2025 AIChE Annual Meeting

Using CFD to Model Fluid Flow in Steam Ejector

Steam ejectors are a proven and reliable method for generating
vacuum within an industrial system. Featuring no moving parts and steam
as the motivating fluid, ejectors provide a cost-effective approach to
establishing precise low-pressure environments by harnessing the effects of
Bernoulli’s principle to entrain fluids. Although steam ejector technology
has a long history, modern advances in fluid flow modeling provide
manufacturers with the tools to analyze system dynamics and optimize
ejector design for improved performance.

The typical ejector system consists of two different fluids: steam and
air. The motivating steam enters the ejector through an inlet and
accelerates through a converging-diverging (CD) nozzle. This accelerates
the steam to a velocity of approximately 3300 feet per second, exiting the
nozzle at a Mach speed of approximately 2.2. As the supersonic steam exits
the nozzle it entrains the fluid within the suction chamber, producing a
negative gauge pressure and thereby creating a vacuum. This study
specifically examines how the geometry of a diffuser, when extended into
the suction chamber and disrupting the flow of the entrained air, influences
the resulting suction pressure.

To achieve our objectives, computational fluid dynamics (CFD)
simulations were performed, modeling the steady state, multi-species flow
of steam and air throughout the ejector. The K-Epsilon turbulence model
was applied to solve the Continuity and Navier Stokes equations within a
discretized mesh enabling detailed analysis of the internal fluid domain.
Simulation outputs for velocity, temperature and pressure are compared
against experimental data to validate the accuracy of the simulations.

This study demonstrates CFD’s capability to model the fluid domain
inside a steam ejector while capturing complex flow dynamics providing
valuable design insights to be used to optimize future ejectors. Creating a
quality mesh with fine detail throughout the geometry without scaling the
number of mesh elements proved to be instrumental in both simulation
accuracy and will be crucial to maintain as future work continues.