The introduction of green technologies in aviation brings significant safety concerns, stemming from uncertainties about their reliability due to the lack of extensive operational history, and the need for novel system architectures which increase the risk of faults due to unexpected interactions between subsystems. Managing these risks while maintaining safety is a significant challenge for the next generation aircraft that emphasizes the need for reliable onboard real-time health monitoring techniques to detect potential unknown fault modes. The accuracy of state-of-the-art health monitoring methodologies is often constrained to the adoption of expensive high-fidelity physics-based models, making the overall time for faults assessment in the order of minutes or hours. This delayed diagnostics coupled with the complexity of faults identification constitutes a significant barrier for the adoption of green technologies in aviation. This presentation discusses and proposes purpose-driven digital twins to enable the fast identification of the health status of complex technologies to ease and accelerate the transition towards sustainable aviation. We illustrate two original formulations of data-driven physics-based digital twins capable of reducing the time scale of diagnostics from minutes/hours to seconds, to meet strict time to decision constraints and reliability of the prediction. At the heart of these predictive digital twins is the synergistic combination of data assimilation from physics-based representations of the system considering both single-source and multisource artificial learnings, uncertainty quantification to characterize the reliability of the health status prediction, and projection-based compression to ensure efficient predictive capabilities. These formulations are demonstrated for key technologies that relate to the transition to all-electric green aircraft.