DIGITAL TWINS FOR RESILIENT STRUCTURES: A SYSTEMATIC REVIEW AND META-ANALYSIS OF REAL-TIME DYNAMIC RESPONSE ANALYSIS, PREDICTIVE MAINTENANCE, SENSOR-DRIVEN DATA ANALYTICS, AND INTELLIGENT DECISION-MAKING IN SMART INFRASTRUCTURE MONITORING

Abstract

The integration of digital twins into structural dynamics and smart infrastructure monitoring has attracted considerable attention for its potential to enhance structural resilience through real-time dynamic response analysis, predictive maintenance, sensor-based data analytics, and intelligent decision-making. This systematic review and meta-analysis aimed to synthesize the existing evidence on the effectiveness of digital twin frameworks in these domains, with a particular focus on predictive maintenance accuracy as the primary outcome of interest. A comprehensive literature search was conducted across multiple electronic databases, following the PRISMA guidelines, to identify eligible studies that reported quantitative performance metrics for digital twin-based systems in structural health monitoring. The included studies were subjected to rigorous quality assessment using established risk-of-bias tools, and a random-effects meta-analysis was performed to pool effect sizes where appropriate. From the collected data, we extracted summary statistics for predictive maintenance accuracy, including a reported effect size of  with a 95% confidence interval ranging from  to , indicating a moderate but statistically non-significant negative association in one key study. The overall pooled estimate across studies was  with a corresponding p-value of , suggesting a trend toward improved predictive maintenance performance that did not reach conventional significance thresholds. The heterogeneity among studies was considerable, as reflected by an  statistic of 4.54, which underscores the variability in methodologies, sensor configurations, and structural systems examined. Our findings indicate that while digital twin integration shows promise for enhancing predictive maintenance and real-time monitoring, the current evidence base remains limited by small sample sizes and inconsistent outcome definitions. We conclude that future research should standardize performance metrics and adopt larger-scale field validations to confirm these preliminary trends and to advance the deployment of intelligent decision-making systems for resilient infrastructure.