The accelerating digitalisation of industrial operations has transformed manufacturing into a highly interconnected ecosystem driven by Industrial Internet of Things (IIoT), cyber-physical systems, cloud computing, edge intelligence, and artificial intelligence. While these technological advancements have improved automation, operational efficiency, and real-time decision-making, they have also introduced increasingly complex cybersecurity challenges arising from expanded attack surfaces, interconnected assets, and continuously evolving threat landscapes. Among these threats, Advanced Persistent Threats (APTs) pose a particularly significant risk due to their stealth, persistence, and ability to infiltrate critical manufacturing infrastructure through multi-stage attack campaigns that often evade conventional security mechanisms. Existing cybersecurity solutions predominantly adopt reactive detection and response strategies, limiting their effectiveness in anticipating sophisticated attacks before operational disruption occurs. This study proposes a Digital Twin-enabled cybersecurity architecture for predicting, simulating, and mitigating Advanced Persistent Threats across smart manufacturing systems. The proposed framework integrates continuously synchronised digital twins with real-time operational data, AI-driven anomaly detection, attack graph modelling, behavioural analytics, and cyber-attack simulation to identify vulnerabilities, forecast attack progression, evaluate mitigation strategies, and support adaptive security decision-making without interrupting physical operations. Furthermore, the study presents a comprehensive architectural and performance evaluation framework that demonstrates how Digital Twin technology can strengthen cyber resilience, enhance threat preparedness, optimise incident response, and safeguard the reliability, continuity, and security of next-generation smart manufacturing environments.