Scholarlink

<p>Structural health monitoring (SHM) is a crucial aspect of civil engineering safety and serviceability, enabling the early detection of damage. This study proposed a framework for SHM using vibration data from the ASCE benchmark building together with Speeded-Up Robust Features (SURF). The ASCE benchmark structure is modelled in ANSYS environment. The structure is subjected to collection of time-history acceleration under both undamaged and damaged conditions, and converted to frequency domain scalograms images. The Bag of Features (BoF) approach is employed to extract the SURF from the registered images. The resulting feature vectors are used to input to the machine learning classifiers (k-Nearest Neighbours (k-NN), and Support Vector Machine (SVM), that classify the undamaged and different damage conditions. Among the combinations of SURF with a SVM classifier, there is greater classification accuracy. Furthermore, the results are then compared with those obtained using the statistical features. The findings provide evidence that the SURF-based feature representation does accurately collect information about structural damage. The proposed framework demonstrates that integrating time-frequency visual representations of images and machine learning to identify structural damage accurately is a viable option.</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

Machine learning-based framework for damage identification of an ASCE benchmark building based on speeded-up robust features

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Kakarla Hussain,
  • Gorantla Nipun,
  • C. H. Ajay,
  • A. George Fernandez Raj,
  • B. Kesava Rao,
  • N. Lingeshwaran

摘要

Structural health monitoring (SHM) is a crucial aspect of civil engineering safety and serviceability, enabling the early detection of damage. This study proposed a framework for SHM using vibration data from the ASCE benchmark building together with Speeded-Up Robust Features (SURF). The ASCE benchmark structure is modelled in ANSYS environment. The structure is subjected to collection of time-history acceleration under both undamaged and damaged conditions, and converted to frequency domain scalograms images. The Bag of Features (BoF) approach is employed to extract the SURF from the registered images. The resulting feature vectors are used to input to the machine learning classifiers (k-Nearest Neighbours (k-NN), and Support Vector Machine (SVM), that classify the undamaged and different damage conditions. Among the combinations of SURF with a SVM classifier, there is greater classification accuracy. Furthermore, the results are then compared with those obtained using the statistical features. The findings provide evidence that the SURF-based feature representation does accurately collect information about structural damage. The proposed framework demonstrates that integrating time-frequency visual representations of images and machine learning to identify structural damage accurately is a viable option.