Scholarlink

<p>This study investigates the application of machine learning (ML) techniques for structural health monitoring (SHM) of the KW51 steel railway bridge in Leuven, Belgium, using vibration-based sensor data to detect structural anomalies. The KW51 bridge, which is 12.4&#xa0;m wide and 115&#xa0;m long, is equipped with 12 uniaxial accelerometers placed on its deck and arches to collect acceleration data. The data is processed to identify important features which serve as indicators of potential structural damage with imbalance handling by SMOTE-TOMIK Technique. Four supervised ML algorithms Random Forest, K-Nearest Neighbors, Naive Bayes, and Decision Trees are applied to classify the sensor data into “damaged” and “undamaged” categories. Among these, the Random Forest classifier outperforms the others, achieving an accuracy of 97.86%, strong specificity/ “undamaged” recall (98.42%), and the best F1 for the Damage class (0.454) with Damage recall (sensitivity) 60.5% and AUC score of 0.95, demonstrating its strong ability to differentiate between damaged and undamaged states. K-Fold cross validation supports the robustness and generalizability of this approach. The findings highlight the effectiveness of ML techniques in detecting structural abnormalities early, which is essential for preventive maintenance and ensuring the safety of critical infrastructure such as bridges. This study confirms the viability of using vibration-based sensor data and supervised ML algorithms to monitor and detect significant structural issues in complex civil engineering structures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging vibration sensor data and machine learning for effective structural health monitoring of the KW51 bridge

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • Ashuvendra Singh,
  • Smita Kaloni

摘要

This study investigates the application of machine learning (ML) techniques for structural health monitoring (SHM) of the KW51 steel railway bridge in Leuven, Belgium, using vibration-based sensor data to detect structural anomalies. The KW51 bridge, which is 12.4 m wide and 115 m long, is equipped with 12 uniaxial accelerometers placed on its deck and arches to collect acceleration data. The data is processed to identify important features which serve as indicators of potential structural damage with imbalance handling by SMOTE-TOMIK Technique. Four supervised ML algorithms Random Forest, K-Nearest Neighbors, Naive Bayes, and Decision Trees are applied to classify the sensor data into “damaged” and “undamaged” categories. Among these, the Random Forest classifier outperforms the others, achieving an accuracy of 97.86%, strong specificity/ “undamaged” recall (98.42%), and the best F1 for the Damage class (0.454) with Damage recall (sensitivity) 60.5% and AUC score of 0.95, demonstrating its strong ability to differentiate between damaged and undamaged states. K-Fold cross validation supports the robustness and generalizability of this approach. The findings highlight the effectiveness of ML techniques in detecting structural abnormalities early, which is essential for preventive maintenance and ensuring the safety of critical infrastructure such as bridges. This study confirms the viability of using vibration-based sensor data and supervised ML algorithms to monitor and detect significant structural issues in complex civil engineering structures.