Scholarlink

In past years, sensor-based health diagnostics have shown significant potential for electromechanical impedance (EMI) techniques with many practical implementations. In order to enhance the effectiveness of health monitoring systems, it is crucial to examine the bonding interface between the piezoelectric sensor (PZT) and the host structure. A promising approach to achieve this goal is the utilization of Sn–Ag bonding, which offers improved resistance to moisture, chemicals, and temperature compared to traditional epoxy adhesive bonding. The Sn–Ag bond layer presents a novel attachment method to the structural design, which may ultimately enhance the sensitivity of the health monitoring system. This study used EMI techniques to instrument a unique Sn–Ag alloy-based bond and compared it against an epoxy adhesive bond for better sensing and actuation. Later, both the bonds were cured, and the bond performances were analyzed through conductance and susceptance signatures. Moreover, three different bond layer damage conditions, i.e., incipient, moderate, and severe, were chosen to check the performance of the bond layer through admittance signatures. The EMI techniques rely on pristine data, which presents significant constraints when applied to pre-existing structures. Furthermore, the experimental results were used to predict baseline data through deep learning classifications, i.e., convolutional neural network (CNN), long short-term memory (LSTM), and CNN-LSTM hybrid model for all the bond layer damage conditions. The experimental results were compared against the chosen deep learning model for futuristic prediction. The efficiency of all the deep learning models is quantified using performance metrics like accuracy and R2 score. The present study demonstrates that the hybrid model yields superior performance compared to the other two models for predicting Sn–Ag and adhesive bond layer damage. The proposed framework aims to replace traditional optimization algorithms to achieve a more simplified effective prototype for the futuristic prediction of EMI signals.

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

Detection and Prediction of Bond Degradation for Piezo Impedance-Based Structural Health Monitoring (PISHM) Using Hybrid Deep Learning Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lukesh Parida,
  • Sumedha Moharana,
  • Sourav Kumar Giri

摘要

In past years, sensor-based health diagnostics have shown significant potential for electromechanical impedance (EMI) techniques with many practical implementations. In order to enhance the effectiveness of health monitoring systems, it is crucial to examine the bonding interface between the piezoelectric sensor (PZT) and the host structure. A promising approach to achieve this goal is the utilization of Sn–Ag bonding, which offers improved resistance to moisture, chemicals, and temperature compared to traditional epoxy adhesive bonding. The Sn–Ag bond layer presents a novel attachment method to the structural design, which may ultimately enhance the sensitivity of the health monitoring system. This study used EMI techniques to instrument a unique Sn–Ag alloy-based bond and compared it against an epoxy adhesive bond for better sensing and actuation. Later, both the bonds were cured, and the bond performances were analyzed through conductance and susceptance signatures. Moreover, three different bond layer damage conditions, i.e., incipient, moderate, and severe, were chosen to check the performance of the bond layer through admittance signatures. The EMI techniques rely on pristine data, which presents significant constraints when applied to pre-existing structures. Furthermore, the experimental results were used to predict baseline data through deep learning classifications, i.e., convolutional neural network (CNN), long short-term memory (LSTM), and CNN-LSTM hybrid model for all the bond layer damage conditions. The experimental results were compared against the chosen deep learning model for futuristic prediction. The efficiency of all the deep learning models is quantified using performance metrics like accuracy and R2 score. The present study demonstrates that the hybrid model yields superior performance compared to the other two models for predicting Sn–Ag and adhesive bond layer damage. The proposed framework aims to replace traditional optimization algorithms to achieve a more simplified effective prototype for the futuristic prediction of EMI signals.