Digital Twin-Enabled Smart Infrastructure Management: A Multi-Level Data Fusion Approach for Decision Making in Predictive Maintenance

Date of Award

12-2024

Degree Name

Doctor of Philosophy

Department

Civil and Construction Engineering

First Advisor

Osama Abudayyeh, Ph.D.

Second Advisor

Hexu Liu, Ph.D.

Third Advisor

Azim Houshyar, Ph.D.

Keywords

Civil infrastructure, condition rating, deep learning, digital twin, multi-level data fusion, predictive maintenance

Abstract

Civil infrastructure are aging at an alarming rate and pose significant challenges, such as inefficiencies and financial burdens, compromising serviceability and the quality of life. Current management practices primarily rely on computerized maintenance management systems, which primarily depend on manual inputs that are costly, prone to human error, and lack comprehensive data integration across all asset lifecycle phases. To address these challenges, there is a pressing need for modernization and digital transformation in infrastructure asset management. Digital twin (DT) technology emerges as a revolutionary solution, enabling real-time data processing and informed decision-making. However, data-driven smart civil infrastructure management using DT is not yet achieved, especially in terms of data fusion. This research aims to integrate data fusion into DT domain for data-driven smart infrastructure management through three comprehensive studies.

First, a bibliometric analysis of 248 state-of-the-art research articles on DT for facility management (FM) was conducted. The analysis provided a holistic view of the research landscape. The findings indicated that the current research of DT in FM focuses on building information modeling-based FM and that the future research should focus on applying DT for intelligent prognostic and facility maintenance by using machine learning algorithms to integrate and fuse complex historical and heterogeneous data for accurate condition assessments.

Second, a systematic review of 105 papers was performed to identify the main challenges in data fusion for smart civil infrastructure management. The review highlighted issues such as data heterogeneity, interoperability, and data quality. In response, a multilayer data fusion framework was proposed, integrating open building information modeling and geographic information system for immersive visualization and stakeholder engagement. The proposed framework facilitates lifecycle data integration, increased data accuracy, adaptability, and further effectiveness of engineering decision-making.

Third, an integrated approach is introduced that leverages deep learning and data fusion for advanced assessment of concrete bridge deck conditions. A novel DSS was developed to predict the condition rating of concrete bridge decks through a multi-level fusion of data from the National Bridge Inventory, environmental reports, imagery, and sensors. The Long Short-Term Memory model was employed for predictive analysis. This system demonstrated a significant improvement in prediction accuracy with an R-squared value of 0.99 and a mean absolute error of 0.04, compared to similar model it increases the prediction performance by 8%. An image-based methodology was developed that utilizes YOLOv8 and computer vision techniques to identify and evaluate concrete defects, such as cracks, scaling, spalling, and leaching. The severity of each defect type is classified and aggregated to produce a comprehensive condition rating, guiding maintenance priorities. This automated approach can significantly reduce inspection time and costs while maintaining high objectivity and consistency in evaluations, achieving a prediction accuracy of 95%.

In conclusion, this research offers a robust and innovative solution for evaluating the condition of concrete bridges, providing a faster, more reliable, and cost-effective alternative to traditional condition rating methods.

Access Setting

Dissertation-Abstract Only

Restricted to Campus until

12-1-2026

This document is currently not available here.

Share

COinS