CO2 critical discharge through compact restrictions: a corrected-CNT two-fluid model

Critical discharge of dense/supercritical CO 2 through compact restrictions couples compressible acceleration, wall friction, metastable evolution, and interfacial heat and mass transfer. Depending on the inlet state and decompression path, the flow remains single phase or follows delayed flashing or condensation, making phase-transition onset central to both accuracy and convergence. This study develops a cross-section-averaged quasi-steady two-fluid model for constant-area channels and converging restrictions and assesses it against 449 experiments, comprising 357 constant-area cases and 92 converging-nozzle cases. Onset-treatment errors are compared on the 335 cases that converge for all three models, while convergence is reported across all 449 cases. On the common set, the classical CNT treatment gives MAPE = 11.61%, MAE = 3910.8 kg·m −2 ·s −1 , and R 2 = 0.86, whereas corrected-CNT gives MAPE = 9.95%, MAE = 3458.1 kg·m −2 ·s −1 , and R 2 = 0.89. Across all 449 cases, the convergence rate increases from 77.1% to 95.3%. The corrected-CNT boundary is semi-empirical: it retains the CNT dependence on surface tension, molecular volume, and temperature, and its branch-dependent coefficients are calibrated from compiled CO 2 phase-limit data. Leave-one-out tests give onset-pressure MAPE values of 1.30% for condensation and 3.86% for flashing. A reduced-order correlation derived from 4000 model-generated samples provides rapid estimates over the investigated ranges; against the experimental database, its MAPE is 10.63%. Relative to classical CNT, the framework improves common-case accuracy and all-case convergence while retaining route-resolved prediction and an explicit validity range for the correlation.