Sunteți pe pagina 1din 9

Thermophysics and Aeromechanics, 2017, Vol. 24, No.

DOI: 10.1134/S0869864317040060

The surface roughness effect on the performance


of supersonic ejectors
1 1* 2 2
D.V. Brezgin , K.E. Aronson , F. Mazzelli , and A. Milazzo
1
Ural Federal University named after the First President of Russia
B.N. Eltsin: Ural Power Engineering Institute, Ekaterinburg, Russia
2
University of Florence, Florence, Italy
*
E-mail: k.e.aronson@urfu.ru

(Received March 31, 2016; in revised form August 3, 2016)

The paper presents the numerical simulation results of the surface roughness influence on gas-dynamic processes
inside flow parts of a supersonic ejector. These simulations are performed using two commercial CFD solvers (Star-
CCM+ and Fluent). The results are compared to each other and verified by a full-scale experiment in terms of global
flow parameters (the entrainment ratio: the ratio between secondary to primary mass flow rate - ER hereafter) and local
flow parameters distribution (the static pressure distribution along the mixing chamber and diffuser walls). A detailed
comparative study of the employed methods and approaches in both CFD packages is carried out in order to estimate
the roughness effect on the logarithmic law velocity distribution inside the boundary layer. Influence of the surface
roughness is compared with the influence of the backpressure (static pressure at the ejector outlet). It has been found
out that increasing either the ejector backpressure or the surface roughness height, the shock position displaces up-
stream. Moreover, the numerical simulation results of an ejector with rough walls in the both CFD solvers are well
quantitatively agreed with each other in terms of the mean ER and well qualitatively agree in terms of the local flow
parameters distribution. It is found out that in the case of exceeding the “critical roughness height” for the given
boundary conditions and ejector’s geometry, the ejector switches to the “off-design” mode and its performance de-
creases considerably.

Key words: supersonic ejector, entrainment ratio, CFD, wall log-law, sand-grain roughness.

Generally, most of ejector design procedures assume that the roughness of internal sur-
faces does not affect its performance [1, 2]. In a number of publications, it is recommended to
manufacture the ejector’s flow parts with the lowest roughness height as possible [3]. The pre-
sent paper presents the numerical and full-scale experimental results related to the roughness
influence on gas-dynamic processes of the supersonic ejector. These simulations are performed
using two commercial CFD solvers (Star-CCM+ and Fluent). The results are compared to each
other and verified by a full-scale experiment in terms of global flow parameters (Entrainment
Ratio (ER hereafter) is the ratio between secondary to primary mass flow rate) and local flow
parameters distribution (the static pressure distribution along the mixing chamber and diffuser
walls). A detailed comparative study of the employed methods and approaches in both CFD
packages is carried out in order to estimate the roughness effect on the logarithmic law velocity
distribution inside the boundary layer. Influence of the surface roughness is compared with
the influence of the backpressure (static pressure at the ejector outlet). It has been found out
that increasing either the ejector backpressure or the surface roughness height, the section
in which the pressure rise due to the shock appears is displaced upstream.

 D.V. Brezgin, K.E. Aronson, F. Mazzelli, and A. Milazzo, 2017

553
D.V. Brezgin, K.E. Aronson, F. Mazzelli, and A. Milazzo

Figure 1 shows the design of a typical ejector. The ejector consists of a supersonic prima-
ry nozzle, mixing chamber and diffuser profiled according to Bezier curves. The term “pro-
filed” means that the geometrical parts represent in the longitudinal section a curved surface
rather than a cylinder or a cone one. This surface is generally described by the second- or third-
order Bezier polynomials.
Based on the data of the literature overview [5−7], the k-ω SST model is selected as
the turbulence model for all simulations. The pentafluoropropane R245fa is used as the work-
ing fluid in the numerical model and experiment. Organic liquid R245fa is one of the most ex-
tensively used fluids for low-temperature Rankine organic cycle due to a low boiling tempera-
ture (tboil ~ 288 K at Р = 100 kPa) and a positive slope of the saturation line over the whole
range of the entropy variation on the T–S diagram. Such a special behavior of the saturation
line in the T–S diagram prevents the formation of the two-phase (liquid−gas) mixture while
the primary flow expands inside supersonic nozzle.
It is known that “perfect gas” model satisfactorily defines the fluid properties under low
pressure and high temperature conditions; however, near the saturation line and the critical
point, it causes significant deviations between the “ideal” and “real” gases behaviors [8].
The work [9] presents a comparative study of several equations of state as applied to the R245fa.
Based on the data of the U.S. National Institute of Standardization and Technology [10],
the authors in that study figured out that the “Peng−Robinson” equation of state ensures
the best agreement of R245fa with the NIST data over the whole range of conditions. The cu-
bic Peng−Robinson equation of state, as all the real equations of state do, takes into account
the intermolecular forces. As a refinement of the Van der Waals equation, the Peng−Robinson
equation of state enables one to describe the real gas properties depending on three individual
properties of the fluid medium: the critical pressure, critical temperature, and acentric factor,
which takes into account the asymmetry of molecules. Due to the fact that the Peng−Robinson
equation of state is built in both CFD solvers and its applicability is justified, it is used for fur-
ther research in the present work.
The near-wall mesh treatment (here we use a finite volume mesh) for the numerical simu-
lations is a crucial feature in terms of the simulation fidelity especially for the tasks of rough-
ness influence investigation on the overall ejector performance. The general simulation ap-
proach suggests that one can resolve the viscosity-affected near wall region by two different
techniques: direct wall integration or resolution with the aid of a “wall functions”. The essence
of the first approach is based on the fact that the boundary layer is resolved with a mesh all
the way to the wall and the employed turbulence model closes the system of differential equa-
tions over the entire control volume, including the near-wall boundary layer. In this case, a suf-
ficient accuracy in the estimation of near wall shear stresses is reached only when the dimen-
+
sionless wall distance for a wall-bounded flow (wall Y ) is less than unity, and the entire height
of the boundary layer includes at least 10 rows of cells. The second approach (using wall func-
+
tions) is characterized by a much coarser mesh (the coefficient Y is above 30) and assumes

Fig. 1. Standard ejector design scheme.

554
Thermophysics and Aeromechanics, 2017, Vol. 24, No. 4

that the near-wall cell lies in the region of the logarithmic law of velocities distribution inside
the boundary layer, and the employed turbulence model is valid only outside this region.
It is to be noted that the “integration to wall approach” is generally the best and justified
decision, which enables obtaining the best agreement with experiments, especially in the prob-
lems related to either the drag coefficient estimation or conjugate heat transfer. However,
the only way for roughness modeling tasks in both CFD-solvers is to use a modified in some
way “wall function approach” [11−13].
The relationship between the resistance law and the mean velocity distribution near walls
for the pipes with a non-zero roughness has been generalized in [14, 15]. It is shown that
the distribution of mean flow velocity near the wall in the boundary layer for rough pipes has
the same slope (1/k) as for smooth pipes when plotted, however, the curve shifts downwards
because of the additive factor (Fig. 2):
1 1
U trb = ⋅ ln(9.0 ⋅ Y + ) − ⋅ ln( f r ), (1)
k k

here U trb is the dimensionless wall parallel mean flow velocity in the boundary layer, k is
the Von Karman constant = 0.42, fr is the wall roughness function which depends on the type
and size of a typical roughness. It is to be noted that because of the great variety of shapes and
types of the real surface roughness, the existence of a universal roughness function is ambigu-
ous. To compare the drag of pipes with different roughness types Nikuradse [14] has proposed
to compare the data with the sand-grain roughness, the dimensionless height of which is deter-
mined as follows:

Ks+ =ρ ⋅ Ks ⋅ u ∗ µ , u∗ = τ / ρ , (2)

here Ks is the dimensionless height of the sand-grain roughness (linear size), u* is the friction
velocity, τ is the shear stress, ρ is the density, and µ is the dynamic viscosity. Based on the re-
sults of his own experiments along a surface with a non-zero roughness, Nikuradse [14] has
determined three distinct flow regimes depending on the dimensionless sand-grain roughness
height Ks+ . The regime without the roughness effect ( Ks+ < 5) may be assigned to the first
mode. The sizes of grains forming the roughness are so small at such a hydraulically smooth
mode that all the bumps lie inside the laminar (viscous) sublayer. The transitional mode
(5 < Ks+ < 70) belongs to the second regime. The roughness elements partially go beyond
the laminar sublayer. The extra drag arises because of roughness elements shape resistance,
which go beyond the laminar sublayer and enter the turbulent boundary layer. The third regime
is characterized by a full roughness impact ( Ks+ > 70). All the roughness elements go beyond
the laminar sublayer. The drag predominant part consists of the individual roughness elements
shape resistance, therefore, the drag law is quadratic.
The authors of the work [16] have corrected the boundaries of all three regimes by using
the experimental data of [14] and proposed the functions for determining the extra drag associ-
ated with the surface roughness. These de-
pendencies are used in Star-CCM+ and Fluent
to modify the wall logarithmic law by rela-
tionship (1) with regard to roughness effect.
The table presents the functions employed in
the numerical algorithm of the both CFD
solvers.

Fig. 2. Downward shift of the logarithmic velocity


profile due to the surface roughness effect [13, 14].

555
D.V. Brezgin, K.E. Aronson, F. Mazzelli, and A. Milazzo

Table
Built-in functions and regime conditions in Star-CCM+ and Fluent for roughness modeling

Regime Star-CCM+ Fluent


Hydrodynamically
smooth , fr = 1 fr = 1
+
K s ≤ 2.25
Transitional, a a
 Ks+ − 2.25 +  K s+ − 2.25 
+
2.25 < K s ≤ 90 fr =  B ⋅ 90− 2.25 + Cs ⋅ K s  , fr = 
+
+ Cs ⋅ K s  ,
  
 90 − 2.25 
π ln( Ks+ / 2.25)  +
a = sin (0.4258⋅(ln (K s ) − 0.8111)),
a = sin  ⋅ ,
2 ln(90 / 2.25) 
  Cs = 0.5
Cs = 0.253, B = 0.0
Fully rough, + +
f r = B + Cs ⋅ K s , fr = 1 + Cs ⋅ K s ,
+
Ks > 90 Cs = 0.253, B = 0.0 Cs = 0.5

As is seen in the Table, both CFD solvers use the empirical constant Cs to define the ad-
ditional surface drag in the transitional and full rough regime. However, the default value for
the constant is equal to 0.253 in the Star-CCM + and in Fluent, it is equal to Cs = 0.5. Figure 3
depicts the logarithm of the roughness function fr versus the quantity Ks+ in the transitional
regime with the set Cs = 0.5 and B = 1 for the both CFD solvers. The comparison in Fig. 3
shows that both the solvers qualitatively and quantitatively closely estimate the wall roughness
effect on the logarithmic law region inside the boundary layer (the relative error is about
−5 ÷ +15 %).
However, according to the documentation for both CFD software, each solver has their
own built-in tools for modeling problems with non-zero wall roughness. In the case of large
+
values of the roughness height and low values of Y , the logarithmic velocity distribution func-
tion within the boundary layer tends to singularity (see Fig. 2). To get rid of that uncertainty,
Star-CCM+ uses a built-in “artificial roughness height limit” mechanism for the roughness
+
height, the essence of which is that when Y tends to zero a solver artificially reduces
+
the roughness height. Thus, it is necessary to satisfy the condition Y > Ks+ in the Star-CCM+
for the full roughness impact on the logarithmic law of velocity distribution within the bounda-
ry layer, otherwise the solver will not be able to properly assess the roughness influence. Such
a kind of restriction requires the use of relatively coarse meshes and makes the solution de-
+
pending on the Y quantity. To avoid uncertainty of the flow velocity distribution within
+
the boundary layer at small Y values and large roughness height, the Fluent solver uses
a built-in “virtually shifting the wall” tool (see Fig. 2). Such kind of approach is based on
the observation that the viscous sublayer is fully established only in a hydraulically smooth
flow regime (without roughness impact). In the transitional roughness regime, the roughness
elements are slightly beyond the viscous sublayer, as a result, the laminar layer starts to be
“washed out” by the flow disturbances within
the turbulent sublayer. So that in fully rough
flows, the viscous (laminar) sublayer breaks
down completely, and the influence of vis-
cous forces becomes negligible.

Fig. 3. Logarithm of the roughness function fr vs.


the quantity K s+ in transitional mode.
1  computation in the Star-CCM+ environment,
2  computation in the Fluent environment.

556
Thermophysics and Aeromechanics, 2017, Vol. 24, No. 4

Fig. 4. Model of the equivalent sand-grain


roughness.

Figure 4 illustrates the equivalent sand-


grain roughness using a wall with a layer of
closely packed spheres, which have an average roughness height representing a technical
roughness with peaks and valleys of different shapes and sizes.
The “virtually shifting the wall” tool employed in Fluent solver for roughness modeling
+
lies in a correction of the Y quantity for the first cell according to the following relationship:
+
Y= Y + + K s+ 2. (3)
+
As a result, the Y quantity always remains larger by a halved height of the dimensionless
sand-grain roughness height. Thus, fine meshes can be handled correctly.
At the first stage of the present work, series of numerical simulations were carried out for
the ejector with zero roughness (smooth walls). To reduce computational time, simulations
were performed in 2D axisymmetric domain. For the Fluent simulations, a structured mesh
+
with the coefficient Y < 1.0 and a total number of cells of about 80,000 is used. For the Star-
CCM+ simulations, a polygonal mesh with 30 rows of prismatic elements near the walls with
+
the coefficient Y < 1.0 and a total number of elements of about 300,000 is used. Boundary
conditions and models were identical for both simulation series. A comparison with experi-
mental data [4] has shown a good agreement between the integral flow parameters, which was
expressed via the Entrainment Ratio (ER). The relative error for the ER value for all simula-
tions was no more than 5% the experimental data.
Figure 5 shows the static pressure distribution along the mixing chamber and the diffuser
walls for the various backpressure values from the numerical simulation results (Fig. 5a) and

Fig. 5. Static pressure distributions along the mixing chamber and diffuser walls
for four backpressure values.
a  numerical simulation results for smooth walls in the Star-CCM+ and Fluent solvers:
1− 4  Star-CCM+ solver, 5−8  Fluent solver: Pst = 1.61 (1, 5), 1.64 (2, 6), 1.67 (3, 7), 1.71 (4, 8);
b  experimental data: Pst = 1.61 (1), 1.64 (2), 1.67 (3), 1.71 (4).

557
D.V. Brezgin, K.E. Aronson, F. Mazzelli, and A. Milazzo

experiments (Fig. 5b). As can be seen from Fig. 5a, the static pressure distribution profile
along the walls for a smooth wall almost coincides for both the CFD solvers, and while in-
creasing the backpressure, the section in which a pressure shock takes place moves upstream
(from the coordinate x = 0.385 m to the coordinate x = 0.335 m). Throughout the full-scale
experiments, the static pressure transducer probes were mounted with the same spacing along
the mixing chamber and diffuser walls. Therefore, it was not possible to precisely identify
the plane section in which a pressure shock has taken place. It should be noted that according
to the numerical simulation results for a smooth wall (Fig. 5a), the static pressure distribution
ahead of the pressure shock almost coincides irrespectively of the backpressure magnitude.
Whereas the experimental data of [4] show (Fig. 5b) that as the backpressure increases,
the static pressure quantity at the wall in each section from the start of mixing chamber also
increases against the profile with a lower backpressure value.
At the second stage of numerical studies, series of numerical simulations were carried out
for the ejectors walls with non-zero roughness height. Taking into account the above-described
features used in Fluent and Star-CCM+ solvers for roughness modeling, the mesh for the Star-
+
CCM+ was rebuilt (Y = 30−200, the total number of cells was about 200 thousand); the Fluent
mesh has remained unchanged. To align roughness blending functions in both CFD-solvers,
the default values in the Star-CCM+ for the Cs and B constants were changed to Сs = 0.5 and
B = 1.0 (see the Table).
Figure 6 shows the static pressure distributions (Fig. 6a) and characteristic Mach number
of the secondary flow (Fig. 6b) profiles along the mixing chamber and diffuser walls according
to the numerical simulation (Star-CCM+ and Fluent) and experimental results for the backpres-
sure magnitude equal to 1.67 bars. It is found out that the numerical static pressure profiles
along the walls for both CFD-Solvers are quite different for the same roughness heights (Fig.
6a). It is suggested that this is because of the different approaches and meshes in Star-CCM+
and Fluent for roughness modeling. It is quite complicated to clearly conclude which CFD

Fig. 6. Static pressure and characteristic Mach number distribution profiles along the ejector walls
at the backpressure equal to Pcond = 1.67 bar.
a — static pressure: experiment (1), Fluent solver at Ks = 20 µm (2),
Star-CCM+ solver at Ks = 20 (3), 32 (4), and 34 (5) µm, simulation results for smooth walls (6);
b — characteristic Mach number: Star-CCM+ solver at Ks = 20 (1), 32 (2), and 34 (3) µm,
simulation results for smooth walls (4).

558
Thermophysics and Aeromechanics, 2017, Vol. 24, No. 4

solver provides more consistent results with the experimental data, since the equivalent sand-
grain roughness height itself is typically several times different from the root mean square or
average roughness measured by the profilometer. The average roughness height obtained by
the contact profilometer measurements on the test ejector surfaces was K ta= 4 − 6 μm [4]. Ac-
cording to the various papers [17, 18], the conversion factor that should be used to transform into
the equivalent send-grain roughness can vary within Ks = (3 − 6) ⋅ K t . Thus, further investiga-
tions are required to determine the conversion factor of the measured average roughness quan-
tity to the equivalent send-grain roughness magnitude.
Numerical results have shown that the influence of the ejector wall roughness on the sta-
tic pressure distribution along the wall has a non-monotonous feature. At a low roughness,
the pressure distribution corresponds to the “on-design” mode of the ejector operation. Since
the equivalent sand-grain roughness height exceeds the “critical roughness height” then ejector
switches to the “off-design” mode; this leads to abrupt alteration of the pressure profile (distri-
bution) along the ejector wall.
It can be seen from Fig. 6a that for to the simulation results obtained by Fluent solver,
the pressure shock occurs at x = 0.45 m in the case of smooth walls and at x = 0.35 m in
the case of roughness height Ks = 20 µm. At the same time, when modeling in Star-CCM+
solver, the pressure shock occurs at x = 0.4 m in the case of roughness height Ks = 20 µm and
x = 0.36 m in case of roughness height Ks = 32 µm. All these data indicate a qualitative agree-
ment of simulation results. While increasing the equivalent sand-grain roughness height,
the section in which the pressure shock occurs is shifted upstream (closer to the primary noz-
zle). The relative error of simulation results obtained by both solvers for the integral Entrain-
ment Ratio value throughout the numerical simulations was in the range of 5%, herewith
an ejector unit continued to work at “on-design” mode.
It can be seen from Figs. 6a and 6b that for the same sand-grain roughness height,
the section in which the pressure shock occurs coincides with a sharp decline in the characteris-
tic Mach number of the secondary flow. In this section, the flow velocity rapidly drops to sub-
sonic values. At the roughness height Ks = 0 – 32 µm, the calculated values of the primary and
secondary mass flow rates reasonably agree (with an accuracy from 3 to 15 %) with experi-
mental data. As it is shown in Fig. 5a, at the roughness height equal to Ks = 34 µm (simulation
results in Star-CCM+) an increasing in static pressure profile along the wall becomes more flat.
The comparison with the data in Fig. 6b indicates that the ejector switched to the “off-design”
mode. The injected (secondary) flow is not accelerated to the sonic velocity, the characteristic
Mach number value of the secondary flow does not exceed 0.6. Herewith there is a significant
decrease in the Entrainment Ratio, up to 30 %.
Numerical simulation results obtained at the sand-grain roughness height Ks > 34 µm
showed an analogous (as for Ks = 34 µm) feature of static pressure distribution along the ejec-
tor walls, which also corresponds to the “off-design” mode. The static pressure profiles along
the ejector walls as well as the abrupt pressure rise since the ejector has switched from “on-
design” to “off-design” mode qualitatively agree with the results of the work [19].
Thus, the authors supposed that there is a “critical roughness height” for the ejector’s in-
ner surfaces. If the wall roughness height is less than this critical quantity, then the Entrainment
Ratio value is quite close to the calculated one (for smooth walls under the given boundary
conditions and geometry), i.e., the walls roughness does not have a significant negative impact
on the overall ejector performance. In the case when the actual roughness value exceeds
the critical value quantity for given conditions and geometry, then the ejector switches to
the “off-design” mode, and the Entrainment Ratio dramatically decreases. It should be noted
that while increasing the roughness height, the static pressure distribution ahead of the pressure

559
D.V. Brezgin, K.E. Aronson, F. Mazzelli, and A. Milazzo

shock section has an increasing trend. Thus, the roughness height affects not only the pressure
shock location, but also the flow conditions ahead of the shock. This outcome qualitatively
agrees with experimental data shown in Fig. 5b.
The authors supposed, the roughness height influence on gas dynamics processes in su-
personic ejector can be compared with the influence of the backpressure (condenser pressure).
It is well known [1] that while increasing the backpressure the ejector operation range is re-
duced. Hereby, while increasing the roughness height, the associated energy losses with
the surface friction are increasing as well. This leads to a lower value of maximum backpres-
sure that a diffuser can withstand without incurring in flow reversal. As a result, there is a re-
duction of the ejector operation range. This outcome underscores the requirement of using
high-tech processes of the ejectors parts manufacturing, especially for the ejectors that operate
at supersonic modes. Losses due to wall friction are usually not the highest ones in the ejectors;
generally, the highest losses come from the pressure shocks, recirculation and mixing. Never-
theless, according to the paper [4], with a relatively small economic outlay, polishing the sur-
faces of the primary nozzle, mixing chamber and diffuser could lead to significant advantages.
It is also necessary to monitor the inner surfaces quality throughout ejector operation.

Conclusions
A comparative study of the effect of surfaces roughness on the logarithmic law of the ve-
locity distribution within boundary layer is carried out using two commercial CFD solvers
(Star-CCM+ and Fluent). The numerical simulation results of the ejector with smooth walls in
both CFD solvers have a good qualitative and quantitative agreement for both the mean inte-
gral characteristics and the local flow parameters distribution.
It is found out that either increasing the ejector backpressure or increasing the surface
roughness height, the section where the pressure shock is located is displaced upstream (closer
to the primary nozzle); the pressure shock is accompanied by a sharp decrease in flow velocity.
The numerical simulation results in both CFD-Solvers (Star-CCM+ and Fluent) of
the ejector with rough walls are well consistent in quantitative terms between each other in
terms of mass flow rates and local flow parameters distribution. Deviations in the static pres-
sure distribution profiles along the ejector walls are associated with the different approaches
for modelling roughness in Star-CCM+ and Fluent.
It is found out that when exceeding a “critical roughness height” for given boundary con-
ditions and geometry the ejector switches to the “off-design” operation mode, and its perfor-
mance is significantly reduced. The transition to this mode is followed by abrupt changes in
static pressure distribution profiles along the wall of the ejectors.

References
1. E.Ya. Sokolov and N.M. Zinger, Jet Apparatuses, Energoatomizdat, Moscow, 1989.
2. V.G. Tsegelsky, Two-phase jet apparatuses, Moscow State Techn. Univ. named after Bauman, Moscow, 2003.
3. ESDU No. 86030, Ejectors and jet pumps. Design for steam driven flow, 1986.
4. F. Mazzelli and A. Milazzo, Performance analysis of a supersonic ejector cycle working with R245fa, Int. J. Re-
frigeration, 2015, Vol. 49, P. 79−92.
5. F. Menter, J.C. Ferreira, T. Esch, and B. Konno, The SST turbulence model with improved wall treatment for
heat transfer predictions in gas turbines, in: Proc. Int. Gas Turbine Congress, Tokyo, 2003, P. 1−7.
6. Y. Bartosiewicz, Z. Aidoun, P. Desevaux, and Y. Mercadier, CFD-experiments integration in the evaluation of
six turbulence models for supersonic ejectors modeling, in: Conf. Proc., Integrating CFD and Experiments, Glas-
gow, UK, 2003. P. 561−566.
7. Y. Bartosiewicz, Z. Aidoun, P. Desevaux, and Y. Mercadier, Numerical and experimental investigations on su-
personic ejectors, Int. J. Heat Fluid Flow, 2005, No. 26, P. 56−70.

560
Thermophysics and Aeromechanics, 2017, Vol. 24, No. 4

8. P. Colonna, S. Harincks, S. Rebay, and A. Guardone, Real-gas effects in organic Rankine cycle turbine nozzles,
J. Propulsion and Power, 2008, Vol. 24, No. 2, P. 282−294.
9. L.M. Lujan, R. Serrano, V. Dolz, and J. Sanchez, Model approximation to an equation of state for the R245fa
fluid in an expansion process, J. Appl. Thermal Engng, 2012, Vol. 40, P. 248−257.
10. NIST Chemistry WebBook, NIST Standard Reference Database Number 69, 2003.
http://www.webbook.nist.gov/chemistry.
11. T. Knopp, B. Eisfeld, and J.B. Calvo, A new extension for k−ω turbulence models to account for wall roughness,
Int. J. Heat Fluid Flow, 2009, Vol. 30, P. 54−65.
12. CD-Adapco Star-SSM+ documentation help.
13. Ansys Fluent documentation help.
14. J. Nikuradse, Laws of Flow in Rough Pipes, NACA, Washington, 1950, Technical Memorandum 1292.
15. H. Schlichting, Boundary Layer Theory, 6th Edition, McGraw Hill, New York, 1968.
16. T. Cebeci and P. Bradshaw, Momentum Transfer in Boundary Layers, Hemisphere Publ. Corp., N.Y., 1977.
17. T. Adams, C. Grant, and H. Watson, A simple algorithm to relate measured surface roughness to equivalent
sand-grain roughness, Int. J. Mech. Engng Mechatron, 2012, No. 1, P. 66−71.
18. M.V. Zagarola and A.J. Smits, Mean-flow scaling of turbulent pipe flow, J. Fluid Mech., 1998, Vol. 373,
P. 33−79.
19. A.V. Sobolev, V.I. Zapryagaev, and V.M. Malkov, Single-stage ejector with high compression ratio, Thermo-
physics and Aeromechanics, 2005, Vol. 12, No. 1, P. 141−150.

561

S-ar putea să vă placă și