Scholarlink

<p>This study presents a real-time Structural Health Monitoring (SHM) framework that leverages low-cost, consumer-grade sensors (e.g., smartphones and ESP32 devices) integrated with a Python-based software platform for continuous infrastructure assessment. The system performs streaming data acquisition via HTTP from distributed accelerometer nodes and applies a modular signal processing pipeline including bandpass filtering, feature extraction, and hybrid anomaly detection. Three key structural features, dominant frequency, RMS and damping ratio, are computed in real-time to characterize vibrational behavior. A hybrid detection approach, combining Z-score-based thresholds with an Isolation Forest algorithm, an unsupervised machine learning method, enables robust anomaly identification. Visual tools such as STL decomposition, CWT scalograms, and PCA projections support interpretability and verification. Ambient vibration testing and simulated anomaly injection validate system performance, highlighting its capability to detect both gradual and sudden structural changes. Designed for usability, portability, and cost-efficiency, this SHM framework demonstrates strong potential for practical deployment in infrastructure monitoring, education, and post-disaster assessment.</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

Real-time structural health monitoring using IoT-enabled affordable sensors and unsupervised anomaly detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月4日

查看全文
  • Mohammad Reza Bagerzadeh Karimi

摘要

This study presents a real-time Structural Health Monitoring (SHM) framework that leverages low-cost, consumer-grade sensors (e.g., smartphones and ESP32 devices) integrated with a Python-based software platform for continuous infrastructure assessment. The system performs streaming data acquisition via HTTP from distributed accelerometer nodes and applies a modular signal processing pipeline including bandpass filtering, feature extraction, and hybrid anomaly detection. Three key structural features, dominant frequency, RMS and damping ratio, are computed in real-time to characterize vibrational behavior. A hybrid detection approach, combining Z-score-based thresholds with an Isolation Forest algorithm, an unsupervised machine learning method, enables robust anomaly identification. Visual tools such as STL decomposition, CWT scalograms, and PCA projections support interpretability and verification. Ambient vibration testing and simulated anomaly injection validate system performance, highlighting its capability to detect both gradual and sudden structural changes. Designed for usability, portability, and cost-efficiency, this SHM framework demonstrates strong potential for practical deployment in infrastructure monitoring, education, and post-disaster assessment.