Scholarlink

<p>Working as the main load-carrying members, steel beams are widely used in many industries. Local damages can exist on the steel beams during the long service life, which may result in huge accidents. Electromechanical impedance (EMI) technology is effective at detecting damage, but its accuracy is affected by ambient temperature. Temperature fluctuations alter the performance of the steel and the sensor, causing a signal shift that can lead to false detection of damage. This study investigates temperature effects on EMI signals through theoretical and experimental analysis. A neural network was proposed with three input features extracted from the sensors and the EMI signals to estimate the crack length under ambient temperatures. Experimental results demonstrate that, within a temperature range of −10 °C to 60 °C, the average relative error for damage severity estimation is 2.89 %. The neural network outperforms the Hilbert transform-based method, offering potential for wide application in EMI-based structural health monitoring.</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

Temperature compensation for electromechanical impedance-based damage detection of steel beams using neural network

  • 期刊论文

  • 发表日期：

    2025年11月5日

查看全文
  • Ru Zhang,
  • Fanfan Tang,
  • Chuanqing Fu,
  • Xiaodong Sui,
  • Chaodong Guan,
  • Li Xing

摘要

Working as the main load-carrying members, steel beams are widely used in many industries. Local damages can exist on the steel beams during the long service life, which may result in huge accidents. Electromechanical impedance (EMI) technology is effective at detecting damage, but its accuracy is affected by ambient temperature. Temperature fluctuations alter the performance of the steel and the sensor, causing a signal shift that can lead to false detection of damage. This study investigates temperature effects on EMI signals through theoretical and experimental analysis. A neural network was proposed with three input features extracted from the sensors and the EMI signals to estimate the crack length under ambient temperatures. Experimental results demonstrate that, within a temperature range of −10 °C to 60 °C, the average relative error for damage severity estimation is 2.89 %. The neural network outperforms the Hilbert transform-based method, offering potential for wide application in EMI-based structural health monitoring.