The industrial metaverse represents a significant advancement towards Industry 5.0, enabling improvements in efficiency, training, and remote collaboration within industrial contexts. Despite its transformative potential, widespread adoption faces substantial challenges, particularly cybersickness, a condition characterized by nausea, dizziness, and disorientation arising from prolonged exposure to virtual environments. Current cybersickness mitigation strategies often lack personalization, failing to accommodate individual differences such as motion sensitivity and prior experiences. In order to contribute to address this gap, this paper introduces a novel Personal Digital Twin (PDT)-driven approach that aims to mitigate personalized cybersickness by integrating real-time biometric data monitoring, individualized susceptibility assessment, and dynamic XR environment adaptation. The authors also validated the proposed approach by leveraging two concrete industrial case studies: the offshore wind turbine maintenance and the overhead crane operations training. A significant result of the research work is that a PDT-driven adaptive training can contribute to enhance user comfort, reduce cybersickness, and promote more effective and inclusive adoption of industrial metaverse applications.