Development of a Digital Twin of a Laboratory Structure for Machine Vision-Based Structural Health Monitoring Approach

Narges Dehghani Dehnavi, Narges; Bitaraf, Maryam

doi:10.22059/ceij.2025.385472.2183

Development of a Digital Twin of a Laboratory Structure for Machine Vision-Based Structural Health Monitoring Approach

Articles in Press

Authors

Narges Narges Dehghani Dehnavi ¹

Maryam Bitaraf ²

¹ Enghelab Ave., 16 Azar Street, University Of Tehran

² School of Civil Engineering, University of Tehran

10.22059/ceij.2025.385472.2183

Abstract

This study presents a cost-effective Structural Health Monitoring (SHM) approach that integrates machine vision, digital twin technology, and machine learning. Machine vision serves as a sensor to capture the response of a three-story laboratory structure under base excitation, using the optical flow method and the Lucas-Kanade algorithm to track displacements. These measurements are validated against radar and accelerometer sensors, demonstrating the effectiveness of radar sensors for vibration-based displacement monitoring in SHM. A digital twin is then developed by integrating vibration data with a physics-based model to simulate structural behavior, enabling the detection of damage type, location, and severity under various conditions. Different machine learning classifiers are trained on data from both simulated and physical models, with the Manhattan distance-based classifier achieving the highest accuracy of 92%. The results indicate that this digital twin system offers a reliable tool for real-time SHM and predictive maintenance, facilitating early damage detection and enhancing structural resilience.

Keywords

Machine learning

Optical flow algorithm

Lucas-Kanade algorithm

Physics-based model

Manhattan distance

Subjects