STAGE DESIGN CONSIDERATIONS 273
FIGURE 12.96 Hosizoncally split fabricated compressor, lower easing ball. (Drescer- Runl Company,
Olean, NK
12.9 STAGE DESIGN CONSIDERATIONS.
most impartant elemet
iinpeller that the wir
treatment the gas receives both before and after the impeller i
¢ directs the flow
{diaphragm} are downs
a centrifuged «
ressur is the impeller. [tis by way of the
joduced at the compressor shaft en
sterred te the gs
portant, Typi
guide
6 the impelles, and a diffuser return bend and return channel
of the impeller (Fig. 12.98). Collectively, these parts make up
1 compressor stage, Some of the desizs considerations involved with these components are
discussed in this section.
Individual impeller designs are celated and evaluated by a pan
speed, which cla
neler called specific
ity of design, such as angles and propor
ne book, a bit of amplifi
impellers based eon simi
tions. Although this w
interest to the reader.
ation may be of274 DESIGN. CONSIDERATIONS AND MANUFACTURING TECIINIQUES,
riGU
2.9% Stulinnaty guts possiges making Up SIX st rer Rand Company, Odeun, 81)
wes. Pn
Specific Speed One aspect of similarity is called Kinematic similarity, where velocity ratios
between wo respective poims within wo separate designs are the same, Por example, if a
particular location velocity Vas compared to the tip velocity U is constant between two dil
‘ferent impellers and this is satistied throughout the flow field, kinematic s
fied. Substituting for the velocity V, the capacity Q divided by the diameter D squared, and
substituting for the tip specd U, the speed er 19, results in GHND, referred
to as the flow coefficient. This is also referred to as & capacity caefficiens. It represents the
velocity similarities between two de
larity is sat
mes the diam:
FN Oe Ne castant —$ flow coefficient ap
UND Ae
‘The second aspect of similarity is called dynannic siméharigy. Desigms ace considered to
be similar if the forces end the pressure fields are proportional to exeh other. This similar
ity is expressed by the pressure coefficient, which is defined as head HT divided by tip speed
U squared.
= constant pressure coefficient (122)
These two elements are combined to form specific speed, As can he seen, only operating
parameters remain: speed, volume, and head.
:
23)STAGE DESIGN CONSIDERATIONS — 275
J4 a9
mena
_*
£8 [=
He
a
. Fy
a a
cine seen
FIGURE 12.99. Eificiency
Company, Ofean,
ic speed conelation based om shop texts. (Dresser-Ruual
where No specific speed
Q > actwal flow, actual fO¥s.
The specific speed can be used in Iwo ways. First, in application of equipment, it points
ut where the impeller is operating relative te previous experience. The second important
use is the cartelation between stage efficiency and specific speed, Such a correlation exists,
a8 shown in Fig, 12.99, hased on test data, Also shown is the approximate shape of the
npellers in their respective specific speed area
For any particular compressor frame size, there must exist a range of available impellers ti
satisfy the flow and head requirements. They may lypicully cover a range of specific speeds
from approximately 20 10 over 200, Designs on the order of 200 to 300 have been appearime
in recent years because of x (rend in gus runsmission services af Ueereased head requirements
aud increased capacity levels, As stated before, impeller geometry changes with specific
speed, One can see the pattern of ehange in straight through compressors of numerous stages.
sus backward bent
ns, such as radial
It is worth looking xt some general consicl
blading, blade entrance angles, and impeller blade shapes. Radial-bladed impellers pro
duce the most head but exhibit « reduced range of operation with reduced stability eom-
pared to that achievable through backward leaning of the blades. It is for this reason that
most centrifugal impellers have backward-leaning blades, generally at angles of approxi-
mately 40 (o SOP, This inclination hus been found w produce a goed bekince between head
and operating range.
‘The desigi flows tow a particular impelle is determined by the blade entrance angle at the
let, in the stationary coordinate system, the
velocity magnitude and direction are represented by Cyy. In the relative system, rotatin
swith the blade at its velocity, the flow approaches in the direction and magnitude of WY. The
design point of a stige represents optimum efficiency: it occurs when the vector Wy is in
sis illustratedSTAGE DESIGN CONSIDERATIONS = 277
inclined cylinder. At bottom left, we depict a three-dimensional blade shape generated om a
om right is u wompletely arbitrary blade shape, The comples
ity and manufacturing difficulty of these shapes increases in the order mentioned, The
choige is made on the Basis of satisfying the aetedynamie requirements. If mere than one
type of blade is satistactory, the least complex design is selected.
Blade chic!
cover te the disk fer the typical closed design at the frame speeds desired.
blades have ts be structurally sound.
The largest flow impellers far a particular frame size are of the open type, with no
allached cover, as shown in Fig. 12.39, Radial blading and elimination of « cower on this
particular design provide for a maximum amount of flow and head in ane stige, yet kee!
stress levels sufficiently low, even in large diameters, to allow acceptance of maxi
material yield point eriteria as specified for H)8 service. Note that API 617 stipulates the
use of 90 00K) psi maximums yield strength, RC-22 hardness (235 BHN) steels. For ease of
manufucturing, the minimum yield strength of carbon steels generally applied in HS serv
ive is 80.000 psi
‘Open impeller blades can be backward curved if desired, with some sacrifice to head
and special consideration 1o blade stresses, The typical open impeller has an almost axial
ot section, referred teas an inducer section. This impeller is typically used in the first stage
and requires more axial space than the smaller designs. The primary advantages of this
design are higher flow and pressure ratins, with some sacrifice to stability and efficiency.
‘The impeller is designed to run with generous clearances (approximately 2 in. for a
40-4 phery and the stationary shroud. It is industry
pructice to provide one or (we stages of backwars-curved impellers behind this radial fhow
impeller to praduce a rise to the sunge performance charseterist ven section,
‘A varistion of the open impeller design is to equip it with a cover, Some gain is made in
efficieney, but the attached cover inereases the stresses and Inwers the maximum speed
capability significantly.
“The typical welded-elased impeller used in the vast majority of applications is ane of
three designs: three-D welded, three-piece welded, or two-piece (milled-welded) construction.
‘The three-D welded (three-dimeasional) hus u blade shape that is a portion of u rolled eon.
ical or cylindrical surface, or may be @ combination of the two. An example is shown in
Fig, 12,102. These blades are posidoned on the isk or cover ata predetetmined inclination and
location. The blades then form a thiee-dimensionsl contour with respect to the eover or disk
iow, and inclination to satisty the aceodynami-
oni
terion, hecause the blades must hold the
ewer, the
fesses are determined hy a stress
um
diameter impeller} between ils pe:
The choice of cane oF cylinder size, lo
cally required angles at the leading and trailing edges of the blade is difficult. In the past, it
ations
swas not unusual for a design drafter to spend several weeks trying different eo
before urriving at a sutisfustory geometry. Use of the computer has reduced this time to
minutes, which ullows the designing af more than one blade [or review und detailed analysis.
The basic advantage of three- dimensional impellers is beter performance, with higher
efficiency. However, material, tooling,
closed impellers. They also require mor
The three-piece design (Fig. 12.103) is the next step down in complexity. This type of
mpeller also has three basic components: blades, cover, and disk, welded together, This
construction allows mote freedam for aerodynamic design than dees the milled-welded
impeller design hut follows. The blades, however, do not form a true thr
eonteur, Instead, they are rolled into a cineular shape, The impeller requires less axial spac
ing than the tivee-D type diseussed earlier and is less complex to manufacture.
d Welding requirements are higher than with ether
axial space in the machin
imensional278 DESIGN CONSID
ATIONS AND MANUFACTURING TECHNIQUES
REA2A02 Threw
ler: (Dresser Rand Cc
I wold
pany, Clear, NEI
FIGURE 12.103 Three piece impeller design. (resser Rarul Company, Olean, N.Y)
The tworpicce impeller (Fig. 12.104) has the blade:
therefore requires welding
and requires u minimum of stuge spacing, The two-piece impeller th
eatures of good performance and ease of m:
the range of impeller sizes availible, The
and the st
milled onte the d
or eover and
nn-one side only. This type of impeller alse has good efficieney
re combines the
ure 12.LD5 gives un idex of
ange refrigeration machine
seeviees stich as gus injection.
I rotor is typical of high pressun
‘Onee the flow channel through the impeller is set, the Wade, disk, and cover eontours
and thicknesses must he developed consistent with the anticipated speed. Welded impellers
ently complex to on snalyses. A. man
ms are typically hacked up by extensive testing of prototype and/or
sdate computer-based stress and deflect
ufacturer's stress prc280 DESIGN CONSH
ATIONS AND MANUFACTURING
HINIQUES
‘We can now look ut some considerations involved with the design of the other pieces that
contribute lo the stuge, such as the auide vane, diffuser, selurn bead, and return ebuanel, as
dkpicted earlier in a six-stage compressor (Fig, 12,98), The efficiency ofa given stage isa func
tion of the Friction and diffusion losses theeu gh the stage components. These loss meetsanisms
can be used toexplain the shape of the ei Hig. 12.99}. Peak
efficiency decreases as the specific speed is reduced orincreased from an optimum range. This
comes shout as 2 result of the combined friction and diffusion losses reachin
Triction losses in a straight pips are proportional to the velocity squared. The larger the
Pipe area, the lower the losses. Diffusion losses for that are changing are propartional
te the velocity sitio Vigessce Voce 1 the fourth power, Furthermore, lsves are incurred by
any bends and are proportional te the degree of turning and the tightness of the tain, The
sson in all shis is that unnecessary diffusion and bending should be avoided in high
ncy wersus specific speed cur
erformance compressors.
‘Charactcristically, friction losses increase at the lower specific speeds. This is duc to the
increased wetted surftce and smuller hydraulic esunnel diameters. Diffusion, on the other
hand, increases a the higher specific speeds, reflecting the effects of larger gas capacities
being wmed in tight bends, When the two losses are added, an area of minimum loss and
muximum efficieney results, shown in Fig, 12.99,
The internal geametry of a compressor is more complicated than that of a bent ex dif
fusing pipe. To.calculate the velocities within a machine, a manufacturer may use advanced
numerical solutions, whereby the geametry would be defined and the velocity fields that
must satisfy radial equilibrium and continuity at every spatial point in the flow field would
be calculuied iterstively, Typically, for a compressor stage analysis, the velocity field from
upstream of the impeller ts upstream af the next st
To optimize the elficiency, the zeomeiry of each component is reviewed in terms of
velocity and velocity grulients. Unnecessary accelerations are eliminated and diftusion
losses are minimized. The return bend vadii are generents, the retuen channel ecvatours are
specially shaped, and the areas through the channel and guide vane are closely specified.
The velocity distributions for a final-stage geometry are shown in Fig. 12.104, from the
trailing edge of one impeller io the leading edge of the subsequent impeller. The velocity
P
st
FIGURE 12.106 Velocity aise
jms foe Final stage goonnctiy.(Drester Rand Coonan Olea, .¥)STAGE DESIGN CONSIDERATIONS — 281
decreases from the first impeller tailing edge through the vaneless diffuser in a uniform
samanner.
This is followed hy the return hend, where the maximum velocities on the
the minimum velocities at the euter wall are proportional to the severity of the bend and the
local surface curvatures. ‘The diffusion losses on both surfaces have heen minimized.
Followeing the ectuen lend, the return channel velocities are shown. The mean velocity
through the channel is represented by the middle line: the upper nd lower velocities are
those ulong the vane surfaces
The return channel vanes are optimized. The vane shupe. thickness di
jer wall and
jstcibution, and
number of vanes ure aaulyzed in coujunetion with the channel axial height to arrive at the
final coordinates,
Following the return channel is the guide vane, which tums the flow from ra
inward flow i
Unnecessary accelerations and decelerations are carefully scrutinized and minimized.
Next is the impeller, which must be designed consistent with the approach velocities
leaving the guide vane and, in turn, impart the energy increase to the fluid, The process
repeats ilsell for as muny stages as there ave in the unit
In summary. the velocity decreases in the diffuser downstream of the impeller, it is then
turned from radially out t6 radially in, the tangential velacity or swiel is removed in the
diaphragm er return channel, and finally, the flow is auened from the radial to the axial
direction and enters the next impeller.
Inlet guide vanes, shown in Fig. 12.29, also provide one method of contralling stage
performance. because they may be used to direct the flow into the impeller at different
angles: against impeller rotation, radially, or vith impeller rotation, The influence of vari
sous guide vane angles on a given impeller head characteristic is shown in Fig, (2.107
ally
‘ong the inner contour.
axial flow. This turning results in aceelem
v0
é
too sf
¢ Z
= 8
# Against
oe
cd tay ton
a whee
2 a relabon
2” rotation
2 50 75 100 128 150
e Mass to
FIGURE 12.107 Guide vane angles ws. impeller head characteristics. (Dreier Rand Companys
Olean, N.¥)282 DESIGN. CONSIDERATIONS AND MANUFACTURING TECIINIQUES,
URE 12.108 Impeller inlet diagram. (Dresser
Remi Company, Olean, NT.)
How this effet
inlet tangential velocity or swirl exists becausc of turning guide vanes, the inlet triangle is
modified. Por a particular value of inlet tangential velocity, C,,. there will be & through
flow yo Cy. whieh veil result in u good approach lo the impeller blading. This
is where the peak efficiency will occur.
saused is shown in the impeller inlet diagram (Fig. 12.1081, When
12.100 IMPE
RR MANUFACTURING TECHNIQU
As mentioned in Section 12.5, the riveted impeller has heen around for many years. There
are t8vo basic types of construction, one being two-piece with milled vanes: the other is
three-piece with separate disk, ov
Lange-capacity impellers require three-piece construction (Fig, 12.103), The caver and
disk are thbricated from forgings, usually of either alloy steel ora 400 series stainless. The
blades illy die-formed from stainless steel eal ta the disk and cover by
short stainless stee! rivets extending eheewgh the disk or cower material and the flange prev
vided along the blade by the formi
7, and blades284 DESIGN CONSIDERATIONS AND MANUPACTURING TECHNIQUES
FIGURE 12.110 Narrows
ch
lth impeller, (Bresser-Rand Company, Olean, N.
Welding tables peovide for proper preheat during the welding, process, These tables are
provided wit heating elements in the tible and cover to maintain material teuperatires a
approximately 600°F, The tables ave mechanized ta permit rotation of the workpiece under
a stationary open flame, oF toch,
Final meehanieal properties are obtained by heat-treating (normalizing or quenching,
with subsequent tempering) after welding. Typically, impellers then undergo final machin-
ing, dymumic baluncing, und overspeed testing to at least 115% of maximum continuous
speed, as specified in API 617.
The general manulucturing wchniques used to produce the different types of welded
impellers vary somewhat and Will be looked at next, For a three-picce-construction impeller,
the disk and cover are machined from fewgings. The dlie-formed blades are then tack-welded
to the eover with the use of locating fixtures. The final welding is a continuous fillet weld
between the blade and cover. Subsequently, the blade cover assembly is joined to the disk
by x continuous fillet weld between disk and blade
An open impeller design consists of a disk and blades. The cover is climinuted. As dis-
cussed previously, this type of impeller is characterized by an inducer section that directs
the gos How into the eye of the Impeller, The blades are either die formed oF piecision cast,
‘The welding procedure is the same as for thee-piece construction, with, the final weld
boing a continuous fillet weld henween the disk and blade, For two-piece construction, the
blades are machined on either the disk ar cover forging. ‘The impeller is completed by a
continuous fillet weld to the mating pieee (disk or cover} arnund the entire blade interface.
‘The welding techniques described are limited to impellers with a flow chunnel width of
more tha ¢ in., (0 allow insertion of a torch, Increusing numbers of applications require the
Advantages of welded construction for impellers with channel widths of less than 2 in,
(Fig. 12,110), One method of Joing this has been to weld through fiom the back side
(Fig. 12.1111. This type of impeller is manufactured fram a disk forging and a cover torp-
ing, thus heing of two-piece construction. Impeller blades, integral with the cover, arc
formed by removing metal from the inner face of the cover. A matehing slot, corresponding
to the blade contour, is machined in the disk at each blade locution. Afler machining, the
disk is locuted precisely over the cover with the blades uligned to the slots in the disk. The
ass THG weld (Fig, 12.112), An internal fillet is
thus formed at the blade to-disk junction, Using this technique, welled impellers with the
slot is then filled with « continuous multi25,
FIGURE 12,111 Slot welding a two-piewe impel.
Ter, (Dresser seamed Campury, Olean, 8%)
FIGURE 12.112 Slots in a disk being filled. (Drevser-Rosf Company, Giean, NY)
smallest practical channel width can he manufactured. Any blade contour may he designed
without affecting the weldahility of the impeller, thus minimizing compromises in acrody-
namic design. The impellers (Fig. 13.113) are available in the same materials and range of
properlies as the fillet-welded impellers.
Te ensure quality, the following gen
on welded impellers
inspection procedures are typivally carried out
+ Matcrial certification reports are reviewed for compliance with specifications
+ Test bars for each forging are serialized snd follow the impeller through all best
treatment cycles288 DESIGN: CONSIDERATIONS AND MANUPACTURING TECHNIQUES:
The effort to minimize bearing span manifests itself in various ways in compressor
design. These include
+ Position
he bearings as far inbeard in the case end enclosures (heads) as possible
+ Minimizing end seal lengths
+ Using wwiable stage spacing
In the past, many easings were designed base