Digital Twin Based Framework for Real Time Software System Reliability Evaluation

Abstract

Modern software systems require continuous reliability evaluation because service failures can spread across connected components and interrupt operations. This study presents a digital twin framework for real-time software system reliability evaluation. The framework combines live monitoring data, synchronized virtual states, reliability scoring, and short-term behaviour analysis. Response time, error rate, resource utilization, service availability, failure events, and dependency health were used to represent system condition. The digital twin was evaluated during normal operation, increasing workload, service failure, and recovery, and was compared with a conventional threshold-based method. Results showed that the twin followed the physical system closely, detected reliability reduction earlier, responded faster to failure, and tracked recovery more accurately. The framework also provided a smoother reliability pattern by combining several operational signals instead of waiting for fixed warning limits. The proposed approach supports continuous monitoring, early warning, failure analysis, recovery tracking, and better operational decisions in complex software environments.