This work presents a rapid, calibration-free methodology for constructing high-fidelity digital twins of laboratory X-ray computed-tomography (CT) systems using only radiographs acquired with the instrument itself. By measuring the nonlinear relationship between operator-set cathode current and actual anode current, we correct the emitted photon flux and derive a representative flux model through inverse optimisation of image noise. The approach incorporates filtration, scintillator response, detector point-spread function, electronic gain, quantisation, and overexposure limits, enabling realistic photonic-noise synthesis without specialised dosimetry equipment. Validation on a dual-tube high-energy (DTHE) micro-focus CT platform demonstrates sub-5% error in signal-to-noise ratio across a broad range of voltages (60–260 kV), currents (50–500 μA), exposures, and gains, and accurate reproduction of material-specific attenuation for aluminium, copper, and steel specimens. Comparisons with Monte-Carlo (GATE) and analytical (gVirtualXray) simulations confirm that the inverse-derived twins achieve comparable visual and quantitative fidelity while requiring orders of magnitude less computation time (≈0.03 s per projection). Limitations stem from unknown internal filtration and gain-dependent noise scaling. Nonetheless, the proposed pipeline provides a practical, reproducible framework for generating digital twins that support rapid prototyping, experimental planning, and advanced data-driven analyses in X-ray imaging research.
