糙度条件指数提供了广泛的判断。对粗糙度进展贡献最大的一个因素是路面的初始粗糙度，这在 MCP 开发的预测模型中没有考虑。另一方面，MCP 的预测结果以整个路面网络为平均值，不便于在决策程序中部署。本文利用 MCP 预测路面粗糙度的初始值。选择从长期路面性能（LTPP）数据库中提取的国际粗糙度指数（IRI）进行分析。基于M&R历史，四大家族，共引进路面断面1770个。修改了 MCP 的预测过程，直接预测 IRI 值，而不是路面部分的恶化部分。提出了一个源自主曲线概念的框架，为将优化程序应用于 MCP 结果铺平了道路。组合主曲线的均方根误差 (RMSE) 显着降低；平均 RMSE 为 0.00675。组合主曲线的均方根误差 (RMSE) 显着降低；平均 RMSE 为 0.00675。组合主曲线的均方根误差 (RMSE) 显着降低；平均 RMSE 为 0.00675。