Loading...

Data Loss Recovery in Wireless Sensors and Using them in Vibration-based Structural Damage Detection Employing Convolutional Neural Networks

Baktash, Shayan | 2022

179 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54904 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Rahimzadeh Rofooei, Fayaz
  7. Abstract:
  8. This study aims to identify the structural damage using deep learning. Damage to structures reduces their lifespan. Therefore, continuous monitoring of structures is necessary to detect any possible damages that occurred due to various causes. Structural health monitoring systems that use wired sensors to measure signals are costly and time-consuming to install. Whereas, wireless sensor-based damage detection systems are relatively cost-effective. However, one of the most important problems in using them is the loss or non-recording of some data, which severely affects the accuracy of a structural damage detection system. On the other hand, health monitoring using machine learning-based damage detection systems that use hand-crafted features cannot guarantee the optimal performance of damage detection systems in structures. To overcome the above problems, in this study, a structural damage detection system using one-dimensional convolutional neural networks and one-dimensional autoencoder neural networks is introduced which can initially recover the lost data of the wireless sensors at different loss ratios and then use them to diagnose structural damage. The results show that the final designed system can recover the data not recorded by the employed sensors with relatively small error, even for a loss ratio of 90%, and is also able to detect the structural damage with relatively high accuracy. The results also indicate that recovering lost data in cases where some data is not recorded by the sensors, has a significant impact on increasing the accuracy of the damage detection.
  9. Keywords:
  10. Structural Health Monitoring ; Deep Learning ; Convolutional Neural Network ; Wireless Sensor Network ; Lost Data Recovery ; Vibration-Based Structural Damage Detection ; Autoencoder Neural Networks

 Digital Object List

 Bookmark

No TOC