|
Title |
Development of a Predictive Model for Bridge Deck Condition Rating and Defect Index Using Various Machine Learning Algorithms
|
|
Authors |
이재현(Jae-Hyun Lee) ; 김우혁(Woo-Hyuk Kim) ; 민근형(Geun-Hyeong Min) ; 김우석(WooSeok Kim) |
|
DOI |
https://doi.org/10.4334/JKCI.2024.36.6.657 |
|
Keywords |
기계 학습; 예측 모델; 열화; 상태 등급; 결함도 지수 machine learning; predictive modeling; deterioration; condition rating; defect index |
|
Abstract |
This work aims to develop predictive models for the efficient maintenance of bridge decks using various machine learning algorithms, including random forest, XGBoost, k-NN, SVM, neural networks, LSTM, and GRU. The models predict the deck defect index and condition rating, with LSTM, random forest, and XGBoost showing superior performance in defect index prediction, and XGBoost, GRU, and LSTM excelling in condition rating prediction. The models were evaluated using MSE, RMSE, MAE, and other metrics, and cross-validation was conducted to assess overfitting risks. The results demonstrate that these models can contribute to more accurate and efficient bridge deck maintenance, enabling proactive interventions and optimizing maintenance costs. Future research will focus on incorporating additional factors, such as environmental impacts, to further enhance the predictive accuracy and practicality of the models.
|