基于TPE-CatBoost算法的混凝土路面错台预测研究

打开文本图片集

中图分类号：U416.22 文献标志码：A

Faulting prediction in concrete pavements based on the TPE-CatBoost algorithm

YANG Junxin，JIN Hu，DING Zhuang，XIAO Wei （School of Civil Engineering and Architecture，Anhui University of Science & Technology， Huainan ，China）

Abstract： Faulting in concrete pavements is a prevalent distress requiring precise prediction. This study proposes a hybrid tree-structured Parzen estimator （TPE）-categorical boosting（CatBoost）model，designated as TPE-CatBoost. Using faulting data retrieved from the Long-Term Pavement Performance （LTPP）database， eight features screened by the Boruta algorithm serve as input to the CatBoost model，with faulting severity as the target variable.TPE optimizes CatBoost hyperparameters to identify optimal configurations.Comparative analysis between random forest （RF）and XGBoost models demonstrates the superiority of the TPE-CatBoost： It achieves R2=0.913 （0.051 and O.O12 higher than TPE-RF and TPE-XGBoost，respectively），and significantly reduces time and cost versus traditional detection methods for concretepavement.The model exhibits strong generalization capabilities， providing a reliable basis for pavement maintenance planning.

Key words：concrete pavement； faulting prediction；LTPP；Boruta；CatBoost；Bayesian optimization

错台是指混凝土路面板块之间产生的相对竖向位移现象，其可能导致路面不平整、行车舒适性下降等问题，严重时可能引发交通事故[1]。（剩余7602字）