Abstract
This work investigates the nucleation of droplets from supercooled CO2 and CO2-rich gas during decompression, using classical nucleation theory (CNT). We model the supercooling limit employing highly accurate equations of state and compare the result with nucleation pressures determined from experimental data. The present analysis is relevant for the safety assessment of pipelines containing gaseous CO2.
Three new full-bore decompression experiments with pure CO2
were conducted, incorporating high-speed pressure sampling and multiple sensors to achieve precise characterization of the decompression wave speed as a function of pressure. Additionally, eight experiments with pure CO2 and nine with
CO2-rich mixtures from the open literature were analysed.
The homogeneous equilibrium model (HEM) and delayed homogeneous equilibrium model (
) were used to calculate the decompression wave speed down to the choking condition. Across all experiments, predictions based on the saturation pressure (applied in the HEM) consistently overestimated the experimentally determined nucleation values. In contrast, those based on the supercooling limit (applied in
) showed a mean absolute percentage deviation of 3%
, with predictions randomly distributed around the experimental results.
Three new full-bore decompression experiments with pure CO2
were conducted, incorporating high-speed pressure sampling and multiple sensors to achieve precise characterization of the decompression wave speed as a function of pressure. Additionally, eight experiments with pure CO2 and nine with
CO2-rich mixtures from the open literature were analysed.
The homogeneous equilibrium model (HEM) and delayed homogeneous equilibrium model (
) were used to calculate the decompression wave speed down to the choking condition. Across all experiments, predictions based on the saturation pressure (applied in the HEM) consistently overestimated the experimentally determined nucleation values. In contrast, those based on the supercooling limit (applied in
) showed a mean absolute percentage deviation of 3%
, with predictions randomly distributed around the experimental results.