Professional Documents
Culture Documents
5.1 FMEC Diffusers Updated
5.1 FMEC Diffusers Updated
Alessandro Corsini,
Giovanni Delibra & Lorenzo Tieghi
Diffusers are used to slow down the flow to convert its kinetic energy into pressure energy
• if the flow is subsonic the diffuser cross-section needs to increase to decrease the flow speed and
increase pressure
• if the flow is supersonic the diffuser cross-section needs to decrease to decrease the flow speed and
increase pressure
Flow phenomena in Diffusers
As such, the flow departs from the ideal isentropic model Rectangular section diffuser
so much that it is not reliable to use a simplified inviscid fluid
behavior
L
𝛼′
′′
𝛼 >𝛼 ′
Flow phenomena in Diffusers
(°deg)
20 hi
𝜀 =0 °
𝛼
′′ 15 10 °
9
L
40 °
′ 90 °
𝛼 𝜀=
0 °
90 °
2
hout
L/hi or L/ri
𝜀
Flow phenomena in Diffusers
Rectangular diffusers:
Circular diffusers:
Diffusers are used downstream of centrifugal compressor rotors or downstream of centripetal turbines
to slow down the flow and recover pressure according to Bernoulli equation:
To this aim, diffuser have an increasing cross-section (subsonic regimen) or decreasing cross-section
(supersonic regimen) that allows to slow down the fluid and thus increase its pressure.
The change in the cross-section results in the deceleration of the meridional component of velocity that
is responsible to keep the flow moving inside the device.
Please notice that you can slow down the flow, but to guarantee the design flow rate you need to have a
non-null meridional velocity component compatible with the passage area
Bladed Diffusers (ii)
However there is another velocity component that can be decreased: the peripheral (swirl) velocity
component.
In most turbomachinery the swirl component of velocity downstream of a rotor is considered as a source
of losses (the fastest the flow, the higher the losses that increase as a function of ). The flow velocity can be
written as:
Unlike the meridional component, the swirl component normally does not play an active role(*) and thus it can
be completely nullified to recover pressure. However, to do so, a bladed diffuser is needed to deviate the flow
path and therefore the design and manufacturing costs of the device increase
(*) in some cases a small peripheral component can be exploited to control separation, or other boundary layer
development characteristics so it can’t be completely killed
Bladed Diffusers (ii)
In part they recover pressure by slowing down the flow as their increasing cross-section results in the decreasing
of meridional velocity component due to continuity equation, in part they kill the peripheral velocity component
with the shape of their blades.
Impeller (rotor)
Increasing (diffusing)
cross section of the
diffuser Sandrolini, 1996
Bladed Diffusers (ii)
In part they recover pressure by slowing down the flow as their increasing cross-section results in the
decreasing of meridional velocity component due to continuity equation, in part they kill the peripheral
velocity component with the shape of their blades.
Bladed diffuser
Blades
Impeller (rotor)
Not shown
Caputo, 1994
Supersonic discharge
Caputo, 1994
Supersonic discharge
Caputo, 1994
Supersonic discharge
Caputo, 1994
Real diffusers
assumptions:
Real diffusers
Losses in Diffusers
Δ h 𝐷= h3 𝑟 − h3 𝑑𝑙
𝐿
or 𝑐2 𝑃
𝐶𝑑∫ ⋅ 𝑑𝑙
~ 𝑝3 − 𝑝3 𝑟 0 2 𝐴
Δ h𝐷 ¿
𝜌3 𝐴 𝑖𝑛
𝐴 𝑜𝑢𝑡
Assuming :
In a linear pipe:
where is the friction factor and the term in red is the shape coefficient
with:
• is determined by the boundary layer blockage i.e. the severity of adverse pressure gradient
• can be determined from Moody’s diagram as a function of Reynolds number
Losses in Diffusers: friction factor
: friction factor
from: wikipedia.org
Losses in Diffusers
𝐿 ~
𝑠𝑚𝑎𝑙𝑙 → 𝜁 ¿ Φ𝐷
h𝑖𝑛
𝐿 𝐴 𝑜𝑢𝑡
=8 → <1.5 𝛼 < 3 °
h𝑖𝑛 𝐴𝑖𝑛
𝐿 𝐴 𝑜𝑢𝑡
= 12→ < 2.4 𝛼< 3.5 °
h𝑖𝑛 𝐴 𝑖𝑛
−3
𝐶 𝑑 10
Diffuser efficiency
A simple definition of diffuser efficiency is not possible, as in different scenarios it is required to re-define
efficiency according to the effect the designer wants and the corresponding definition of “losses”.
Diffuser efficiency
A simple definition of diffuser efficiency is not possible, as in different scenarios it is required to re-define
efficiency according to the effect the designer wants and the corresponding definition of “losses”.
If we assume as the velocity at the diffuser inflow, we can define the diffuser efficiency as:
where and are velocity and pressure at the diffuser discharge, the exponent of the polytropic corresponding
to the real transformation of the flow.
Diffuser efficiency
The previous expression does not account for reheat work. To include it we need to modify the expression as:
Moreover, it is possible that not all the kinetic energy at the inflow is available (i.e. the flow can’t be stopped).
Alessandro Corsini,
Giovanni Delibra, Lorenzo Tieghi, Francesco Aldo Tucci
Turbulent flows in rotating channels are of obvious engineering interest with applications associated with
rotating devices:
• Turbines
• Pumps
• Compressors
The need for this study lies in the fact that the performance of the turbomachines can be largely affected by
the rotation of the system.
Effects of Rotation
In the analysis of fluid machinery behavior, it is often advantageous to view the fluid from a coordinate
system fixed to the rotating parts. This allows to work with fluid motions which are steady, but the rotating
system is not inertial.
In an inertial coordinate system, the acceleration on a particle of mass m is directly related to the vector sum
of the forces: F=ma.
• Coriolis force
• Centrifugal force
These forces can enhance or decrease the turbulent activity within the boundary layer.
Rotation effects on boundary layer mixing and stability
The mechanism that leads to the mixing process in turbulent boundary and shear layers can be described by
examining the forces on a particle that is displaced from its initial position in a rotating 2-dimensional parallel
shear flow [1].
If the relative vorticity, caused by the action of Coriolis force 2 on the particles, and the background rotation have
opposite senses the effects of rotation are destabilizing. If the relative vorticity and the background rotation have
the same sense the effects of rotation are stabilizing.
Coriolis forces on a particle in a rotating flow: (a) destabilizing effects; (b) stabilizing
effects.
[1] Tritton, D. J. "Stabilization and destabilization of turbulent shear flow
in a rotating fluid." Journal of Fluid Mechanics 241 (1992): 503-523.
Non-dimensional parameter: Rossby Number
When the effects of rotation become dominant, fluid motions exhibit properties quite different from those with no
rotation.
To define this regime it is necessary to develop a measure of the importance of rotation in each situation.
The quantity that gives a measure of the ratio between inertia force and the Coriolis force, is known ad Rossby
Number:
CFD is a tool for modeling turbulent flows within predefined volumes. The smaller the approximations the more
accurate the results (model of real flows).
Turbulence consists of hierarchy of eddies described by the Navier-Stokes Equations which are non-linear partial
differential equations that do not admit a general analytical solution, but only particular solutions obtained under
suitable simplifying hypotheses.
Turbomachinery flow modeling lives in a constant compromise between accuracy and the computational
costs of numerical simulations.
Computational grids and sketches of instantaneous velocity profiles for DNS and LES,
and RANS. (Hanjalic, 2005).
Background and Motivation
The aim of our work is to derive a wall function for rotating diffusing passages, through means of
machine learning.
Wall functions are universally exploited to replicate the physics of boundary layer where grid resolution does
not suffice.
Wall functions are derived by the observation of some canonical flows and are frequently applied in flows
where the ground assumptions cease to be true, such as rotating passages. In these flows, the mathematical
formulations of wall functions do not account for the distortion on the boundary layer due to the combined
action of centrifugal and Coriolis forces.
Background and Motivation
In this work a multi-layer neural network (MLNN) is trained to provide a correction for the Turbulent Kinetic
Energy (T.K.E.) at the wall, effectively substituting the standard wall treatment.
A LES simulation of a rotating diffuser at a Rossby number equal to 0.41 is performed to build a high-fidelity
database.
The same simulation is carried out using a RANS approach to derive the T.K.E. values to be corrected by
MLNN.
Test Case
The training datasets are generated through numerical simulations of a rotating radial diffuser.
Dimensions of the diffusing duct are similar to the test section of the experimental study performed by Moore [1].
With respect to those results, we have performed the simulation in kinematic similarity.
[1] Moore, J. "A Wake and an Eddy in a Rotating, Radial-Flow Passage—Part 1: Experimental Observations." (1973): 205-212.
Numerical Methodology
LES simulation relies on dynamic one equation model from Kim and Menon [2] using PISO approach.
RANS simulation relies on the Launder-Sharma model [3] using the SIMPLE approach.
Boundary conditions remain the same between RANS and LES simulations, with the exception of obviously not
present in the latter.
able 3 – Boundary Conditions
Patch U p k
Inlet mapped ZG mapped mapped mapped
Outlet ZG TP ZG ZG ZG
Wall NoSlip ZG kLowRe nutLowRe WF
[2] S. Menon, Kim, “High Reynolds number flow simulations using the localized dynamic subgrid-scale model”, 34th Aerospace Sciences Meeting, AIAA Paper
96-0425, Reno.
[3] Launder, B. E. and Sharma, B. I. (1974), "Application of the Energy-Dissipation Model of Turbulence to the Calculation of Flow Near a Spinning Disc", Letters
in Heat and Mass Transfer, Vol. 1, No. 2, pp. 131-138.
Flow Description
One of the metrics which illustrates the most important peculiarity of the flow in a diffuser is the static pressure
recovery ().
In a rotating frame of reference, it can be computed by subtracting the centrifugal pressure field to the static
pressure:
A comparison using mean relative axial velocities for the different numerical approaches versus the experimental
values, shows a good correspondence between the LES approach and the experimental data from Moore.
LES approach (dashed red line) captures the effect of the rotational forces on the boundary layer, as evident in X =
0.4699. In contrast, the RANS approach (blue dotted line) is completely unaffected by the rotation.