|
Title |
Machine Learning-Based Prediction of Time-Dependent Safety Grade of Road Facilities
|
|
Authors |
김민선(Minsun Kim) ; 박예원(Yewon Park) ; 왕길환(Gil Hwan Wang) ; 전종수(Jong-Su Jeon) |
|
DOI |
https://doi.org/10.4334/JKCI.2024.36.6.625 |
|
Keywords |
머신러닝; 시간 의존적 안전등급; 도로교량; NATM 터널; 옹벽 machine learning; time-dependent safety grade; road bridges; NATM tunnels; retaining walls |
|
Abstract |
Aging road facilities experience a reduction in their service life. Without appropriate maintenance, the performance of aging structures can deteriorate, diminishing their utility. Predicting the safety performance of these structures with time characteristics is an effective strategy for maintenance planning and can appropriately pre-determine the duration and cost for their safety performance. This research predicts the safety grade of domestic road facilities including bridges, tunnels, and retaining walls. A database for each facility type was constructed to perform the safety performance evaluations, detailed safety inspections, and precision safety diagnostics. Ordinary bridges, NATM tunnels, and reinforced concrete retaining walls were selected since they are the most common structural types in the database. The safety grade prediction model was developed based on machine learning techniques considering the service life of the structures. Three regression-based machine learning algorithms were used to develop an optimal prediction model: decision tree, random forest, and extreme gradient boosting (XGBoost). The accuracy and confusion matrix for the three machine learning models were compared. In the comparative results, the XGBoost model resulted in the highest accuracy of 100 % for the road bridges, road tunnels, and retaining walls, effectively reflecting the time characteristics.
|