Diagnosis algorithms for indirect bridge health monitoring via an optimized AdaBoost-linear SVM

Yifu Lan, Youqi Zhang, Weiwei Lin*

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

1 Citation (Scopus)
25 Downloads (Pure)


A data-driven approach based on Optimized AdaBoost-Linear SVM is proposed to indicate the bridge damage using only raw vibration signals received from a vehicle passing over the bridge. To enable Linear SVM as an effective component learner in AdaBoost and to achieve its best generalization performance, an optimizing strategy is designed to modify its configuration. Laboratory experiments are conducted to establish the dataset employing a steel beam and a scale truck model with an engine. The present algorithm learns to identify bridge health states by feeding training data, and its performance is assessed using the testing dataset. Principal Component Analysis (PCA) as a dimension reduction technique is utilized to visualize the identification results. From the dataset on diverse health states, the proposed strategy can identify the bridge damages effectively and provide better generalization performance than other commonly used algorithms. When compared to other algorithms such as SVM and Random Forest, it improves result accuracy by 5% to 16.7%. The experimental results also indicate that the vehicle-based indirect Structural Health Monitoring (SHM) framework can be equally effective as the direct SHM systems, and suggest the potentials of achieving automatic, robust and practical SHM models in the future.
Original languageEnglish
Article number115239
Number of pages14
JournalEngineering Structures
Publication statusPublished - 15 Jan 2023
MoE publication typeA1 Journal article-refereed


  • Drive-by bridge inspection
  • Vehicle-bridge interaction
  • AdaBoost SVMs
  • Health state identification
  • Data-driven SHM


Dive into the research topics of 'Diagnosis algorithms for indirect bridge health monitoring via an optimized AdaBoost-linear SVM'. Together they form a unique fingerprint.

Cite this