Proceedings of the ASME 2015 Gas Turbine India Conference

GTINDIA2015
December 2-3, 2015, Hyderabad, India

GTINDIA2015-1202

UNIVERSAL MODELING METHOD – THE INSTRUMENT FOR CENTRIFUGAL COMPRESSOR

GAS DYNAMIC DESIGN

Y. Galerkin, A. Rekstin, K. Soldatova, A. Drozdov

R&D Laboratory “Gas dynamics of turbo machines”
Joint Institute of Science and Technology
Peter the Great St.Petersburg Polytechnic University
195251, St.Petersburg, Polytechnicheskaya, 29
Russia
e-mail: [email protected]

ABSTRACT programs. The single stage 32 MW pipe line compressor was recently
Similarity theory principles are widely applied in gas dynamic designed for the Ukrainian partner who offered high-RPM drive and
design. But completely new solutions must be realized on a base of favorable single-stage scheme. The test of the model at the 1:2 scale
engineering approaches to predict performances. The heart of the validated project parameters. Design pressure ratio and efficiency
Universal modeling engineering method is the physical model that is curves have matched completely. The predicted maximum efficiency
based on flow visualization and measurements inside rotating 90% was proven.
impellers. The math model is a sum of algebraic equations for
calculation of head losses. Normalized velocity gradients along and NOMENCLATURE
normal to blade surfaces are arguments. Empirical coefficients values c2 absolute velocity at an impeller exit;
are established in a course of the identification – calculated
performances are compared with the measured ones for several dozens cw friction force coefficient; Critical streamlines change directions
of model stages tests with wide range of design parameters. The 4th near blade edges.
version of the TU SPb modeling method (the set of the several Dcu1 inlet critical streamline pre-rotation due to blade load
computer programs) was applied in design practice in 1993 – 2010.
Some Russian and foreign manufacturers realized several dozens of related to u2 ;
designed compressors with power up to 25 MW. The level of design Dcu 2 outlet critical streamline post-rotation due to blade load
point parameters prediction was so satisfactory that the manufacturers
related to u2 ;
do not prove designs by model tests anymore. The whole performance
prediction was not so good. The other difficulty – to predict design hw loss of head;
point efficiency with accuracy about 0,5% the very careful choice of k isentropic coefficient;
empirical coefficients is necessary. The difficulties have been
overcome in the new 5th and 6th versions. Most effective multistage Km empirical coefficient (viscosity influence on a loading factor);
compressors plant test performances were modeled successfully by 5th Mu blade Mach number;
version program with the single set of empirical coefficients.
Calculated performances and geometry of several dozens of stages of m mass flow rate;
these compressors can be are applied in designs as usual model stages. Ni compressor shaft power;
The current designs are executed by the 5th and 6th version computer p pressure;

1 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 R gas constant; operate in all basic industries. Industrial compressors life time is 3 – 4
Reu impeller diameter Reynolds number; decades. Many compressors are old. Development and renovate
processes in chemistry, metallurgy, etc. are increasing demand for new
Ro¢ non-dimensional normal velocity gradient; compressors. Energy saving prevails when industrial compressor is
T temperature; designed. The first step is gas dynamic design.
u2 blade velocity; There are many effective compressors now that are the base for
new designs by application of the similarity theory. Nevertheless the
w1 relative velocity at an impeller inlet;
world engineering companies offer their service in absolutely new gas
wsec velocity of a secondary flow; dynamic design and find clients [1, 2]. The team of the authors designs
Xi empirical coefficient in math model; 2-4 new centrifugal compressors annually [3].
The challenge of a new design consists of the next:
z number of blades;
- given pressure ratio must be guaranteed at given mass flow rate,
b fr disc friction coefficient; - efficiency must be as high as possible at a design point and in
bleac shroud disc labyrinth seal leakage coefficient; the widest range of flow rate,
- surge limit must be as far as possible from a design flow rate,
bbl blade angle to tangential direction; - special demands for gas dynamic curves – maximum power
a2 flow angle at an impeller exit; must be at a design flow rate for GT driven compressors,
- mechanical limitations must be taken into account.
lw max maximum velocity coefficient on a blade surface; Recommendations to choose main dimensions of a flow path are
p pressure ratio; presented in books [3, 4, 5, 6, 7], etc. The corresponding designs
t blockage effects coefficient in cascade; would be not absurd. But the listed above demands cannot be
guaranteed. The performance curves of the new design must be
j2 flow rate coefficient at an impeller exit; defined and must be compared with a buyer technical specification.
yp polytrophic head coefficient; The model tests were a solution at pre-computer era. CFD calculations
are offered now instead. The authors are not so sure in centrifugal
yi work coefficient; compressor performance curves validity calculated by CFD [8, 9]. The
yT loading factor; positive experience of CFD calculations of stator elements [8, 9] does
not change the situation. The engineering methods of design and
y T des design loading parameter;
quantitative analysis are actual as ever. There is a quotation from the
w angular velocity; text that belongs to one of leading western specialists [1]: "Some turbo
Dh loss of efficiency; machine engineers believe that the project can be begun with
application of methods of computing gas dynamics – nothing can be
Dhimp loss of efficiency in an impeller; further from truth. By the one-dimensional analysis it is necessary to
h- polytrophic efficiency; find optimum triangles of velocities, inlet blade angles, middle
F flow rate coefficient; radiuses, height of channels".
The way to calculate compressor performance curves
F des design flow rate coefficient;
p ,h , N i = f ( m ) is simple when non-dimensional stage performance
z imp loss coefficient of an impeller;
curves y ,h = f ( F ) are known:
z mix s coefficient of mixing losses; i

z fr coefficient of friction losses. k

p = (1 + ( k - 1)y i M u2 ) k -1
h
(1)
SUBSCRIPTS 4m
Where F = .
1 impeller blade row inlet; pr0t D22u2
2 impeller outlet;
Stage performances are objects of math modeling. The author of
bl blade;
[1] prefers application of boundary layer theory methods. Empirical
des design;
coefficients are defined on a base of experimental performances of a
h hub;
stage flow path elements. Authors of [10] use simple hydraulic
imp impeller;
analogies, as a diffuser divergence angle, etc. The authors of the
inc incidence;
presented work continue to develop ideas formulated still in 1970th
max maximum;
and presented in detail in [5]. Math model takes into account flow
mix mixing;
behavior. Physical model reduces flow behavior to a simplified
s shroud.
scheme. Math model calculates loss coefficients of stage elements by
algebraic equations with normalized non-viscid core velocity gradients
INTRODUCTION. CENTRIFUGAL COMPRESSOR
as arguments.
ROLE AND GAS DYNAMIC PROBLEMS
This kind of compressors is applied in all basic industries and FLOW BEHAVIOR
their installed power is great. In Russian pipeline industry only operate
The physical model takes into account flow behavior that has
4 484 compressor units with total power 51 000 000 kW (December
been studied by measurements and flow visualization. The
2014). Additional 540 units with total power 11 000 000 kW must be
summarizing of vast experiments is presented in [3]. Two samples of
put in operation in visible future. Comparable amount of compressors
the visualization are presented at Fig.1.

2 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 Thin powder inserted in the flow path sticks to surfaces with low case of the impeller that is shown at Fig. 2). Q3D calculations are
shear stresses visualizing separation zones (Fig. 1). The known “jet- applied for 2D and 3D impeller blade shape optimization – non-
wake” phenomenon is detected. The wake is of 3D character. Wake incidence inlet along a leading edge height, necessary loading factor,
formation takes place at a suction side at positive incidences, and at a mean blade load, control of velocity deceleration along suction side.
design flow rates in most cases too. Flow velocity near a shroud is the All this is important to obtain maximum efficiency.
biggest and its deceleration leads to earlier jet formation near a shroud. A loading factor is proportional to an area of a velocity diagram.
There is the zone of low shear stress in the corner near a hub. This An area of a calculated diagram is close to an area of a measured one.
zone is formatted by a secondary flow on a hub surface. The empirical coefficient correlates their values. Q3D calculations are
The special water test rig was constructed to study flow stream used as the final step of the load factor definition. It is a technique to
lines close to surfaces. Flow traces on wet oil paint that were left by guarantee given pressure ratio.
water flow are shown at the photo – Fig. 1. Wake starts at a leading
edge near a shroud at this flow rate. Jet-wake model is proven by
experiments. Meridional velocity in a wake is about 20% of a jet
velocity. There is no backward flow from a diffuser to an impeller.

FIGURE 1. SUCTION SIDE OF 2D IMPELLER FLOW

VISUALIZATION. LEFT - LOW SHEAR STRESS ZONE AT A
DESIGN FLOW RATE (AIR TEST), RIGHT – SURFACE FLOW
LINES AT A SURGE LIMIT FLOW RATE (WATER TEST) [5]

The important fact is that there is no flow separation at blades’

pressure side at any flow rate. No flow separation was detected at hub
and shroud surfaces at any flow rates. The phenomenon was explained
by Prof. E. Smirnov (TU SPb). Secondary flow produces the Coriolis
acceleration 2w × w oriented in the same direction as a core flow.
sec
The result is thin boundary layers with shear stresses preventing
separation.
The single-point pressure transducer with the switch for 33 points
was invented and constructed for pressure measurements inside
impellers (maximum RPM 18 000). Velocity diagrams at design
FIGURE 2. VELOCITY DIAGRAMS AT THE 2D IMPELLER
( F = 0,0852) and off – design flow rates are shown at Fig. 2.
des BLADE AT THREE FLOW RATES. SOLID LINES – Q3D NON-
The measured diagrams are based on measured static pressure VISCID CALCULATION, STROKE LINES – MEASURED
and calculated total pressure inside an inviscid core. The computer
program 3DM-023 for non-viscid quasi – 3D calculations the authors HEAD LOSS MODEL
apply in design practice from 1990–th [3]. Axis-symmetrical flow is The principles that are presented below were realized in the 4th
calculated by quasi-orthogonal method. Blade-to-blade solution on 7 version of the model and computer programs that were widely applied
surfaces along blade height is made by integral equation method. in R&D practice since mid-1990th [3]. The level of design point
Blades are substituted by vortexes. Non-penetration condition is parameters prediction was so satisfactory that the manufacturers do not
achieved in course of iterations. The advantage of the solution is prove new compressors by model tests anymore. The whole
correct calculation of flow near leading and trailing edges. Kutta- performance prediction was not so good. The other difficulty – to
Joukowski condition for trailing edges is applied. The flow character predict design point efficiency with accuracy about 0,5% the different
near a leading edge is predicted well. It gives the possibility to design sets of empirical coefficients for stages with different design
an impeller with non-incidence inlet at a design flow rate. Experiments parameters are necessary.
demonstrate that non-incidence inlet leads to the highest possible The problems were solved in the 5th version [11, 12, 13] applied
efficiency. It is one of conditions to reach highest efficiency. Both to stages with traditional 2D impellers and VLD. The 6th version
diagrams are very alike at a design flow rate. The principal difference widens this principle for other types of stages that was forced to add
is in a part where flow separation starts (very close to a trailing edge in

3 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 some further improvements. The general principles of modeling and There parameter Ro ¢ is a non-dimensional normal velocity
some special features of new models are presented below. gradient. For impeller blades it is:
The principle of loss summarizing is usual in engineering
calculation and analysis: wmean . (9)
Ro' = 4 -
Rbl mean
1 - å Dh (2)
h=
1 + b fr + bleac The empirical coefficients X (i ) take into account flow path
difference from the simplified model – thin plate. Velocity gradients
Here Dh is loss of efficiency of in of flow path elements: inlet influence boundary layers parameters, shear stress in particular.
nozzle, impeller, vaneless diffuser, vane diffuser, return channel or The interconnection of z and cw establishes the equation of a
fr
exit nozzle (volute, etc.). By tradition work coefficient is presented as
a product of mechanical work transmitted to a gas by blades – loading power loss calculated on a base of these parameters:
factor y , by outer disc surfaces b and work, provided by blades to
T fr

a gas flow in a shroud disc labyrinth seal b :

сwbl ze mean Sbl æ æ ws mean ö ö (10)
2 2
leac æ w p mean ö
z fr bl = 0,5 çç ÷ w + ç ÷ w p mean ÷
Ф0 ç è w1¢¢ ø s mean
è w1¢¢ ø ÷
y i = y T (1 + b fr + bleac ) (3) è ø
If cu1 = 0 then y T = cu 2 . - blade surfaces,
The disc friction coefficient b and labyrinth leakage b leac are
fr
e mean (11)
z fr o = 0,125cw о S о (1 + w& ) × w1¢¢
3
calculated by well proven semi-empirical way [14]. In case of an
impeller: F0

- hub and shroud surfaces.

z imp
Dhimp = w12 (4) Mixing loss model is based on equation of sudden expansion
2y T hmix = 0,5 ( w1 - w2 ) . The “jet-wake” model supposition is that flow is
2

leaving an impeller in a jet, while flow meridian component is zero in

The loss coefficient of an impeller blade cascade reflects the fact a wake. Mixing losses are proportional to difference of a jet meridian
of absence of mixing losses everywhere but a suction side of blades component and a meridian component of fully uniform flow after
(different kinetic energy at suction and pressure sides of blades is mixing. The empirical coefficient takes into account the real character
taken into account too): of flow:

(j - j¢2 )
2 2
æw ö æw ö
2

= (z fr s + z mix s ) ç 1s ÷ + z fr s ç 1 p ÷ + z fr hub + z fr shr (5)

(12)
z imp z mix imp = X (i)
2 jet

è w1 ø è w1 ø w1¢¢2
Where j¢2 = c¢m 2 , j2 jet = cm¢ 2 jet .
The object of friction loss modeling is the friction drag force Calculations by 4th version indicate that Eq. (11) underestimates
coefficient. There are equations of the Russian aerodynamic – classic mixing losses in impellers with high loading factor. In 5th and 6th
[15] for thin plate coefficient – rough and hydraulically smooth versions the empirical coefficient is presented as function of this
surfaces (turbulent flow without velocity gradients). Empirical parameter:
coefficients X (i ) are inserted in math model:

0,0307
(
X 20 = X (i) 1 + X (i)y TХ (i ) + X (i) (1 - w& )
X (i )
). (13)
c f smooth = X (i)
Re1/7
w (6) Flow deceleration to a separation point (suction side) was
presented as function of Ro¢ in the 4th version (Eq. 14). More
X (i ) (7) representatives are a loading factor (Eq. 14a):
c f rough =
1,89 + 1,62lg (1 / krg )
2,5

w& s = X (i ) (1 + X (i ) Ro¢ Х (i ) ) (14)

The 4th version used the Eq. (6), treating flow path surfaces as
hydraulically smooth. The new 5th and 6th versions take surface w& s = X (i ) (1 + X (i )yT X (i ) ) (14a)
roughness into account by Eq. (7).
Velocity gradients in a compressor flow path influence friction
Experiments have demonstrated that the velocity coefficient
losses. Math model representation is:
influence on losses depends on its maximum value on a profile near
the leading edge at a suction side:
cw = c f (1 + X (i ) Ro¢ ) + X (i) (1 - w& )
X (i ) X (i )
(8)
K M = 1 + X (i ) lwXs1( i ) , (15)

4 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 2k Pressures are different at blade surfaces p < p (plus – minus
velocity coefficient is defined by equation lws1 = ws1 / RT0 tot . s p
k +1 symbols at Fig. 3). Critical streamlines change directions near blade
Reynolds number and roughness influence is taken into account edges. Velocity triangle components D c u 1 and Dcu 2 are results of a
by eq. (6), (7). blade load. The proposed model substitutes a blade by a swirl with the
The problem above is treated in 2D mode. Real 3D flow character same circulation as a blade. A swirl creates Dcu1 and Dcu 2 . Non
is modeled by the proper coefficients. In case of a 3D impeller:
incidence condition for a design blow rate is:
X (i )
æ D - Dh ö (16) j1¢
K3D = 1 + X ( i ) ç 0 ÷ b1des = b1bl = arctg , ( Dcu1 < 0) (18)
è Dh ø D1 - Dcu1

The loss coefficient by Eq. (5) corresponds to non-incidence inlet

r The models for stator part elements are based in the same
flow rate. Flow velocity vector w is changing in an impeller throat principles and were revised in the new versions of the model.
r
(vector w1¢¢ in a throat) when flow direction does not correspond to an One of improvements realized in 5th version and used in 6th
version is the more accurate velocity diagram schematization.
inlet blade angle. The model for incidence losses is based on the
( )
Velocities near edges at both blade surfaces are necessary to know.
r r
principle of sudden expansion: hw inc = 0,5 w1¢¢ - w1 . The Eq. (17) Maximum velocity is necessary to calculate velocity coefficient lw max
realizes these principle meaning inlet velocity triangles components: that influences head losses. Solid thick lines in Fig 4 demonstrate
simplified schematization of velocity diagrams. Inlet and outlet
Dw12inc (17) velocities are known from 1D calculation. Blade load Dw depends on
Dhinc = X (i)(1 + X (i)(lu w1¢¢) X (i ) ) a loading factor and number of blades. Velocities at suction and
2y T
pressure sides are wsuct = w + 0,5Dw, w press = w - 0,5Dw . Numerical
Inlet flow becomes non-uniform in circumferential direction Q3D analysis of about 100 impellers of modern design has given
because of blades’ load. Critical streamline obtains a negative information about velocity diagrams that are close to real diagrams.
circumferential component Dсu1 < 0 . Streamlines in mid-span The results were approximate by algebraic equations. The effect
demonstrates Fig. 4 (thin lines).
obtain Dсu1 > 0 . The non-dimensional critical streamline
circumferential component Dcu1 is taken into account in a process of
incidence loss calculation. It is shown in a scheme presented in Fig. 3.

FIGURE 4. NON-VISCID VELOCITY OF IMPELLER

BLADES DIAGRAM AND ITS SCHEMATIZATION. SOLID
LINES – ‘TRAPETHOID” REPRESENTATION IN THE 4TH
VERSION, DOTS – CALCULATION BY EQUATIONS THAT
APPROXIMATES NUMERICAL EXPERIMENT

An unnoticed earlier difference of math model efficiency

æ k T ö
calculated by Eq. (2) and measured efficiency h = ln p / ç ln o ' ÷
è k - 1 T0 ø
influenced negatively on accuracy of calculations. The inaccuracy was
eliminated.
The positive effect of innovations has demonstrated the
FIGURE 3. SCHEME FOR CRITICAL STREAMLINES identification process. The accuracy of design regime calculation for
DIRECTION DEFINITION

5 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 38 model tests is inside 0,5% with one set of empirical coefficients for The empirical coefficient K > 1 takes into account flow
m
stages with very different design parameters. The same is accuracy in
efficiency calculations of about two dozens of industrial compressors. viscosity.
Good performance prediction of model stages and calculation of plant The alternative idea of modeling exploits linear character of the
test performances of different industrial compressors verified 6th function y = f (j ) too. The scheme is presented at Fig. 6.
Vectors cr , w
r are connected on the straight line inclined under
T 2
version of the model too [16].
2 2
the angle b . The tangent of this angle is:
LOADING FACTOR PERFORMANCE MODELING T

The experiments with all known model stages and compressors

demonstrated that loading factor y T = cu 2 / u2 is practically a linear j 2 max tg b bl 2 (21)
tg bT = =
function of the flow rate coefficient j2 = cr 2 / u2 . This function is u2 -y T 0 1 -y T 0

independent of compressibility criteria for a given impeller [16] –

The function y = f (j ) is presented as:
Fig.5. T 2

There is well-known model: an impeller with infinite numbers of

blades, non-viscid flow. The linear character of the function j2 1 -y T 0 (22)
y T (j 2 ) =y T 0 - =y T 0 - j 2
y T ¥ = 1 - j2 × ctg b bl 2 is evident. The linear character of function tg bT tg b bl 2
y T = f (j 2 ) shown at Fig. 4 is linear (inside test accuracy) for all tested
impellers. It is not easy to explain this phenomenon meaning wake
propagation at positive incidence angles (Fig. 1). This is not a reason
not to use the phenomenon in modeling. Loading factor performance is
defined by the equation:

y T 0 -y Tdes
y T = y Tdes + (j2des - j2 ) (19)
j2des

Values of y T 0 differ not too much for different impellers. There

is an empirical equation for their calculations. In design method
presented in [5] and in [3] the value of y Tdes is a design parameter. The
task of a designer is to find necessary number of blades and blade exit
angle to obtain this value. The scheme is presented above in Fig. 3.
Blade angle that corresponds to given y Tdes is:

(20)
j2¢
bbl 2 = arctg FIGURE 6. VELOCITY TRIANGLES AT AN IMPELLER EXIT
y Tdes
1 -y Tdes + K m
z (1 - D1 ) K pc
This scheme simplifies modeling process excluding some
parameters that was necessary to take into account earlier.
0,7
th th
NEW 5 AND 6 VERSIONS OF UNIVERSAL
3 MODELING AND SAMPLES OF APPLICATION
0,6
The 4th version of the model and computer programs was widely
1
applied in R&D practice since mid-1990th. Several dozens of designed
compressors with power up to 25 MW were realized by some Russian
0,5 and foreign manufacturers. Plant test performance of 16 most effective
2 industrial compressors were compared with calculations by the
programs of the 5th version [11, 12, 13]. Good correlation of design
0,4 parameters was demonstrated. Satisfactory modeling of the whole
performances was achieved by correction of empirical coefficients
responsible for incidence losses. The sample is presented in Fig. 7.
0,3
0,10 0,15 0,20 0,25 0,30

FIGURE 5. LOADING FACTOR VERSUS FLOW RATE

COEFFICIENT AT THREE MACH NUMBERS.
1 – M u = 0,6 ; 2 – M = 0,7 ; 3 – M = 0,8
u u

6 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 FIGURE 7. FOUR STAGE BOOSTER COMPRESSOR AIR TEST
PERFORMANCES (GREY) AND MODELING RESULT (BLACK)

The four stage booster compressor was designed by the 4th

version of Method. High efficiency was achieved due to a proper
design and optimal design parameters F , y . Performances of 68
des Tdes

stages of 16 compressors are added to database of model stages for use

in current designs.
The one-stage 32 MW pipeline compressor was designed by 5th
version program for one of industrial partners [17]. The tests of the
model at the 1:2 scale validated project parameters – Fig. 8. Project
performances, test results and CFX-calculations are related to the
complete flow path. Its cross-section is shown in Fig. 8 above. CFX-
calculation was made by the industrial partner [17].
Highest efficiency was achieved due to optimal design
parameters F , y as the partner has chosen high-RPM drive.
des Tdes
Proposed by the partner single-stage scheme is the most effective due
to axial inlet and zero hub ratio.
The prediction of the 5th version is quite good and the design
principles in a whole demonstrated their effectiveness once more.
The 6th version input menu for stage parameters with 3D impeller FIGURE 8. COMPRESSOR 32 MW MODEL CROSS
and graphics with calculated stage performances at different velocity SECTION (AHEAD) AND TEST RESULTS (BELOW). 1- DESIGN
coefficient are presented in Fig. 9. The 6th version input menu for stage PERFORMANCE, 2- ANSYS CFX,
parameters with 2D impeller and graphics with calculated stage 3- EXPERIMENT M u = 0.705, 4 - EXPERIMENT M u = 0.710,
performances at different velocity coefficient are presented in Fig. 10. 5 - EXPERIMENT M u = 0.700

7 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 FIGURE 10. INPUT MENU FOR COMPRESSOR
PARAMETERS AND A GRAPHIC WITH A STAGE
PERFORMANCE AT DIFFERENT VELOCITY COEFFICIENT.
COMPUTER PROGRAM CCPM-G6E

The program gives the possibility to optimize 11 main geometry

parameters of a stage. For instance, the efficiency of the stage with
F des = 0,105 was increased on 0.97% by more than 80 stage variants
FIGURE 9. INPUT MENU FOR STAGE PARAMETERS WITH 3D
IMPELLER AND A GRAPHIC WITH A STAGE PERFORMANCE comparison.
AND VELOCITY TRIANGLES. COMPUTER PROGRAM CSPS-
G6E

8 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo

 CONCLUSION [17] Tirtishnij I., Parafejnik V., Nefedov A., Rjgalskij S., Kotenko N.,
New versions of the Universal modeling method computer Timoshadchenko D., Mixajlenko S. The thermodynamic analysis
programs are applied in current design practice. The experience of of efficiency of the centrifugal compressor as difficult system as a
application demonstrates their new abilities and points out on part of the gas-compressor unit (32). 16 International compressors
possibility of further progress. The actual problem is to increase CFD- conference. – Russia. – 2014. – Book 1. – P. 328-339.
calculation modeling abilities. The authors have good experience in
CFD application for stages’ stator elements. CFD impeller
performance prediction is not completely reliable.

ACKNOWLEDGMENT
Work is performed with support of a Grant of the President of the
Russian Federation for young PhD MK-7066.2015.8.

REFERENCES
[1] Japikse D. Design system development for turbomachinery
(turbopump) designs - 1998 and a decade beyond. JANNAF
Conference. Cleveland. – Ohio. – July 15 -17. – 1998.
[2] PCA Engineers Limited. [digital resource].
http://www.pcaeng.co.uk/software.
[3] Galerkin Y. Turbo compressors. LTD information and publishing
center. - Moscow. – 2010 [Russian].
[4] Ris V. Centrifugal compressors. - Leningrad. – 1981. [Russian].
[5] Seleznev K., Galerkin Y. Centrifugal compressors. Leningrad. –
1982. – P. 271. [Russian].
[6] Кampsti N. Aerodynamic of compressors. – Мoscow. – «Mir». –
2000. – P. 688. [Russian].
[7] Aungier R. Centrifugal compressors: a strategy for aerodynamics
design and analysis. – 2000.
[8] Galerkin Y., Solovieva O. Improvement of vaneless diffusers
calculation based on CFD experiments. Part 2. Compressors and
pneumatics. - 2014. – № 4. – P. 15-21. [Russian].
[9] Galerkin Y., Marenina L. Investigation and perfection of
centrifugal compressor stages by CFD methods. International
Conference on Numerical Methods in Industrial Processes. World
Academy of science, engineering and technology. Paris 2015
Conference.
http://www.waset.org/Publications/?path=Publications. -
Vol:9 No:01 2015. – №. 249.
[10] Lunev A. Structure of a method of design and test of flow part of
superchargers for pumping of natural gas. Compressors and
pneumatics. – 2001. – № 10. – P. 4-7.
[11] Soldatova K. Universal modeling method loss and work input
models verification on results of industrial compressors test.
Scientific and technical transactions of the TU SPb. 2010. – No. 4.
– Page – 300-308.
[12] Galerkin Y., Soldatova K., Drozdov A. Specification of algorithm
of calculation of parameters of a stream in the centrifugal
compressor stage. Scientific and technical transactions of the TU
SPb. – 2010. – No. 4. – Page – 150-157. [Russian].
[13] Galerkin Y., Soldatova K., Drozdov A. Turbo compressor
efficiency application and calculation. Compressors & Pneumatics.
– 2011. – No. 8. – Page 2-11. [Russian].
[14] Galerkin Y., Soldatova K. Analysis of gas flow calculations in a
backlash between a covering disk and a case of centrifugal
compressor stage. Chemical and oil and gas mechanical
engineering. – No. 5. – 2007. – Page 27-29. [Russian].
[15] Lojtsanskij L. Mechanics of liquid and gas. – Moscow. – 1978
[Russian].
[16] Galerkin Y., Karpov A. Development of method for centrifugal
compressor impeller pressure characteristic modeling on results of
model stages tests. Compressors & Pneumatics. – 2011 – № 6. – P.
13-17. [Russian].

9 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86637/ on 02/01/2017 Terms of Use: http://www.asme.org/abo