-MOLON Tatas) ae AERODYNAMICS POL lenly rieContents ‘ Preface ‘Acknowledgements Notation ¢ 1 Useful basic ideas 1 6S 1.1 Introduction 1 1.2. Blades and flow 1 € 1.3 Work input into compressors 4 1.4 Dynamic scaling i 1.5. Losses 21 ‘ 1.6 Efficiency 34 +, 2 General design considerations 46 Go 2.1 Introduction - 46 2.2 . The axial compressor a7 wo 2.3. The radial compressor 2 Ll 2.4 The matching of multistage compressors 8 os 3 Throughtlow on the hub—casing surtace and some “ aspects of flow in three dimensions 93 ‘ 3.1 Introduction 93 3.2 Approximations applicable to axial compressors: simple radial equilibrium o ‘ 3.3 Early developments 102 3.4 Practical methods for the meridional flow 106 3.5 Applications of streamline curvature methods in axial compressors: 4 3.6 Mixing in multistage axial flow compressors 12 3.7 Axial compressor off-design tends 126 « 3.8 Flow chart — use of a streamline curvature method in analysis mode for an axial compressor 129Addison Wesley Longman Limited, Edinburgh Gate, Harlow, Essex C20 258, England and Associated Companies throughout the world © Longman Group UK Limited (989 All rights reserved; no part ofthis publieati nay he reproduces, stored in a retrieval system, or transmitted in any form ot by any means, electronic, mechanieal, photocopying, recording, or otherwise ‘without either the prior written permission of the Publishers or a licence permitting restricted copying, in the United Kingdom issued by the Copyright Licensing Ageney Ltd, 90 Tottenhain Court Road, London WIP SHE, First published 1989 Reprinted 1996, 1997 and 1998 British Library Cataloguing in Publication Data Compsty, NAA. Compressor aerodynamies 1, Compressor, Aerodynain L Title 62151 ISBN O-582-033b4-X Library of Congress Cataloging-in-Publication Data ‘Cumpsty, NA ‘Compressor aerodynamics / N.A, Cumpsty pcm, Bibilography: p. Includes index. ISBN 0-470-21334-5 (Wiley): 1, Compressors ~ Acrodynamics, I. Title TI267.5.C5C86_ 1989 621.5" Ide 19 Printed in Singapore through Addison Wesley Longman China LimitedBlade-to-blade flow for axial compressors with subsonic inlet flow 132 4.1. Introduction 132 4.2 The effect of blade shape 140 4.3 Loading limits for blade rows 149, 4.4 — The selection of incidence 159 4.5 The prediction of deviation 168 4.6 The determination and prediction of losses 171 4.7 ‘The effect of Reynolds number on blade performance 176 4.8 The effect of inlet Mach number on blace performance 180 4.9 Concluding remarks 191 Blade-to-blade flow for axial compressors with supersonic inlet flow 194 5.1 Introduction 194 52 Choked flow with attached shocks ~- “unique incidence’ 198, 5.3. Operation with detached shocks 208 5.4 Shock structure and the nature of flow in supersonic rotors 209 5.5 Losses in supersonic blading 214 5.6 The design process for supersonic blades 27 The centrifugal impeller 220 6.1 Introduction 220 6.2 The flow pattern in impellers 223 6.3 Calculation methods and predictions of flow in i 236 6.4 Slip and the estimation of slip factor 245 6.5 Loss in impellers 249 6.6 Design choices for the impeller 254 The diffuser of the centrifugal compressor 266 7.1 Introduction 266 7.2 ‘The nonuniform flow from the impeller + 269 7.3 The vaneless diffuser 276 7.4 The vaned diffuser 285 7.5 The volute or scroll 301 Viscous effects in compressors a0 8.1 Introduction 82 Throe-dimensional viscous flows in compressors 3110 1 8.3. Axial blade boundary layers 8.4 Flow in the endwall regions of axial compressors 8.5 Viscous cffects in centrifugal compressors Stall and surge 9.1 Introduction 9.2 Insta 9.3. Post stall behaviour 9.4 The flow in the rotating siall cell 9.5 Stability enhancement: easing treatment Vibration and noise 10.1 Introduction 10.2 Vibration Mechanical vibration modes Forced vibration Flutter Supersonic unstalled flutter 10.3 Noise Scales and rating of noise Elementary acoustics Compressor and fan noise Non-acronautical aspects of compressor noise Acoustic treatment Design, Measurement and computation 11.1 Introduetion 11.2. Understanding and design 11.3. Experimental techniques 11.4 Mathematical techniques Appendix Blade profile families for axial compressors Bibliography Index 320 331 356 359 359 369 301 398 aot 410 ALO 410 412 als al7 422 428 429 431 440 455 457 459 459 459 461 466 479 484 505ke Preface ‘The idea for this book was given to me by Dr R C Dean early in 1973. His reason was that Axial compressors by J H Hotlock was out of date. That was 15 years after the publication of Axial compressors and another 15 years have elapsed between then and now. [am sure that it is fortunate that I did not try to write a book at that time. The title of this book may seem a little bald with so many books being en- tifled “An introduction to” or ‘Fundamentals of” but that is not a style which, like. Tam however very aware that what I have written is only scratching the suface of a very big topic. It has also been a surprise to me how much Thave had to leave out: in particular, discussion of ideas which are new (and may not be correct) but are currently interesting. Although I had set out to write a short book, I seem to have failed in this. In stressing that this is inevitably a brief account of a very big field, onc occupying hundreds of people and vast expenditures of money, it must be realized that most of this work is being carried out by companies and the results are predominantly confidential. It is one of the main difficulties that much of what really matters is never releas into the public domain, a trend that is more pronounced now than when Axiaf compressors was being written, for example. One of the consolations for an author is that those people inside the large companies who have access to this knowledge are not likely 0 write a book about it. Writing a book about compressors thitty or so years ago, it probably seemed reasonable to give a guide to design and this is evident in, for example, Eckert and Schnell’s book Axial- und Radialkompressoren as well as in Horlock’s Axial compressors. Writing today this does not seem so appropriate. This is partly because of the greater concentration of knowledge and expertise in con- fidential form, referred to above, but also the greater use of numerical methods. The computer has come to dominate the design and assessment of compressors, 80 that it is not very likely that a major design would be carried out nowadays without the extensive use of numerical methods. Such methods are not necessarily described at all well in design manual form, In short this book does not set out to be a design guide but to provide ideas and clarification which will in turn lead to improvements in design. It is my hope that com. pressor aerodynamicists and designers will find it helpful in this way and that it will have a use in introducing newcomers to the topic. Some of the sections are relatively specialized, others are quite general and have application to all turbomachines and to wide areas of fluid dynami On the whole I have not included much mathematics; someone wanting the details will be able to find them in one of the referenced works. Again theshift to numerical methods has somewhat diminished the attraction of the cow plete mathematical coverage of a topic and mathematical analysis has become relegated to certain specialist fields, mainly those such as stall or vibration where linear analysis is valid, As a rule equations are given when they are needed to follow the argument. In so far as Tam able I have held my argument together by using results in geaphical form and as a result there are a lot of igures in this book, | would be giving a wrong impression if I suggested that to be modern is to relegate what was done in the past to obscurity. There is, I believe, a lot to be learned from the work which was done in the 1940s and 50s and I have included some of this, Naturally the areas where this is most apparent are those in which the early workers were most active, so in the chapters on blade-to- blade flows in axial machines and on centrifugal impellers there is quite extensive reference to much of the excellent early work. ‘This does not, however, set out to be an historical account attempting to put the record straight about who did what. Where it is appropriate [ have referred to the originators of ideas, but this is not the primary purpose of the book or of the reference list, Ihave, of course, included any reference from which I have taken a diagram as well as those which I think will be useful, cither because they are themselves interesting or because they in turn provide a very complete bibliography. Inevitably many first-class pieces of work are not referred to, though the person wishing fo dig deeper will find them from the reference lists in other papers. In general I have tried to give references which are easily obtained and this necessarily distorts history to some extent “Thus, for example, I have referenced Day and Cumpsty (1978), which is easily available in libraries around the world, whereas the work is first described by Day in a doctoral dissertation held in the library of Cambridge University. ‘To the majority of readers this distortion will probably seem a price worth paying for convenience. ‘As the author I feel more comfortable with some topics than others. In weiting about blade vibration [was very conscious of having a fairly superficial grasp of the topic. I am very grateful to Dr D $ Whitehead for his help with this and on the strength of this feel confident to see what I have written published. Similarly the computation of flows in urbomachines is a specialized fold with which 1 felt unfamiliar, In Chapter 11 there is a relatively short section on numerical methods which is based on a lecture by Dr J D Denton who was kind enough to review what I wrote; again with this support I am prepared to see what I have written published, believing it to be free from the worst errors and a helpful introduction for the non-specialist, ‘A sense of realism tells me that there will be mistakes which slip through into the published text, In the hope that there will ultimately be a secon! edi- tion J should be very pleased to have readers send me corrections, most par- ticularly corrections of fact. If they have illustrations which they think would help a second edition to be clearer, perhaps even replacing existing examples, T should be very pleased to receive them.Acknowledgements 1 focl that in writing this book in the Whittle Laboratory I have been very for- ‘unate; the combination there of resources, ability and a willingness to be helpful ig all that one could ask for. My contacts in the industry have served me very well and I can only list those who helped me most substantially. At the outset the encouragement of Dr L H Smith of General Electric was crucial and since then conversations with him and with Mr C C Koch and Dr D C Wisler of the same company have helped me a very great deal. Mr D P Kenny of Pratt and Whitney of Canada likewise helped by his encouragement and by discussion. My long- standing friend, Mr C Freeman of Rolls-Royce, has taught me a very great deal over the years and in particular whilst I have been working on this book. Dr M V Casey of Sulzer Escher Wyss has encouraged and helped me and given guidance in areas where I have had less experience. I also appreciate the help of Mr D Tapikse of Concepts Inc. and Dr HG Weber of the Cummins Engine Company. In academia no one could have done more to help me than Professor EM Greitzer of MIT, both by encouragement and by argument. Dr T P Hynes, with an office next to mine in the Whittle Laboratory, has been an invaluable colleague to test my ideas on and I owe him a great deal, both for his exceptional forbearance as well as his great abilities. (He also took a very active role in the section on matching of multistage compressors described in Chapter 2.) My debt to Drs Denton and Whitchead I have referred to in the preface, but to this must be added the use of computer programs they have written. Dr Dawes has helped me with calculations and advice, Others in the Whittle Laboratory have helped me substantially, in particular my students and former students who had the misfortune to be here when the book was being prepared; Drs § G Gallimore, Y Dong, and N McDougall and Mr Y $ Li have helped both by their influence on my ideas and by carrying out calculations for me. ‘There is a special note of thanks I wish to record to those who have helped me with the book in its later stages. Foremost of these is Professor Greitzer who has done his hest with savage injunctions to improve the book and has read most of the chapters for me. Dr Casey read three chapters and made invaluable contributions. Mr Freeman and Dr Gallimore took great pains over Chapter 3 when T was having a lot of trouble with it. Mr M Howard checked Chapters 4 and $ for me. Dr McDougall has helped me in many ways, but [ particularly appreciate his assistance with the later stages when he himselfwas under pressure. The contribution of the following in reading and check- ing the proofs is gratefully acknowledged: Dr D Andrews, Mr L He, Dr D Lambie, Mr Y $ Li, DrJ Longley, Mr J A Storer and Dr M Zangeneh-Kazemi We ate indebied to the following for permission to reproduced copyright material: Academic Press Ine. Ltd, for figs. 10.21 ftom plate 3 (Cumpsty 1972) and 10.29 from fig. 5 (Mariano 1971); Advisory Group for Aerospace Research and Development, North Atlantic Treaty Organization (AGARD/NATO) and the respective authors for figs. 3.8 ftom p. 268 (Hirsch and Denton 1981), 4,47 from fig. 7.7 (Scholz, 1977), 5.8 from figs. 6 & 7 (Wood er al. 1986), 8.5 from fig. 10 (Seyb 1972), 9.6 from fig. 11 (Riess & Blocker 1987), 9.38 from figs. 9 & 10 (Mikolajczak & Pfeffer 1974), 9.45 & 9.46 from figs. 11 & 13 Gansen e/ al, 1980) and 10.1 from fig. 3 (Sisto 1987a); American In- stitute of Aeronautics and Astronautics for figs. 3.14 from figs. 2 & 5 (Smith 1974), 4.49—4.51 from figs. L, 6 & 8 (Schmidt er al. 1984), 5.2 from fig 12 Gchreiber & Starken 1981), 8.6 from fig. { (Walker 1987), 10.4, 10.12 & 10.13 from figs. 7, 1L & 12 (Mikolajczak et al. 1975), 10.24 from fig. 2 (Lowrie 1975) and 10.28 from fig. 8 (Ginder & Newby 1977), American Society of Mechanical Engineers for material from ASME Transactions; Elsevier Science Publishers B.V., Physical Sciences & Engineoring Division for figs. 3.15 & 9.10 from figs. 2 & 10 (Smith 1969) and 4.39--4.41 from figs. 10, 20 & 21 (Schlichting & Das 1969); Gas Turbine Society of Japan for fig. 2.4 from fig, | (Freeman & Dawson 1983); General Electric Company for figs. 2.12.3 (Wisler 1988) Copyright © 1988 by General Electric Co., U.S.A. All Rights Reserved; the Controller of Her Majesty's Stationery Office for figs. 4.5 from figs. 4, 6 & 14 (Andrews 1949), 4.34 from fig. 6 (Carter 1950), 4.37 & 4.38 from figs. 13, 25 & 36 (Rhoden 1952), 8.22 from fig. 13 (Lakshminarayana & Horlock 1967) and 10.9—10.11 from figs. 2, 10 & 12 (Halliwell 1975); Institution of Mechanical Engineers for figs. 2.10 & 4.10 from figs. 68 & 81 (Howell 1945), 6.11 from fig. 3 (Casey & Roth 1984), 8.8 from fig. 6 (Abu-Ghannam & Shaw 1980) and 9.32—9.36 from figs. 3, 8, 13, 14 & 20 Day & Cumpsty 1978); Institute of Refrigeration for figs 2.23, 2.24 & 6.18 from figs. 4, 5 & 7 (Casey & Marty 1986); the author, Prof W.C. Reynolds for table J. and fig. 1.17 from pp. ix and 161 (Reynolds 1979); Rolls-Royce ple for figs. 4.25 (Carter 1961) and 4.44; Royal Aeronautical Society for fig. 10.2 from fig. | (Armstrong & Stevenson 1960); Society of Automotive Engineers Inc. for figs. 2.22, 2.13, 2.14, 6.20, 7.19 & 7.20 from figs. 16, 20, 25, 30, 32 & 34 (Kenny 1984) and 10.16 & 10.17 from figs. 13¢ & 15 (Tyler & Sofrin 1962) © 1984 & 1962 Soriety of ‘Automotive Engineers Inc.; Von Karmen [nstitute, Belgium for figs. 7.17 7.29 & 7.30 from figs. 18—39 (Stiefel 1972) and 8.16, 8.18, 8.21, 8.25 & 8.26 from figs. 1 and 33-44 (Freeman 1985).Notation ‘There are very many notations in use for the consideration of turbomachines and it is just about impossible to evolve a system which has no duplication of symbols without recourse to excessive use of subscripts. It is hoped that the system adopted will represent a reasonable compromise which is fairly transparent and agrees with that generally used for the topic being discussed The overlap that does exist here (for example m denotes mass flow rate and meridional distance) should not confuse the reader too much. There are other inconsistencies (as Emerson wrote, a foolish consistency is the hobgoblin of lite minds) but this should not be too irritating, The list given is not an exhaustive one and various additional symbols are introduced throughout the book General points ‘Throughout the book all angles are measured from the meridional flow diree- tion, which reverts to the axial direction for the blade-to-blade flow in axial ines and the radial direction towards the outlet of centrifugal compressors (Lhe terms radial or centrifugal compressor are used as alternatives without any implied difference; both are in common use). ‘The velocity magnitude and direction in the relative or rotating frame of reference are denoted by W and @ whilst in the absolute or stationary frame of reference they are denoted by V and a. As is comtnon in British and American practice for compressors, a conven- tion of positive and negative signs for flow or blade angles is nor used; angles ae taken as positive and the appropriate sense adopted The word stagnation is normally used, as for stagnation enthalpy lg = h +¥/2, and not the word total. The usage total-to-static, as in total- to-static efficiency, is so widespread and the corresponding term based on stagnation so much harder to say that this is retained. ‘The outer diameter of axial machines is sometimes called the tip. This may be ambiguous, for the tip of the stators is at the hub, The word casing is therefore preferred for the outer diameter. ‘The phrase hub—tip ratio is so common that this is occasionally used in place of hub—casing ratio for the ratio of the hub diameter divided by the casing diameter,Variables commonly used Geometric variables b passage width in spanwise direction, used for centrifugal compressors c blade chord 4,D diameter g staggered gap, pitch resolved normal to the flow direction h blade height, used mainly for axial compressors mn distance in meridional direction dn =V(dx*-+di), dx/V, =dr/Vp rR distance in the radial direction 5 blade pitch s distance along streamline dy =V (dx? +d? +17d), dx /Ve=dr/V p= rd01V y ' blade thickness t tip clearance x distance in axial direction y distance in the pitchwise direction z distance normal to x and y ° solidity c/s Angles Relating to Blading (see Fig. 4.1) € angle between a blade filament and the radial direction in axial view (blade lean) stagger (angle of chord line measured from the axial} direction) camber angle in the circumferential direction blade inlet angle (measured from the axialt direction) 4 blade outlet angle (measured from the axialt direction) blade lean in radial compressors rex De Fiow variables Stationary frame of reference ay flow inlet angle (measured from the axialf direction) a __ flow outlet angle (measured from the axial} direction) V; inlet flow velocity ¥;, outlet flow velocity Rotating trame of reference Br flow inlet angle (measured from the axial{ direction) Ba flow outlet angle (measured from the axialt direction) jon rather ed a 8 iow 4+ For radial and mixed flow machines, angles are measured from the meridional dl than the axial ditection. For axial machines when the meridional streamlines are ‘substantial angle to the axial direction, the angles are also sometimes referred «© theW, inlet flow velocity W outlet flow velocity Subscripted velocities Vp, tangential component of velocity into blade row Vey radial component of velocity into blade row Vy; axial component of velocity into blade row likewise for other velocities, V2, W etc Voy meridional component velocity, /,,=W(V? + V2) Special angles : i incidence (angle between inlet flow direction and blade inlet direction, (=a, or #=8,—x, for stator or rotor respectively) 4 angle of attack (angle betwoon inlet flow direction and the chord | ine, A=, —£ or A=f,-4) y é deviation (angle between outlet flow angle and blade outlet i angle, =a) —x2 or =8)—x2) ‘é inclination of meridional streamline to axial direction 7 inclination of meridional streamline to axial direction (used for radial machines) | General variables | 4 streamtube cross-sectional area : a velocity of sound a* velocity of sound at condition when flow sonic (similarly p*, p* | ete.) i AVDR axial velocity—density ratio ppVia/a,Vy } b streamtube depth measured normal to two-dimensional surface ' B blockage, 1-(mass flow + mass flow across same section in ideal flow) c velocity of sound (Chapter 10) Cp dissipation coefficient or integral : & skin friction coefficient, 7,/(} 0U*) & specific heat capacity at constant pressure ' & static pressure rise coefficient, (p—p)!por—P 1) : DF Lieblein’s diffusion factor | F flow function, m(¢,To)'7/Apo fh specific enthalpy Q hg specific stagnation enthalpy, #-+¥4/2 A r specific rothalpy, 4 + W/2—U7/2 3 k acoustic wavenumber i m mass flow rate M ‘Mach numbergroggy s BeOR™ ze gvrR ere ee ° peeses number of blades angular velocity rev/min statie pressure stagnation pressure, sometimes (ermed total pressure volume flow rate gas constant degree of reaction specific entropy static temperature stagnation temperature, sometimes termed total temperature blade speed boundary layer thickness boundary layer displacement thickness force deficit thickness fficiency boundary layer momentum thickness ratio of specific heat capacities ¢,/¢y switl parameter Vo/Ve, used for radial compressors acoustic wavelength dynamic viscosity Kinematic viscosity, »=n/p density slip factor, (absolute whirl velocity + ideal absolute whirl velocity) shear stress flow coeflicient; V,/U for axial, different definitions for radial ‘compressors velocity potential stream function loading, Aitg/U* loss coefficient, Apol(Po1—P1) angular velocity vorticity1 Useful basic ideas 1.1. Introduction It is assumed in this book that the reader is familiar with the concepts and methods of fluid mechanics and engineering thermodynamics and no attempt will be made here to survey these. This chapter will look at topics which are of particular relevance to this book, either because they will be used or because they are basic ideas which are useful and have not been very satisfactorily treated elsewhere. Some of the sections are very basic and may serve some readers merely to establish the notation, others are much more challenging and presuppose a fairly thorough understanding of fluid mechanics and turbomachi 1.2. Blades and flow ‘The essential clements of turbomachines are the blades or vanes because it is these which impart the force and, more relevantly, the moment to the flow. Sometimes it is more convenient to think of the blades turning the flow and this is particularly the case with axial machines where the flow is very often at nearly constant radius. On other occasions itis the blade force per unit area (Le. the pressure difference) which is more helpful. Blades in axial compressors have some features in common with aircraft ‘wings but the situation is more complicated. Thus attempis to take over suc- cessful concepts from aeronautics ate by and large not very useful. "varies very strongly along the span, both because of endwall boundary layer effects and because the radius and the blade speed change markedly. Also a compressor blade is just one clement surrounded by many other blades and blade rows; the whole flow is the result of all the blade rows and any one blade can itself have little effect. The design of blades or blade rows in multistage applications should be thought of as choosing a configuration which is compatible with a desired flow in which it is immersed. Fortunately the blades are, like wings, able to tolerate a range of inlet flow angles (i.e. a range af incidence). In addition to a first approximation the outlet flow angle remains constant because of the constraint of the bladed assembly 12. Useful basic ideas ‘Although the outlet flow angle is nearly constant it is not the same as the ‘ion of the blades themselves. For an axial turbomachine the outlet flow direct ifference between the outlet flow angle of and the blade outlet angle x2 #8 Teferred to as the deviation defined by B=cr— xa- In the case of the radial impeller it is normal to define «slip factor @ by the ratio of the measured average wrest tangential velocity out of the impeller to the absolute tangential velocity if the outlet flow were uniform and in the direction of the blades at outlet. Both the deviation and slip factor are predominan'y inviscid effects to which the boundary layer fluid makes only a small adcitional contribution, @ point taken further in Chapters 4 and 6, In the idealized case with very thin boundary layers the streamline leaving the tailing edge ‘would do 0 in the blade direc- tion, a property associated with the KuttaJoukowsky condition that the pressure difference should go to 2er0 there, Ou! actos the passage the flow rnclined to this direction, the sense ofthe inclination being thet which reduces encore on the blades. Methods for estimating deviation and slip will be vs, but itis worthwhile in this introduction giving broad discussed in later chapt etimates: for a radial impeller with a typical number of vanes the slip factor is approximately equal to 0.9; for axial blades the deviation is given approx- imately by 6=0.36V/(s/c) where 0 is the camber, 3 is the blade pitch and ¢ athe chord, The expression for deviation allows some simple generalization for axial blading. The outlet flow direction is given by on 2B - O+8 where £ is the blade stagger, the inclination to the axial direction! of the chord, the line joining the leading and tailing odges. (This is valid only for circular ac camber Lines but this is a very common chojee.) Introducing the estimate for the deviation for a solidity c/s ~ L.0 gives on=t — 0/2 + 0.36 so that a 0.26 In other words the outlet flow direction depends to only & fairly small extent ton the camber whereas itis the stagger angle £ which really thas a big effect. cre Jow intet Mach murabers most axial blades are able to tolerate quite a large jncidence range, so again the camber is of secondary importanes. i even high subsonic ones, f ‘no longer so true. It is nevertheless fe ath numbers b h Mach numt ‘chinery blade passages is extraordinarily com- reason for this is the unsteadiness of the flow. Tt is essen- oment plicated and a major vial that there should be stationary and rotating components and movBlades and flow 3 By ay Ms uy “ —-> t~ @ Fig. 1.1 The velocity triangles for flow catering and leaving a moving component, im this example an impeller with backsweep x2 of one past the other inevitably creates unsteadiness. For mast purposes it is possible to ignore the unsteadiness by working in a frame of reference fixed to the component under consideration: for stator blades @ coordinate system is used which is stationary (sometimes called the absolute frame) and for rotor blades or centrifugal impellers the frame of reference moves at the local blade speed (this is often referred to as the relative frame), As a convenience in changing the frame of reference it is usual to describe the flow with vector triangles and Fig. 1. 1 illustrates this for a backswept cen- trifagal impeller in which the flow at inlet has components in the axial and tangential directions only, while at outlet it has components only in the radial and tangential directions. Throughout this book the velocities in the stationary frame will be denoted by V and in the relative or moving frame by 1. The corresponding flow angles will be denoted at inlet by a and 8 for the absolute and relative flows respectively. wy foe ra w, Bap We, Ve a Fig. 1.2 ‘The velocity triangles into and out of an axial rotor row4. Useful basic ideas With axial machines, whenever the inlet and! outlet radii are equal, iis ane common to overlay the inlet and outlet triangles. This is illustrated in Fig, To One of the attractions of overlaying the triangles is that the and the arial velocity is not equal upstream and downstream of the blades, tout sirm s y peasant that chi is often assumed singe in many eases this assupion sl) allows cone! Le Oe tents it is gencrally tue that the vpalation in the axial velocity about 2 mnean is kept small for good aerodynamic reasons for most blade rows in a compressor 4.3 Work input into compressors d integrated. In ceality this is an impossible take and instead it is usual to con- sider the ‘The advan tage of this is that ise nat ses can be ignored and the results arc valid even for flow with large irreversibilities. <. Fig. 1.3 An idealized rotor. Flow enters at racius 7; and leaves ar Consider the sketch of a eross-seetion through a rather unusual but gener’ imixed-flow blade cow shown in Fig. 1.3. The flow enters at radius 71 with tangential (or whiel) velocity Vp, and leaves at ra with tangential velocity Vea ‘The moment of momentum about the compressor axis of fluid ene i re ‘and rpVm leaving. The torque therefore required to produce this change for a flow rate, nt kgis is ay m(r2¥en — 11 Vou) and the work input per unit mass flow is W = elt, Vor — 11 Yo) a2) ‘This, of its derivatives below, is known as the Buler ‘equation Jor turboWork input into compressors machinery, Sometimes it is itten in terms of the local blade speed We UVa — UVa a3) and for the special case of axial machines where the flow enters and leaves at the same radius W = U(Veq, — ¥q) = UAV). (1.4) Itis to be remembered that the Euler equation is valid no matter how the moment of momentum is produced, and even viscous drag on the blades can produce a positive work input. Because the tangential stresses at the casing and the hub are usually very small the total work of the blade row can usually be i ferred from measurements of the velocity components upstream and downstream. In special cases, for example when there is casing treatment to delay stall, there may be large effective stresses at the walls and an expression such as equation 1.4 may give a misleading estimate for the blade work. The equation for the torque T=mrVin ~ Vn) can be applied to a small region of the machine, not necessarily across an entire blade row. The torque nay then be replaced by the local blade force and radius rE meV) ) Writing m=pVA, where A is the streamtube cross-sectional arca, and also writing 80-¥5)=(rV4)/ds 5s, with s along the streamline direction, gives 3 pVAIrY,)/as 63 or fy = Vatr¥ las © = Derv,\/Dt (1.6) in steady flow, where fy is the tangential force per unit mass of fluid. The use of local forces such as f, often referred to as body forces, is very convenient in applying simplified mathematical formulations through the blade rows or through impellers. In axial compressors it is often possible to consider flows in rotor passages by merely adopting a moving frame of reference. For radial or mixed flow machines it is necessary to be a little more careful because of the change in local blade speed with radius. It is convenient to begin with the familar energy equation — O+ W=hn — ho6 Useful basic ideas where htop and fi ate the stagnation enthalpies into and out of the region Note that in this equation the work W is positive because in the Euler equation work input to the fluid was taken as positive. For adiabatic machines @ =0- Introducing the Buler equation and expanding the stagnation enthalpy gives after rearrangement b hy + BQ — Ug = hit V2 — Un an Consider juste eRshand is, expanding [Link] Vip + Via Wiz and then expressing the absolute tangential velocity jin torms of that in the moving frame of relerence Vop-= Wan + U2. After some manipulation 10 the Left-hand side of the equation one obtains hy + V2 — UM = hy + Whiz — U32 “Thus equation 1.7 ean be writen + hy + WY — UBD = hy + WY — uy2 (1.8) or alternatively (gaden ~ U32 = order ~ U2 as) whore (igs) is the stagnation enthalpy in the moving or relative frame of Veforence, Following the suggestion of C H Wu it is common to detine the quantity known as the rothalpy by > roielpy © Taht Wa U2 so that equation 1.9 becomes simply heh In rotating blade rows rothalpy has properties analogous to stagnation enthalpy in stationary passages. In a moving passage the rothalpy is therefore constant provided: (a) the flow is steady in the rotating frame; @) no work is done on the flovr in the rotating frame (for example by Fri- tion from the casing); (© there is no heat flow to or from the flow. Very clearly when the concept of rothalpy is applied to a stationary Place row the equation reverts to conservation of stagnation enthalpy. he objectiv. i: ging the equationWork input into compressors 7 for the conservation of rothalpy gives dg—hy = 4U3 ~ UR) + KV} — Wh. (1.10) If W, and W, are relative to the blade row under scrutiny the equality is valid whether the blade row is moving or stationary, If U)=Up, as it might in an axial machine, then it follows that 0s is usual for any stationary passage or diffuser, Note that the corresponding expression for the static pressure rise Pap) = 4o(WE — WB) (1.12) is only valid for loss-free incompressible flow. If there are losses in incom- pressible flow the static pressure rise can be written P2-P1 = WE — W2 ~ APross- (1.13) For compressible flow more care is necessaty and it is convenient to begin with the thermodynamic relation dh=dp/p-+Tds so that on integrating In-iy= [PP (? ras (1.14) and the losses for adiabatic flow are contained in the entropy rise. The pro- cess can be shown graphically; Fig. 1.4 is for an axial blade row in which (ighea is constant. i “at pp, The losses, by which is meant the tise in entropy, are produced by Processes associated with the flow, mainly shear W ted vis iSSipation) and mixing oF the to Entnelpy| (roket h,| | wre Entropy $ Fig. 1.4 ‘The enthalpy diagram for an axial rotor with equal blade speed at inlet and outlet8 Useful basic ideas ss we fay 2 Lt =1, fy ut vile 2 Fig. 1.5 The steps in the enthalpy change through a compressor, in this case with large radius change in the moving element as, for example, in a eentrifual compression Ghodea = hy + WH and rothalpy f, = hy + W3/2 — U7. Station 3 is alter a stator or diffuser. Consider now the case there is constant rothalpy but a change in the blade speed such that U2> Uj. The corresponding enthalpy—entropy diagram is shown in Fig. 1.5, In this case there is a marked change in the relative stagna- tion enthalpy such that Gorda > Hovdeer With loss-free compression the static pressure would rise to pa, but with loss this is reduced to pz. As before the losses, represented here by the entropy As, arise only from the flow processes. Some of the static enthalpy and pressure rise comes only from the tern 4(U3—U}) and since this is unconnected with the flow processes it does not have loss-making processes associated with it and gives rise to no losses and no entropy increase. The losses tend to increase as the amount of deceleration of the relative flow is increased and also in proportion to the cube of the relative velocity; in other words the loss may be expected to be related fairly directly to (W—W3) and not to the overall change in static enthalpy Ay—hy. The significance of this is that the part of the enthalpy rise attributable to the change in blade spood, 4(U3— U}, is essentially loss free. Furthermore this can be. increased without aerodynamic Jimit, unlike the enthalpy rise produced by decelerating the relative flow where excessive reductions in velocity lead to flow separation. It is these two factors which have favoured the use of ravial compressors: if most of the static enthalpy rise is attributable to the change in blade speed between inlet and outlet the expected pressure rise will be obtained and the efficiency will be reasonably high even if the acrodynamicsWork input into compressors 9 behaviour is poor with large regions of separated flow. In fact the principal limit on the maximum pressure rise from radial compressors is the strength of the material from which the impeller is made Radius changes may have important effects on axial machines as well, For atypical axial compressor stage the static enthalpy A/t might be approximately 0.4 U?, where U is the local blade specd. Suppose that between inlet and outlet to the rotor the distance of the streamline from the compressor axis in- creases by 10 per cent. Then it follows that the quantity 4(U3 — U3) increases by about 10 per cent of U too, In other words a small change in radius can produce ‘free” changes in static enthalpy of the same order of magnitude as those produced by the deflection and deceleration of the flow in the blades. The changes arc ‘free’ because they are without losses and do not contribute to the tendency of the boundary layer fluid to separate. This often has led to a pronounced effect at the bub of axial compressors; Figure 1.6a shows a com- pressor with a rising hub line where advantage is being taken of the change in radius, whilst Fig. 1.66 shows a constant hub radius and a falling casing where advantage is not being taken of this effect and difficulties of overloading near the hub are more likely to be encountered. In examining the flow in radial machines, or axial machines in which the distance of streamlines from the com- pressor axis changes significantly, it is proper to take account of the conserva~ tion of rothalpy by considering quantities such as T—U7/2c), sometimes called the reduced temperature (and the pressure related to it) rather than the normal static temperature and pressure. To- re -> C5 fa) Fig. 1.6 Meridional sections through «wo hypothetical compressor stages: (a) has a rising hub, (b) a falling ot contracting casing There is a different way of looking at work input which is usually associated with Dean (1959). The one-dimensional equation of motion along an instan- taneous streamtube can be written av , pov _ 1 ap ar Sas as a and the stagnation enthalpy Aho ak 1 ay? as "as 2 as G18)10 Useful basic ideas Now introducing the thermodynamic relation Tds = dh—~dp/p and restrict ing consideration to reversible adiabatic processes, Tds = 0, gives aig 1 op, Lave 3: p ds | 2 as ap Combining equations 1.15 and 1.17 shows immediately that for reversible adiabatic flow dho _ _ AV as ar G18) ‘Thus, with the restriction to isentropic flow, the rate of change of stagna- tion enthalpy in the flow direction is equal to minus the rate of change of velocity with respect to time at that point. Tf the flow were steady, 9¥/0/=0, there could be no work input. Equation 1.18 may be modified by multiplying by V and adding dhg/31 to cach side so that Do _ aho { av. \ Dho _ dho 4 py} -— : Dt or * at (1.19) ‘The term dfig/9/ may be replaced by a process analogous to equation 1.16 to give Dho Dr (1.20) In words this states that the rate of change of stagnation enthalpy of a particle along a pathline is equal to the partial derivative of static pressure with respect to time at the instantaneous position of the particle. Since it is the partial derivative with respect (o time it is apparent that stagnation enthalpy change js impossible without unsteadiness. For a stationary observer in a turbomachine the pressure distribution from rotor is unsteady and the function of time may be idealized by a linear sawtooth ' distribution. As the number of blades in increased the amplitude of the sawtooth can be reduced for the ideal blade row but the slope 47/81 remains constant i so that the rise in stagnation enthalpy of fluid passing through the machine remains the same. ‘The unsteady view of work input has found litfle practical appitcutivm although it dispels any illusion that conditions inside a turbomachine ean be rendered steady, The convenience of the Euler equation for turbomachines lies with I its simplicity and with the fact chat by changing the frame of reference for rotors and stators it is possible to carty out most calculations by treating the flow as steady. The real complexity of unsteady flow is beyond the present capabilities of calculation methods.Dynamic scaling 1.4 Dynamic scaling This section would often be called “dimensional analysis’ but the present title has been chosen because it is more appropriate. The cancelling of dimensions is rarely used (o analyse real problems and in addition the scaling groups which are finally used are often left with dimensions. So long as the correet scaling is taking place it does not matter whether the groups have dimensions or not even though there are obvious advantages of generality in having groups with the same value in any system of units. Geometric scaling ‘The simplest scalings to consider are the simple geometric groups, most of which are obvious such as the solidity (the inverse of pitch—chord ratio), aspect ratio (the ratio ofmean blade length to mean blade chord), hub—cesing ratio (the ratio of the diameter of the hub to that of the casing; often this is referred to-as the hub—tip ratio but tip is ambiguous because the stator tip may be on the hub). The geometric scaling could also be taken to include angles but these are usually viewed differently Whilst the specification of the geometric ratios is simple it is not always clear what is the most appropriate choice. For example the tip clearance is known frequently to have a crucial effect on the blade performance but the ‘most appropriate ratio in which to express it is not at all evident, In design, particularly mechanical design, it is often convenient to scale tip clearance with the blade height but this does not take proper account of the relevant fluid mechanics for long blades (i.e. high aspect ratio blades) because it is only con- ditions near to the blade tip which will be dircetly significant. More suitable lengths with which to scale tip clearance, all of which are used, are the blade chord, the pitch, the staggered gap (essentially the distance between adjacent blades at the trailing edge measured normal to the flow direction) and the maxi- mur thickness of the blade at the tip. All the possibilities have some merit and the choice depends on the process or model being used. It is here that the judgement of the acrodynamicist must be exercised. With the exception of the last of the lengths all the other three are very similar in magnitude and it may be difficult to identify which is most appropriate from the measured trend, Overall aerodynamic sealing Very often the choice of aerod . Sometimes instead of mass flow it will be volume flow which is required and as an alternative to efficiency, the power input could be used. ‘The pressure ratio is non-dimensional but it neads to be decided what pressures are most suitable for the application or investigation. Most compressors are ML42. Useful basic ideas \ able to use the inlet dynamic pressure, so the inlet stagnation pressure Por is ‘a suitable denominator, It is not always the case that the downstream device ( can wllize the dynamic pressure at outlet and if this is the ease itis then more | Sensible to use the downstream static pressure pz in the numerator of the ¢ pressure ratio. When the gas leaving one compressor enters another or when the air is to be used directly for propulsion (as for the bypass stream of a jet engitc) then the outlet stagation pressure po» is appropriate, Clearly palPor « (known a ure ratio) is less than pox/Por (known as th total-to-total pressure ratio); in an idealized case with uniform outlet flow the amount can easily be quantified in terms of outlet Mach number to the highest value of pz relative to ffer is a measure of how good is the Uniform outlet flow also corresponds pop and the amount by which these dif outlet flow from a compressor. “he pressure ratio (or pressure rise) and the mass flow through a compressor 6 osceretonaic enmpreenssoee ‘The compressor speed is most naturally con- verted into a Mach number based on the blade tip speed U, i.e. UN GRD), (ml which is an unknown quantity sine q C is n 5 Vy RT). For compressors passing only one type of gas 7: | « that the ratio UN To is often used. If only one machine is of concern the tip é radius is fixed and then NIN Ty suffices, where NV is the rotational speed iv rev/min, radian/s, Herz etc. A convenient method for expressing the speed i is to use the corrected variable so that ‘ ¢ Neon = —Ne- where here @ = os vo ret r and Tage i8 a reference stagnation ‘ appa ‘usually the sea-level temperature for the standard atmosphere, | ( he corrected speed has the advantage that with the value of @ usually hot very far from unity the corrected speed is fairly similar in magnitucle to i C | the aotual speed and as a variable its Tess abstract than, for example, NIV Ty. | In considering the operstion : co itis apparent thatit isthe volume jg), thee tibea reaeaatarm sme large in relation to the speod ‘of sound then the density in the inlet flow is vel of incident sno longer constant. A convenicit non-dimensional group can therefore be formed as the ratio of the acttal iiss flow rate m to the choking mass flow rate m/imenore Where Mm chore is the 1rs me CESS NA nt EON ne Dynamic scaling 13 flow of a uniform sonic velocity flow through the same cross-sectional area A. For the sonic flow the density must be evaluated at the sonic conditions (with properties denoted as p*, p*, PM and a’) and the choking mass flow rate becomes ‘Mtcyore = A pt at = A(PYRT*)V(y RT) = AfptvT* (yk) Now it is easily shown that for the flow of a perfect gas = Toy DAI and for isentropic flow Dilpo=(PITyH-” so that p* and T* are both proportional to their corresponding. stagi properties. It is then easy to show that the group mN (RT 9) Po is a satisfactory non-dimensional combination of parameters being proportional to m/ngyxe: The name flow function is sometimes given to the group or those related to it, In some work R is replaced by the specific heat c, with no par- ticular advantage. Clearly if the same gas is being used the same effective scaling is produced with R (or c,) dropped so that the group has dimensions and this is common industrial practice. Furthermore if a single machine is being considered then the area term A is redundant, What is then left is frequently reframed as the corrected mass flow rate, which can be written im 8 shore here 022 and §=-P0- Meo 8 ref Poret The reference stagnation pressure is usually taken as sea-level static, 101 kN/m?, Again the use of corrected mass flow has the advantage that the numerical magnitude, and with it the physical significance this has, is not obscured as it might be with uly non-dimensional scaling. The use of a group to relate mass flow to the choking mass flow of a uniform flow along a streamtube of the same cross-sectional area has wice application in compressible flow. In the present work, for example in Chapters 4 and 5, it will be used in considering axial blade-to-blade flows. There it will be used14 Useful basic ideas in the form mv (cpTo) Apo denoted nd referred to as the flow function. An equally valid and perhaps neater formulation compares the streamtube area 4 to the streamtube area for uniform sonic flow with the same mass flow rate; this ratio is usually writ- ten A/4* and can be easily related to F, Sometimes the interest is directed! to the conditions downstream of a compressor, for example when a second ‘compressor is to be connected, and then the flow function is expressed in terms of the outlet variables pgp anc Typ. If the dynamic pressure at exit is to be lost it may be more appropriate to use the static outlet pressure pp. For a compressor with no variable geometry (or variables held fixed or to ‘a schedule itself fixed 1o corrected mass flow and corrected speed) for a fixed gas and where the dependence on Reynolds number can be neglected, the overall functional dependence can be written Po asf NO Por po’ VT ‘The same functional dependence would exist if pgp were replaced by the static pressure p2 or if the compressor efficiency 9 were being sought JaP mar Ma Aerodynamic scaling ce, Ofien different machines will be compared on the be ese variables and then itis important to bear in mind the geometric difference between the machine —- again knowledge and insight are called for in recognizing when it is appropriate to compare dissimilar machines. When it comes to finding appropriate scaling to assess the aerodynamiés a rather dif ferent formulation is appropriate; for example pressure ratio gives no measure of the severity of blade loading and the flow function or corrected mass flow does not relate directly to the incidence on the blades. Stage and blade loading The enthalpy are of the rotational sp | BAI? where radins, the latter being guite common for axial machines. The e can be the static or stagnation enthalpy, depending on the context, hough stagnation is more common, For compressors in which the pressure rise is small compared (o the absolute pressure, such as low-speed machines, the den- sity may be reasonably approximated as cot Wenient to define the pyosoltefrterovetticien:4 | npanoaravcos Dynamic scaling “denominator. As before the use to which the group is put will determine whether static or stagnation quatitics should be used. It is a common convention to define though sometimes y is used for Ap/pU?: it follows of course that for incompressible operation Ap/p U?=nAh/U*, where 7 is the effi- ciency, and both non-dimensional groups are very similar in magnitude and often in their trends. The ratios Ah/U? or Ap/pU? provide a measure of the actual input to the potential work available, i.e, to U. Clearly the demand don the stage is more taxing if a large enthalpy input is required from 2 low blade speed machine than a high-speed one and the magnitude of y gives a ‘measure of this. For radial machines it is quite common to define the loading by Y= AplpN’D3) and sometimes to replace Ap/p by the isentropic (sometimes called the polytropic) head rise Hy, = (dp/p Sometimes pressure ratio is used to present low-speed compressor results and values like 1.005 are quoted. This is not a helpful presentation because it shows the pressure rise to be orders of magnitude below the ambient pressure and gives no indication of the severity of loading of the blades or stages. If Ap denotes the pressure rise which is small in relation to the inlet pressure py then it is easy to show that Po. Pit Ap _ yy VAP yy a PL Pi Opved-a alu? p, UF Ap =ltyM ay pU where second order terms in Ap have been neglected. M, denotes the Mach number based on the blade speed, For low-speed machines, when M, is much less than unity, it is better to use the pressure rise coefficient Ap/p U® than the pressure ratio since the latter is essentially providing less information. For values of M, significantly less than one the non-dimensional performance of the machine is essentially indopendent of My. However when M, becomes large the pressure ratio is a very useful parameter because the rise in pressure 15 is comparable to the absolute inlet pressure. Furthermore, since there is an - upper limit to the possible tip speed, the experienced engineer gets an implicit assessment of the loading from the value of pressure ratio. Flow coefficient The work input to a stage depends on the flow through it and for an axial stage it is easy to show that y is related to V,/U, the ratio of axial velocity to the. blade speed often called the flow coefficient and denoted by #. Similarly the16 Useful basic ideas flow coefficient determines the incidence into the first rotor and then in (urn into the blade row downstream. For a particular stage the incidence is one of the crucial flow quantities in determining the performance of a blade row. ‘The flow coefficient therefore effectively determines the performance of the stage. The axial velocity might be the local value or the mean value across the annulus, depending on the use to which it is being put. As @ is reduced the incidence rises, bringing changes in the blade operation, Most axial com- , with the lower end of this range being more common nowadays. For radial machines a somewhat different convention for the flow coeffi- cient is sometimes used. The symbol @ is still used but is defined differently sometimes in terms of the impeller outlet speed U2, mass flow rate m and a density, typically the inlet density p,; 6=a/(p,U;D3). The flow coefficients defined in this way have rather low values, typically of order 0.01—0.1. Another not dissimilar expression which is sometimes used is 6=m!(pND}. 2.2 woot 20 ween M8 18 seen 45K Pressure ratio veh Brera taf ecco gk rab 980 eee Bk 11 —L 1 = on 0.3 05) OF 0.3) 0.5 IM ratececes (MIM) Fig, 1.7 The measured performance of a turbocharger compressor presented in (wo different ways; the symbols on the right are for the same rotational speeds as those on the left, (From Fink, 1988) Figure 1.7 shows measurements of a turbocharger compressor made by Fink (1988) and presented herc in two different ways. The left-hand plot is the overall performance that might be passed to a hypothetical customer and it shows the pressure ratio plotied versus corrected mass flow rate for a range of rotational speeds (the corrected mass flow rate is actually non-dimensionalized by a reference value). The right-hand diagram shows the same data plotted with the acrodynamic scaling: the ordinate is the compressible version of Ap /pti? and the abcissa is the inlet flow cocfficient V,/U,. Expressed in these aerodynamic variables it can be seen that the impeller performance is almost independent of the rotational speed; Mach number is not an important variableDynamic scaling in the range at which this machine was operated. (It should be noted that the constant speed lines in the left-hand diagram are short at high speeds only because the facility had insufficient power to drive the compressor at higher mass flow rates.) ‘The flow coefficient V_/U can be related to the oves inn (CpTo)!Apg introduced above. The nd the density ‘en by p=p/RT. a given flow rate the static and stagnation variables are related by functions only of flow Mach number, so the mean axial velocity is given by ¥,=kmRTo/Apy where k is a constant, A compressor operated at constant cor rected speed, i.c. N/VTy = constant, is actually one for which the Mach number based on blade speed is constant, so the mean blade speed is given by U=«V/Ty where « is another constant. The flow coefficient V,/U can then be written Il flow function Vx _ Kola) RCP) mV pT) _ U «Ty AD Reynolds number Real compressors suffer losses attributable to shear siresses and mixing and the amount of this is strongly affected by the extent to which the flow is laminar or turbulent. If the flow remains laminar the shear stresses are low and the loss directly attributable will be correspondingly small, but on the other hand there will probably be extensive regions of separated flow leading to a lower pressure rise and a lower efficiency than that which might be achieved with amore turbulent flow. The transition to turbulence is ultimately an indication that the inertial stresses have overwhelmed the viscous dissipation and the non- dimensional group which describes this is the Reynolds number, X= Vi/y where. V is a representative velocity and J a representative length. The values of Roynolds numbers used in turbomachines are usually large, often 10° or more based on characteristic lengths of the machine, Normally the correct non- dimensional group has a magnitude near to one if it is relevant to the flow because it represents the ratio of two related or competing effects; a good ex- ample of this is Mach number which is relevant only when it is of order unity. In (he case of Reynolds number very large values are relevant because macroscopic quantities such as the dimensions of thé compressor ot of the blade are used whereas the process which is being assessed is a comparatively microscopic one taking place with dimensions comparable to or smaller than the boundary layer. This use of the ‘wrong’ scale is possible because the 718 Useful basic ideas c Tn a compressor it is not obvious what are the relevant velocity, Length and c viscosity. If machines of « particular geometry are being compared the sense wf the Reynolds number can be retained with any relevant specification: ‘ SyDty would be quite suitable with N being rotations) speed, D the inlet or Carlet diameter and the viscosity ata reference condition ‘such as inlet stagma- tion, This could be added to corrected mass flow and corrected speed in the i specification of pressure ratio given above, TRE choice becomes more difficult ( ven machines are not identical and the comparison requires the salient al effect at low Reynolds variables to be chosen. For axial compressors a crite ‘ tumbers is separation of the flow on the blades, & topic considered further in Chapter 4. For the axial machines the natural ehoice therefore based on vad chord and the inlet velocity relative tothe blade, In most eA the changes ‘ across an axial blade row are small enough that the viscosity could be defined anywhere upstream or downstream but by convention upstream conditions are ‘ eed as the reference, Decisions need to be made about which blade row to ‘ tise and at whieh cadial station in the evaluation of Reynolds number. Figure ‘ 1 8 shows normalized stage pressure rise coefficient at stall for multistage axial compressors as a function of blade chord Reynolds number, taken from the « paper by Koch (1981). ‘ TT Te co a 10 ‘ Pressure rise é coetficient at stall normalized by value at Re =1.3-10° os re ae 01 02 04 7 2 4 xt0° 10 : ‘Reynolds number Volv Fig, 1.8 The variation in static pressure rise at stall for axial stages with Reynolds number. (From Koch, 1981) “Phe choice ofthe appropriate Reynolds number for radial machines requires snore to be an idea of where the most serious viseous Toss source is likely be. This is believed to be at or near the impeller outlet, largely because the passage is narrowest here, see Casey (1985). An appropriate choice is therefore Tigbalv, where Ua is the impeller tip speed and Bp is the impeller tip width. areait, With the viscosity set at the inlet value this RrouP is now the basis tifa recommendation reported by Strub et al, (1987) of an international roup ‘who have devised correlation procedures for assessing Reynolds munis effectsDynamic scaling in radial compressors. Rather than cocretate efficiency it is the non-dimensional loss, =m, which is used, The dependence of loss on Reynolds number is assumed to follow the same form as the flow in roughened pipes and the predic. tion of the effect of Reynolds number change on performance needs the effec- tive surface roughness to be known as a fraction of the impeller outlet width bo. “Phe effect of Reynolds number on the performanee of axial compressors was discussed by Wassell (1968) with the Reynolds number based on the blade chord and the inlet velocity to the blade. ‘The effect on efficiency, flow rate and maximum pressure rise was considered. Loss rather than efficiency was presented in the form Lay = eR with 1 varying significantly from compressor to compressor. A rapid rise in Joss with decrease in Reynolds number occurred below 0.3- 10° for com- pressors, compared with about 1-10° in the case of cascade tests. Schaffer (1980) found Wassells correlation for efficiency to work well for hydrodynamically smooth blades but not adequately for rough blades at high Reynolds numbers. The loss decreased quite rapidly in the turbulent region, proportional to R* with n in the range 0.10 (0 0.13, until a value of Reynolds number of about 5-10° was reached at which loss remained more or less independent of Reynolds number, This is the condition at which roughness comes to dominate, The Reynolds number at which roughness becomes dominant is a function of the blade pressure distribution and, of course, the roughness height. The behaviour with roughness can be described in a manner analogous to the flow in rough pipes and a critical Reynolds number for roughness height was found to be about 90. In the discussion on Schaffler’s paper the evidence on roughness was supported by Koch and Smith of the General Electric Company. For radial machines it is common to define a group known confusingly as specific speed. It is the ratio of two non-dimensional groups, a loading denoted conventionally by y and given by Ap/(pN*D3) and a flow coefficient =ml(pND}). They are combined in such a way as to remove the dimension of size, in this case the impeller outlet diameter Dz. What results is the specific speed ol? lp My pe apse and a plot of the efficiency against specific speed is shown in Fig, 1.9, taken from Rodgers (1980) for a range of different impellers of the type with axial inducers. In this particular figure the average density has been used to calculate 1920 Useful basic ideas i OT jo Novbact jeep © mrt 5 © 25-50° backsweep owe 2 2 55° backsw BPR Le 2 xf 55° backsweep PR ee § 80} oe & wes 8 5 aie 3 70} é ~ a de a4 02 a4 06 08 10 12 Specific speed based on average density N= } 047 fre Fig. 1.9 The variation in plytropic efficiency with specific speed. The impelers shown ie reproduced by permission of Sundstrand Turbomach and were use in the test Pro- gram (WV, ~ 0.47, 0.74 and 1,2), (From Rodgers, 1980) the specific speed. Very often inconsistent units are used 10 calculate specific speed so that a dimensional quantity results but the form which is given here i: ithe true non-dimensional one. (Tae true ton-dimensional specific speed is a equal to the comion value in Imperial units divided by 129.) Peak efficiency a scours in Fig. 1,9 at a value of N, equal to about 0.7, ‘his figure alse shows 4 thee of the impellers used in deriving this data with specific speeds, from Toft to right, of 0.47, 0.74 and 1.2. The sitnilarity in the overall geometry should be noted because for the specific speed fo provide useful guidance it is necessary that it be used on a single family of designs and it cannot be relied I upon to give the correct trend for machines with radially different geometrical layout,Losses 21 Now the pressure rise depends primarily on the outlet diameter and rota- tional speed, The mass flow rate depends, however, primarily on the inlet diameter and the rotational speed. The specific speed then contains or reflects the ratio of the inlet diameter to the outlet diameter and the magnitude of V, for a particular family of designs is determined a ir of these diameters, 88 € outlet diameter, :. For the impellers with the highest ney at around N,=0.7 in Fig. 1.9 the inducer tip diameter is between about 75 per cent and 80 per cent of the impeller outlet diameter. (More recent designs suggest that still higher efficiency can be obtained at rather lower ific speed with a diameter ratio of about 0.6, sce Chapter 6.) In some organizations it is recognized that specific speed gives geometric information which can be handled more easily and transparently by using simply ratios such as the ratio of inlet to outlet diameter. Given the additional effects at high Mach number which are omitied from the derivation of N, this would seem to be a very satisfactory decision, 1.5 Losses It is often assumed that everyone kno often reveals that ther on the end result, a ar lence of entropy tually the eq ee and fall in eee press for adiabatic flow, but this is usually an acceptable assumption for compressors.) In considering loss it is usually most helpful to consider the mass weighted average of the stagnation pressure, s how the losses arise but discussion this. Agreement exists = Spon Way Jo vay At the micrascapic level , ise ir uit to the aerodynamicist working a . A pos- sible'list of loss sources, which other people might draw up rather differently, is: 1. drag at solid surfaces 2. mixing; 3. shock losses; 4, shear work Rather than try to unify the loss generation it is probably morc helpful to keep the different categorics as soparate entities and to illustrate and amplify them.22. Useful basic ideas 1 Dearie, ‘The effect of drag y be seen by considering incompressible flow through a cascade of uncambered and unstaggered blades, Fig. 1.10. Upstream of the cascade conditions are uniform so that the static pressure and velocity can be denoted by py and V7). At ¥2 is aot uniform, although to a very good approxiny 88 p> is. Well downstream the conditions will revert to uniform with velocity V; and pressure p3. Because the flow is incompressible it follows immedintely that Vj=V5 and also Vis = V35 = (o voay = Vols —5) where 5 = {ply is the total displacement thickness on both sides of each passage. Applying, conservation of momentum between planes | and 3 leads to the very simple result, since Vy=V5, that FIS=Pi-Pa where F is the force on each blade. Because V\= V5 it also follows that Fis=poi- Pos the loss in stagnation pressure. As the flow is uniform at stations 1 and 3 it i clear that the stagnation pressures are also equal to the mass weighted values. In this simple case then the overall loss can be oblained entitely by knowing, the drag force on the blades and the argument could also be extended to com- pressible flow. ‘The simplicity of the result gives ground for caution and it can be seen that the same result would have been predicted for a grid of rods on cach of which the drag force F would have been produced by the pressure distribution consequent upon separation of the flow around the point of maxi ‘mum thickness. In other words this simple analysis has provided no information about the mechanism of loss but has merely given an expression for evaluating, it, Fig. 1.10 Loss produetion by a cascade of uneambered unstaggered plates, Condi tions uniform at stations 1 and 3» thickness @ 2s she crucial variable, This was presumably because for an Losses 23 ‘Suppose that the loss were assessed at the trailing edge, In this case the mass weighted stagnation pressure is not equal to the valuc in the middle of the passage and different values would be obtained depending on the shape of the velocity profile at the trailing edge. It is for this reason that it is usual to choose a reference condition for the losses when all the nonuniformities have mixed out. This condition is obtained a long way downstream. However, it is usual only to make measurements sufficiently far downstream that the largest changes have taken place when a reasonable approximation to the mixed out loss can be obtained. This identification of loss with blade force has had seriously misleading effects. First it is only true in rather special cases, Second it caused people to identify loss with momentum changes and the calingTedgeraraeS lated aerofoil the drag force is proportional to the total momentum thickness in the wake. As will be seen below momentum thickness is not the most relevant variable, although by using correlations the true loss can be inferred from the momentum thickness, and more insight can be obtained by using the correct variable which is the so-called energy displacetnent thickness usually denoted by 85. Mixing loss has been referred to above in connection with two-dimensional cascade flows but itis also a familiar process which takes place in many other instances in turbomachinery and elsewhere. The mixing taking place after the sudden enlargement in a pipe is a standard example of a thermodynamically irreversible process; the conservation of mass and momentum allows the rise in static pressure and the fall in stagnation pressure to be calculated. Mixing usually occurs in much more complicated forms in turbomachines and the flow is generally three-dimensional. An example of the pattern downstream of a rotor blade with tip clearance is given in Chapter 8, Figs 8.19 and 8.20, Here a jet is formed by the tip clearance flow which has ultimately to be mixed out to the uniform condition and it chieving this that the major losses are to be expected. The crucial point here is that the high kinetic energy of the jet cannot be properly utilized (at least no one knows how) and is therefore entirely lost. The only sure reference condition is the fully mixed out one. In considering mixing one recognizes that the essential feature is non- uniformity in the flow. Care is needed in distinguishing betwecn those non- uniformities which lead to mixing and those which do not. A loaded blade ‘or an isolated wing produces a nonuniform flow as a necessary part of producing lift. This does not necessarily produce loss because it may be possible to decelerate the high velocity flow by a process which is virtually thermo- dynamically reversible without requiring mixing. This is the way in which one hopes that the flow will behave on a well-designed acrofoil. If, on the other hand, the flow separates and produces a wide wake there is then no mechanism but mixing to return the flow to uniform and the losses will be24 Useful basic ideas high at the completion of the mixing process. Flow separation does not necessarily lead to large mixing loss because the flow may reattach to form a bubble. The increment in Joss attributable to the presence of the bubble is very stnall. Itis known that many blades operate with a separated region near the trail- ing edge and that the corner regions often contain separated flow. One can make estimates for the loss that this ineurs in returning to the uniform condi- tion and the easiest case to consider is a two-dimensional cascade, The model is shown sketched on Fig. 1.[1 where the flow separates in the blades so as to leave at the wailing edge as 2 core or jet with uniform velocity 2 and a wake with zero velocity in it. For simplicity the flow is treated as incompres- ible. The flow enters at ey to the axial direction and for simplicity is assurned to leave the blade row in the axial direction (even after separation). The flow is assumed uniform across the core region with velocity V2 and stagnation pressure pgp and with no flow across the wake. The static pressure at the trail- ing edge plane is taken to be uniform and equal (0 pz across the entire passage, Since it is assumed that there are no losses to the core flow outside the separated region it is clear that at the trailing edge pea po; in the core, ‘Typical measured loss for fully aitached flow TT 4 oof 02 089 04 Wig. 1.11 Calculated loss for fully mixed out flow with idealized wake of widlth No flow in wake. Incompressible flow. Inlet flow angle a, = 35°, outlet flow a a = 0 i SLLosses Because there is no flow at all in the separated region the mass-weighted stagna- tion pressure across the trailing edge plane would also be equal to por. Clearly a loss-making process has occurred in the passage but the usual mass flow weighted book-keeping process does not record iG, it might be better to say that a loss-initiating process has taken place which could hypothetically be undone at least in part by some device which would decelerate the core flow reversibly Some distance downstream the flow mixes out to a uniform condition and here the velocity is V3, the static pressure p3 and stagnation pressure pos. Applying continuity of mass and axial momentum (only axial momentum is needed here because of the simplification to axial outlet flow) between station 2, the trailing edge, and station 3, the station for fully mixed out flow, results in = a 2 Oyined ae = PAB = caster, ( “ ) (21) eV? sow ‘The curve plotted in Fig. 1.11 is for a,=35°, a reasonable value for inlet flow angle with an outlet flow which is axial. Also shown is the level of ‘measured loss in an unstalled cascade with this inlet angle and a solidity of one. Rather surprisingly the wake needs (o extend over about one-cighth of the passage before the mixing loss attributable to it equals that of other regular losses in the cascade. At higher widths of wake the loss rises very rapidly. Very similar conclusions would be reached for other inlet and outlet flow angles and for more general three-dimensional separations: small arcas of wake can mix out with surprisingly small loss. The treatment in this section is predominantly from the point of view of incompressible flow because the ideas and mechanisms ean be understood much more clearly with this simpler flow. In many practical cases the flow Mach numbers will be quite high and it is reasonable to ask how valid is the assump- tion of incompressible flow. Stewart (1955) looked into this from the point of view of flow downstream of a row of axial blades, Calculations were per- formed varying parameters to explore the effects and one of the most reveal- ing is shown here as Fig, 1,12 in which the ratio of the fully mixed out loss Hrsced oa 1.8 rare cs 14}- 4ol___4 yy 0 04 08 12 16 (V/A "eating 0490 Fig. 1.12 The effect of compressibility in the mixing out of non-uniform flow at the trailing edge & = (Po, ~ Po)/(Po, — P\), &* is the velocity of sound when flow is sonie, (From Stewart, 1955) 2526 Useful basic ideas &y is compared with loss just downstream of the blades w, for a range of Mach numbers. The ratio «29/4 rises quite markedly with Mach number and this trend for mixing loss to rise with Mach number is a general one which is not restricted to two-dimensional flow downstream of cascades. ‘These are another of the standard thermodynamic irreversibilities. The largest loss in total pressure occurs when a flow is decelerated by a single normal shock and the deceleration can be more nearly reversible if it takes place in 1a series of weaker oblique shocks. Inlet Mach numbers to compressor rotors, exceeding about 1.4 are unusual and at this condition the Joss in stagnation pressure across a normal shock is only 4.2 per cent of inlet stagnation pressure. Such a normal shock produces a static pressure ratio of 2.1 and this combina- tion therefore represents quite an efficient compression system. One of the aims of the design of supersonic blades is to keep the Mach number from rising too high downstream of the leading edge; in some cases it may get as high as 1.6. The losses rise as (M—1)° so a small increase in M may give a substantial increase in loss which is somewhat redeemed by the correspond- ing rise in static pressure ratio. The most serious aspect of shocks is their tendency to cause boundary layers to separate. A shock incident on a flat plate will cause a turbulent boundary layer to separate if the static pressure ratio exceeds about 1.8 and for a normal shock wave this corresponds to an incoming Mach number of about 1.3. For a laminar boundary layer separation occurs with shocks at very much lower Mach numbers, Furthermore a turbulent boundary layer which has overcome an adverse pressure gradient prior to the shock (and has an elevated value of form parameter H) will separate with much weaker shock waves. If after separating the flow can reattach the loss may not be that great, but if a wide wake is created, for example downstream of a blade row, then the contribu- tion to loss may be very high. 4. Shea aoe is taki wherever there are velocity gradients but itis only in regions where the gradients are very steep that its magnitude is of concern, “These regions are shear layers including boundary layers, wakes and the regions dividing fast and slow regions of separated flow. At the fundamental level all the loss mechanisms can be related to viscous dissipation by shear work but it is not always profitable to think of it in this way. A complete treatment could be given in three dimensions but the salient features can be found from a very simple one-dimensional analysis which suffices for the proscnt purpose, ‘The flow will be assumed to be in a boundary layer as sketched in Fig 1.13. The normal viscous stresses are expected to be very much smaller than the shear stresses 50 the momentum equation in the streamwise direction can be written SIRNALosses 27 WIT Fig. 1.13 The velocity profile anc! shear siresses operating near to a solid surface with x and y parallel and normal to local streamlines respectively av ap yo ae 1.22) ’ Ox ay ax ‘ ‘The balance of energy can be written, including the heat transfer q across the streamlines but not in the direction along them, as yaly 89 4 aG¥) On’ ay ay (1.23) where ftg is the stagnation enthalpy, Pr=pep/k is the Prandtl number and the bulk properties have been taken to be uniform. Equation 1.23 may be simplified to ye ay? [eon } 24 which is the form given by Denton (1986). For a Prandt! number of one and inlet conditions with uniform stagnation enthalpy a solution to equation 1.24 is clearly that ig=constant. In mast gases the Prandtl number is of order one (for air at standard conditions it has a value of 0.73) and it is common experi- ence that the stagnation enthalpy is nearly constant through a boundary layer on an adiabatic wall. This is more generally true in wrbulent flow when the effective Prandtl number is very nearly one for all gases, Constant stagnation enthalpy indicates that thore is heat transfer through the gas of an amount to balance the work input by viscous shear. Using the thermodynamic relation Tds=dh —dp/p with equations 1.22 and 1.24 gives (1.25) the rate at which the entropy of a particle changes as itis carried by the flow)Losses 27 Fig. 1.13 ‘The velocity profile and shear stresses opcrating near to a solid surface with x and y parallel and normal to local streamlines respectively av _ ar _ ap ax ay ax 1.22) ‘The balance of energy can be written, including the heat transfer g across the streamlines but not in the direction along them, as dhy 8g, arV) ae ay +” by a v P(e trae to Ta) * ate ad where fig is the stagnation enthalpy, Pr=pc,/k is the Prandtl number and the bulk properties have been taken to be uniform, Equation 1,23 may be simplified to yao on (1.24) 2. = ke {ko + (Pro which is the form given by Denton (1986). For a Prandtl number of one and inlet conditions with uniform stagnation enthalpy a solution to equation 1.24 is clearly that 4o=constant. In most gases the Prandtl number is of order one (for air at standard conditions it has a valuc of 0.73) and it is common experi cence that the stagnation enthalpy is nearly constant through a boundary layer on an adiabatic wall. This is more generally true in turbulent flow when the effective Prandtl number is vory nearly one for all gases. Constant stagnation enthalpy indicates that there is heat transfer through the gas of an amount ‘© balance the work input by viscous shear. Using the thermodynamic refation Tds=dh —dp/p with equations 1.22 and 1,24 gives as 1 ar ar > ay (1.25) This can be written in terms of the substantive derivative, DID =Va/ax (i.« the rate at which the entropy ofa particle changes as itis carried by the flow)28 Useful basic ideas by muldplying both sides by V to give Ds _ Yar (1.26) Dr pay Although equation 1,26 has been derived for a Prandtl number of one it can be used in a wide range of gases sinee the processes which dominate the flow in compressors are turbulent, The generalization of equation 1,26 to Full three- dimensional flow is straightforward if tedious and leads 0 Ds Vj ar, Ds Katy TS a aan It would be wrong to leave the impression that shear work is invariably a loss and invariably detrimental. If one considers the upper and lower faces + of the streamtube in Fig. 1.13 work is done on the lower surface at the rate of Vr and on the upper surface of Vr-+a(V7)/ay -dy. If the flow is laminar, or if the shear stress can be taken as locally proportional to the velocity gradient, itis clear that the net shear work, the difference in the work in at the bottom and out on the top, is given by iy = 2 (wr or Nay ay ay ay It follows immediately that the net shear work can be expanded into two terms Ve y Pv <_) + ue (1.28) ( ay Way? > ‘The first term on the right-hand side of equation 1.28 is positive definite and is the irreversible component, which translates directly into internal energy and entropy inerease of the gas. ‘The second termn is not a foss but is the ‘uscful” work done by the shear; it is this component which, for example, provides the momentum transfer to Keep the low velocity fluid in a boundary layer moving against an adverse pressure gradient. In considering loss it is easier to address attention to a particular geometry than to talk in more abstract terms. ‘The geometry chosen is the familiar two- dimensional cascade, Fig. 1.14, but the principle to be discussed is much more general. The inlet stagnation pressure is assumed uniform at po) and downstream of the passage the velocity distribution is sketched, vp is the resultant velocity but the direction is assumed to be uniform, The static pressureLosses 29 Fig. L14 ‘The velocity profile downstream of a blade row somata p2 is assumed uniform across the exit plane, Near the middle of the passage the velocity is Vz and the stagnation pressure is equal to the inlet value, Po=Por, Where the capital letters will be used to denote the “freestream’ values for the outlet flow. For simplicity the flow is treated as incompressible, Integrating over the whole flow the loss in total pressure is given by Joven PY2o1~ Peal (1.29) ‘The integration here may be with respect to y over a pitch and advantage may be taken of the fact that there is a core of fluid for which Po, =pq). Using the incompressible relation for stagnation pressure, po=p +4pv?, the integral 1,29 can be rewritten as ff evatV3—viddy . 30) and the integral 1.30 is familiar in boundary layer theory as the energy displace- ment thickness where it is usual to write it as, ao EP OG Jo aap (in this case the same notation as the rest of the section has been retained rather than switching to that more usual for boundary layers.) The energy displace: ‘ment thickness has found extensive use in boundary layer calculation methods, particularly those from Germany of which the best known is by Truckenbrodt (1952), Lieblein and Roudebush (1956) had found the form of equation 1.30 for the loss from a two-dimensional cascade. Such, however, was the status of momentum thickness, 9 [ wd — wy, Gerived from its importance in acronautics where the drag of an aircraft is30 Usetul basic ideas 18 Wake 4.6 form parameter on 14 12 oS 10 0 02 04 06 08 10 Chord lengths downstream Fig. 1.15 The form parameter measured in the wake of blades and aerofoils. (From Lieblein and Roudebush, 1956) proportional to @, that they found it worth recouching the loss in terms of ‘momentum thickness. It is an important part of turbulent boundary layer calcula- tion methods to assume that the various integral thicknesses are in a fixed rela- tion to one another. When the form parameter H=6*/6 (with *={(— v/V)dy being the displacement thickness) tends to onc it is easy to show that the energy displacement thickness 6, tends to 20. Figure 1.15 shows the form parameter H=64/0 measured downstream of cascades and isolated aerofoils from Lieblein and Roudebush (1956) and from this it ean be seen that within one third of a chord downstream of a cascade the form parameter is less than 1.2. At this condition the energy displacement thickness would be in the range 1,96 <6; <2.09. Measirements of loss would normally be made at least this far downstream from a blade in cascade but inside compressors, measurements. ‘often have to be made very close to the trailing edges, Lieblein and Roudebush used a power law profile to represent the wake and were then able to derive the loss in terms of the momentum thickness and form parameter, Because the form parameter was so close to unity, Licblein (1956) was able to adopt the approximation 8=26 when he wrote Chapter VI in the report by the staff of the NACA Lewis Research Center “The aerodynamic design of axial-flow compressors’. He recommended the approximate expression for the mass averaged stagnation pressure loss 2 6, - 22-2 jo] (1.32) © cose | cose as adequate for unstalled cascades, (This also required the restrietion that the. momentum thickness was much less than the pitch, again reasonable for un- stalled cascades.) The loss is here nondimensionalized with respect to the inlet dynamic pressure, Lieblein and Roudebush also considered the mixing loss from the measure~ ment station to the fully mixed out condition downstream where the flow has returned to uniform. In doing this it is necessary to consider the conservation of mass and momentum in the axial and tangential directions so that there. are changes in the downstream static and stagnation pressures and flow direction.Losses ‘The mixing loss is a function of the form parameter and the momentum thicl ness but in going from a typical measuring station to the mixed out condition the loss will not increase by more than about 5 per cent for a value of the form parameter H up to 1.2 and by not more than 12 per cent for A up to 1.4, Ifthe mixing were being considered from the trailing edge then the addi- tional loss in going (0 the mixed out, uniform condition would make a very much larger contribution, even as high as 100 per cent for the idealized case considered in the section above on mixing loss. There is, of course, nothing wrong with expressing the Ioss in terms of the momentum thickness provided the momentum thickness is not thought of as indicating the mechanism by which loss is generated or can be controlled However, the use of momentum thickness as the usual correlating parameter has had the unfortunate effect of directing minds towards the skin friction drag on the blades or hub and casing walls as the origin of the loss ‘The form for the mass-weighted loss given in equation 1.30 leads, as already noted, to the identification of the loss with the energy displacement thickness 6; défined in equation 1.31 which has formed an important part of turbulent boundary layer calculation methods, Boundary layers will be discussed in more detail in a later chapter but it may be noted here that the equation for flow in a two-dimensional boundary layer, with the assumptions which are usually known as the boundary layer assumptions, is (1.33) where # and v are here the local mean velocities in the x and y directions, x being in the direction of the flow outside the boundary layer. If equation 1.33 is integrated with respect to y from the surface y—0 to a distance out from the surface & > 5, where é is the boundary layer thickness, what results is the so-called momentum integral equation do, @ aU 1 “et oa HID yer (1.34) where U is here the free-stream velocity, ¢; = 7y/(Jp U%) is the skin-friction coefficient and 7 the wall shear stress, This equation forms the basis of most integral calculation methods. If, however, equation 1.33 is multiplied through by u bofore integrating with respect to y a different equation is formed Lda 28 u ar Ad (yay = 2 2 or an 0 pu , U ay” = af. T 2(4)ay = 2p (1.35) }o pU? ay\ U 332. Useful basic ideas ‘The intogral on the right-hand side, denoted by Cp, is known as the dissipa- tion coefficient or integral. The integrand, it will be noted, is, apart from the terms to make it non-dimensional, the same quantity derived in the section on shear work. The mainly turbulent processes making up Cp were discussed by Fernholz. (1964), the main contributor being the Reynolds stress multiplied by the mean flow velocity gradient. Fernholz. derived the values of Cp for a number of sets of measurements and Cp tends 10 increase as the form parameter H gets larger. Nevertheless the spread is quite small, most values lying within a factor of wo (by way of comparison, the skin friction coeffi- cient varies from 0 to a finite value). As part of a calculation strategy, simplified correlations were devised for dissipation coefficient Cy as well as for the skin friction coefficient cp. Schlichting (1979) gave an approximation for the dissipation coefficient of the form p + (1.36) ‘The Reynolds numbers R in this correlation is based on the energy displace- ment thickness and the local free-stream velocity. From Figure 22.7 of Schlichting (1979) 6 may be taken as 0,15 for a wide range of pressure gra- dients, rising to 0.17 in strong favourable pressure gradients and down to 0.1 in regions closc to separation. The parameter 8 varies little over the same range and may reasonably be approximated by 0,005, Equations 1.30 and 1.35 are equally valid for laminar as for turbulent flow ‘Truckenbrodt (1952) gave a relation for the dissipation coefficient in laminar flow and this is of the simple form Cy = BY Re (1.37) where here H is the usual form parameter and the Reynolds number here is based on momentum thickness. ‘The first power of Reynolds number is predic- table from similarity conditions. 6 was shown to vary very little with the form. parameter H, which is perhaps surprising. Figure 1.16 seeks to give some more concrete comparison of the dissipa- tion coefficient, and therefore the losses, for laminar and turbulent flow. It hhas been assumed that for the turbulent case a profile exposed to a modest adverse pressure gradient would be appropriate and therefore 8 =0,005 and 5=0.15 have been chosen as representative. A form parameter H=68/0= 1.6 has been taken from which it can be found that 6)/0=1,65 as a means to giving the results in terms of the usual Reynolds number based on momentum. thickness, For the laminar flow a slightly favourable pressure gradient has been assumed, since it is under such conditions that laminar flow is usually found in compressors, and for this 8=0.19 was taken. The corresponding values of Cr, have been plotted against Reynolds number based on momentum. thickness in Fig. 1.16, The turbulent loss is very much greater than the laminar,Losses 33 Dissipation coefficient Co At Laminar 1 2 x03 ve Roynolds number —— Fig. 1.16 Dissipation coofficients for turbulent and laminar flow useful for estimating loss production + an order of magnitude greater at all but the lowest Reynolds numbers. If Cp can be approximated by a constant value, a better approximation for turbulent than laminar flow, equation 1.35 can be rewritten as 35, = 2CpUMdx = 2Cp % UBdx (1.38) in which the left-hand side term will be recognized as the flow-weighted loss in stagnation pressure. In other words the principal factor determining the loss production for a laminar or turbulent boundary layer is the cube of the freestream velocity : The dissipation coefficient for turbulent flow varies only slowly with both Reynolds number and form parameter (and hence pressure gradient). Laminar flow offers huge reductions in loss production, but the laminar flow is unable {o resist substantial adverse pressure gradients and in some instances the reduc- tion in turning by the blades would offset any gains due to reduction in Loss. Very often the pressure surfaces of axial compressor blades have largely laminar boundary layers and it is clear from Fig. 1.16 why the pressure surface loss is usually assumed negligible. There is a benefit to be had by delaying transi- tion to as late as possible on any surface and this strategy is adopted on the suction surface of low-loss blades such as the new so-called supercritical ones ‘The means by which transition can be delayed is to continue accelerating the flow, which in turn leads to peak high freestream velocities. ‘The dependence of the loss production on the cube of the free stream velocity gives a strong incentive to keeping the peak velocity as low as possible and decelerating as rapidly as possible once the flow is turbulent. Since the loss production is an integral along the surface in the flow direction it also follows that for low loss one wants as few surfaces as possible, giving an incentive to have the lowest solidity and the highest aspect ratio possible. Loss is only one factor to be considered in the design of compressors and there are other34 Useful basic ideas strong reasons for having low aspect ratio and fairly high solidity; a successful compressor is one in which low loss (high efficiency) is balanced with wide operating range, large pressure rise and satisfactory resistance to stall. ‘The analysis in this section has taken as an example the blade-to-blade Now in an axial machine. This was for convenience and because it was the area which had in the past received the most study. However, the ideas are much more general. The two-dimensional assumption would not be exact, for even on the blades of axial compressors the boundary layers are three-dimensional. Close to corners, on the hub and casing walls and generally in centrifugal com- pressors the flow is markedly three-dimensional. Nevertheless the crossflow velocities in the three-dimensional boundary layers are usually much smaller than the streamwise component, at feast until very close to the wall, If for this purpose a turbulent eddy viseosity can be imagined to exist then the effect of the crossflow velocity on the three-dimensional energy dissipation integral is likely to be second order. Asa means of assessing loss production the treat- ment here is probably an adequate guide, pointing to the low levels for laminar flow, the dependence to the third power on free-stream velocity and the insensi- tivity (o the shape of the velocity profile (and therefore to the pressure gradient) Forthermore the ideas are probably transferable to more general flows where the boundary layer approximation is not valid and a full turbulent Navier— Stokes type solution is required. 1.6 Efficiency ‘The term efficiency finds very wide application in turbomachinery. Sometimes it is used for cascades or blade rows to relate the static pressure rise to the drop in stagnation pressure, but this is much less important than its use applied to complete stages or machines. For all machines or stages efficienoy is defined as work into ideal compressor ‘work into actual compressor whore the pressure rise from the ideal compressor is equal to that from the actual. There are several different ways of evaluating efficiency and these reveal different information. The ideal compressor will be reversible in the thermo- dynamic sense and depending on the duty it may be isothermal (constant temperature) or adiabatic (no heat flow io the gas). For duties where the gas is to be stored for a long time before use or where it is to be cooled further, for example in refrigeration plant, the isothermal compressor is the appropriate ideal. When the gas is to be used directly for propulsion or to be heated ot burned then the adiabatic and reversible ideal is the sensible choice. The adiabatic application is more common and will be considered first. seinenEfficiency 35 Adiabatic compression An ideal compressor which is both adiabatic and reversible cannot alter the entropy of the gas flowing through it; such compressors are usually referred to as isentropic with the corresponding efficiency described as the isentropic efficiency, een = tte ven = where the subscript s denotes entropy held constant, ‘The restriction to adiabatic means that the work input is equal to the rise in stagnation enthalpy, W=hoz— Ao, but for brevity at this stage the distinction between stagnation and static enthalpy will be overlooked and so W’—hy—h. The efficiency is then given by (1.39) Ideal gases Equation 1.39 is valid without restriction to the nature of the gas being com- pressed but great simplification is possible if the gas can be assumed to be ideal. In this case h=c,7, with c,, the specific heat at constant press either constant or a function of temperature alone. If c, is constant (i.e independent of temperature as well as pressure) the gas is often referced t as perfect. For all ideal gases it follows that p/p= RT, with p the density and R the gas constant, Por gases at low pressures (low in relation to their critical pressure) the ideal gas assumption is very good but in many engineering applica tions of compressors there are significant departures from the ideal gas relations to an ideal ga: af the reduced variables, ‘the law of corresponding sates’, p/p, and and T, being the critical pressure and temperature. Figure 1,17, taken from Reynolds (1979), plots values of Z against reduced pressure and Table 1.1 shows approximate values of reduced pressure and temperature for a range of common gases taken from the same source. Fortunately most air compressors work within a range of reduced temperatures between about 2 and 5 so that the excursion of the compressibility factor from unity is very small even when the pressure is well above the critical yale. cy and the rato of specific heats wary 2 little over this temperature range: in the case of air, which is very thoroughly documented, cp ranges from 1,003 to 1.066 kI/(kg K) and from 1.4()1 1036 Useful basic ideas @ 9 10 Reduced pressure P/P ses Fig. 1.17 Generalized compressibility chart, valid for most gases and vapours. (From Reynolds, 1979) 1.368. For the newer gas turbines with overall pressure ratios of around 40:1 it is necessary to include the variations in cy and y, as discussed below. many gases, including refrigerants and hydrocarbons, the oper temporaturcs are not significantly higher than the critical temperature and exten- sive deviations from ideal gas behaviour are normal. Aerodynamic design, however, is always on the basis of an ideal gas and the recognition of the non- ideal nature is relegated to the overall aspects of the machine, including assess- ment of efficiency, as discussed below. For an ideal gas it is possible to write the isentropic temperature ratio in terms of the pressure ratio ) ty This yields the simple expression for the isentropic efficiency (r= Walp rah (1.40) Tisew (P/F pai i ‘The irreversible compression and its reversible counterpart can be shown on t the T—s diagram, Fig. 1.18, and if the ordinate were replaced by enthalpy it would apply to nonideal gases as wel e 5s diagrai other words the minimum | temperature rise to produce a given pressure rise increases zs either the initialEfficiency 37 ‘Table 1-1 Critical pressures and temperatures for common engineering substances Values as given by the computational equations in Section 3 of Reynolds (1979), these may differ slightly from the true values. Critical Critical Chemical temperature pressure Substance Sormula (k) (Pa) _ Helium-4 Het 5.2 0.228 Hydrogen (pare) H@) 32.9 1.28 Neon Ne 44.4 2.65 Nitrogen Ny 126 3.4 Air - 133 3.77 Argon A 151 4.86 Oxygen Op 155 5.04 Carbon dioxide CO, 304 7.38 Ammonia NB 407 116 Water 1,0 647 22.1 Methane CH, 191 4.60 Ethane CiHg 306 5.01 Propane Coy 370 4.24 Isobutane Cato 409 3.68 Butane Cath 424 3.22 Pontane Cally, 467 3.24 Hexane CH 506 2.93 Heptane Hic 538 2.62 Octane CHs 568 2.40 Ethylene Cay 283 5.08 Propylone CoM 365 461 Refrigerant 14 cr, 228 3.78 Refrigerant 503 Mixture 293 4.33 Refrigerant 23 CHF, 299 484 Refrigerant 13 ccIr, 302 3.87 Refrigerant 502 Mixture 355 4.07 Refrigerant 22 CHCIE, 369 498 Refrigerant 500 Mixture 379 4.43 Refrigerant 12 COLE, 385 4.12 Refrigerant C-318 Cue 389 278 Refrigerant 114 CaChE 419 3.27 Refrigerant 11 COLE 471 44l temperature or the entropy increases. In a multistage compressor this means that the work input required to produce a given pressure rise is greater for the later stages because the temperature is higher and also that the work input Fequited by the later stages is raised because of the losses in the early stages38 Useful basic ideas i r i q i i pe ‘ rk 2% c Bs r ‘ : 1 | Entopy § ‘ 18 Schematic temperature entropy diagram showing reversible adiabatic com- pression 1=2s and irreversible adiabatic compression {—2 The isentropic efficiency of compressors of identical aerodynamic quality therefore gets lower as the overall pressure ratio is increased and this trend may be confusing or misleading. It can be avoided by using another definition for efficiency for adiabatic compressors, the so-called polytropic or small-stage efficiency, mp. The polytropic efficiency removes the penalty for higher pressure ratios so that compressors of equal acrodynamic quality but signifi- cantly different pressure ratio would have the same polytropic efficiency though a different isentropic efficiency For ail processes Tds=dh —dpip and for an adiabatic and reversible one dh,=dp/p. Ina real compression the enthalpy rise dh will be larger than dig and it can be written re dn = Lan, = 1¥ 1 1 np If np is assumed constant over a finite change in pressure it is easy to show i for a perfect gas that j Ty ( pa ye ( [ qT, Pt and this may be rearranged to give the polytropic efficiency as I 1p =! In@aip) (1.41) vy Int/T 1) ‘aa ae ae as! Hence if Ty/7, and p/p, are known the overall or average polytropic cffi- ciency for the compression process can be calculated. In a real multistage com- pressor if cach stage had an equal value of mp the polytropic efficiency for a atEfficiency the whole machine would be equal to that of each stage. Using the temperature ratio derived above for a given polytropic efficiency one can then calculate he corresponding isentropic efficiency ep apy (1.42) ison and the results obtained are displayed in Fig. 1.19 for a perfect gas with the ratio of specific heats equal to 1.4. The difference between the polytropic and isentropic efficiencies inoreases with raising of the pressure ratio and lower~ ing of the efficiency, the two being equal for p2/p,=1.0 or for a= 1.0. 100, — rr op Polytrapic ofticenoy 80] Icontropic efficiency tisen % 80) Pressure ratio p/p, — — — bry air T, = 722K oi Bir 7} = 288K — Perfect gas y= 1.40 Fig. 1.19 Iscntropic efficiency as a function of pressure ratio and polytropic efficiency for a perfect gas (y = 1.40) and for dry air. (From Koch, 1964) Even though air at pressures and temperatures around the standard atmos- phere is quite a good approximation to a perfect gas, the pressure ratios now considered for gas turbines ate sufficiently liigh and the level of precision required sufficienty fine that the deviations of air from the properties ofa perfect gas must be allowed for, Figure 1.19 also shows a comparison of the predicied ‘elation between isentropic efficiency, pressure ratio and polytropic efficiency 3940 Useful basic ideas for ait obtained by Koch (1964) using the real properties of air tabulated by Keenan and Kay (1948). With the real gas the temperature has an effect and Tines are drawn for two different inlot temperatures. As Fig, 1.19 shows, the isentropic efficiency can be significantly underestimated if air is treated as a perfect gas, the discrepancy becoming larger as the pressure ratio is increased, the inlet temperature increased or the poly:ropic efficiency reduced, The calculations giving tise to the lines in Fig. 1.19 were for dry air, and water vapour also affects the properties of air to an extent which needs to be con- sidered at high pressure ratios. As has been remarked above, the various stages of a multistage compressor contribute differently both to the overall pressure ratio and to the overall effi- ciency. Algebraic expressions can be derived to show this but it is perhaps more instructive to give a numerical example. Consider a ten-stage compressor with inlet air temperature 300 K and equal temperature rises per stage of 47.6 K. For this purpose air will be treated as a perfect gas. If each stage has a polytropic efficiency of 90 per cent the overall pressure ratio is 20.0 and the overall isentropic efficiency is 85.22 per cont, Because the stage inlet temperature is so much higher at the back and the temperature rise is specified as constant, the first stage gives a pressure ratio of 1.59 and the last stage only 1.22. Suppose now that the polytropic efficiency of the first stage is only 85 por cent bnt the efficiency of the other nine stages remains 90 per cent: in this case with the same temperature rise the overall pressure ratio would be 19,5 and the isentropic efficiency 84.15 per cent, If, on the other hand, the first nine stages have a polytropic efficiency of 90 per cent but the last stage only 85 per cent then the overall pressure ratio would be 19.8 and the overall isentropic efficiency 84.73 per cent. Real gases In many instances related to industrial compressors exceedingly high outlet pressures are obtained from multistage compressors, several hundred at- mospheres, for example, Furthermore the gases used are often those which do not approximate well to ideal gases even at standard atmospheric condi- tions. Contracts may contain specification of the efficiency and disputes over as little as half of one percentage point have been known. The specification of officiency with real gases is not necessarily straightforward and can lead to confusion and dispute. The application of isentropic efficiency is exactly the same as that for ideal gases except that the enthalpy rise corresponding to the isentropic pressure rise must be obtained from an appropriate empirical data correlation. It is normal, however, to use a polytropic efficiency as the measure of performance (o take out the bias which would otherwise be in- troduced in favour of low pressure rativ machines. Given a satisfactory equa- tion of state the differential expression defining the polytropic efficiency, dh =(dp/p)/ny, can be integrated numerically to deduce an overall value for np from the measured values of T,/T, and py/p,. This process can be as accurate as the equation of state allows and is consistent in the sense that noEfficiency 41 additional assumptions have been introduced: within the limits of the assump- tion of equal m) for each stage and a given equation of state the method is ex- act provided the numerical scheme is accurate. Examples of this procedure have been given by Nathoo and Gottenberg (1983) and independently by Huntington (1985) The aim is normally to derive the overall value of np, so this is a constant in the integration, but if the intention were to investigate the thermodynamics of the compressor the numerical integration could be performed with rp allowed to vary from stage to stage. The actual machine will not have the same efficiency for cach stage but one which varies; which stage is more efficient depends on the designer's choice. Fozi (1985) gives un example of a six-stage compressor with ethylene entering at 25 bar and 37°C and leaving at 224 bar and 262°C, The polytropic efficiency across the entire machine is about 75 per cont, but representative polytropic efficiencies for ihe individual stages were estimated to be, from front (© back 81.0, 75.9, 71.6, 73.1, 64.0 and 63.7 per cent (the lower efficiencies towards the rear were because the flow passage area was reduced as the density was increased), Taking a single average efficiency is a perfectly satisfactory procedure for establishing an overall assess- ment of the performance of a machine but it is possible to become confused in thinking about the real machine. A different outlet temperature T) would be produced for the same pressure ratio if the schedule of efficiencies for the different stages were altered: the overall performance of the machine depends on the distribution of efficiency in relation to the temperature and pressure through the machine, remembering that pressure and temperature are not related in this case by a simple ideal gas relation ‘Nowadays numerical methods are so common and computers so readily available that the use of numerical integration to derive the polytropic effi- ciency seems natural. Twenty fives years ago this did not seem to be the case and Schulz. (1962) devised a simple approximate method for calculating ‘p. The method essentially treats the fluid as an ideal gas, so that the compression is assumed to follow a polytropic compression p/p" = constant, with the value of 1 determined from the end points of the compression. The Schulz method has become the industry standard as part of the ASME Power Test Code 10-65. Itis very much an approximation which is difficult to assess precisely because the errors depend on the exact form of the equation of state for the substance being compressed and on the route in property space (i.e. how p varies with ) which the compression takes. An accurate equation of state is clearly essen- tial. Huntington (1985) compared his numerical integration with Schulz's method and another approximate method by Mallen and Saville (1977) for four cases of compression with a range of conditions, one with Refrigerant-12, two with ethylene and one with carbon dioxide. Errors between Huntington's numerical scheme, which is essentially exact, and Schulz's method of 0.2, 1.7, 1.6, and 1.3 per cent were found. Errors of roughly similar magnitude but sometimes different sign were found with the method of Mallen and Saville Huntington was able to derive an approximate method which agreed very42 Useful basic ideas closely with his numerical method (which is essentially exact if the equation of state is known), It was pointed out in the discussion of Huntington's paper that ultimately what matters most is that the supplier and customer should agree in advance on what method of calculation should be used since the differences are oo small to be interesting from the point of view of fluid mechanics and cither will give an adequate relative assessment of quality. In comparing tests with different substances the manufacturer will want a consistent, reliable and realistic method of calculating efficiency and for this a numerical scheme or the approximate method devised by Huntington would seem superior to earlier methods. ‘The ideal isothermal compression is shown in Fig. 1.20 as the path 1—27'on the temporature entropy diagram. The ideal process will, of course, be rever- sible. The corresponding process for an irreversible adiabatic compressor is shown by the path 1-2 and roughly identifying the enthalpy rise with temperature rise it can be seen that a great deal of unnecessary work input has taken place which must then be ‘undone’ by cooling from 7, down to To. An alternative irreversible process is shown with the path 1=2', where at the intermediate pressure Pia, the gas is cooled before a second stage of compres: sis ressure Py eni a high pressure ratio compressor : ‘To calculate efficiency for an isothermal process more careful reasoning, must be introduced than in the case of the adiabatic process. If the gas can be adequately approximated as ideal the enthalpy is a function only of the temperature and will be equal at stations { and 27° in Fig. 1.20. For real gases this is no longer the case and the enthalpy will be lower at 27'than at { even though the temperatures are equal. Now it can be shown that the smallest work input to produce a given change in property is equal to the change in the availability function, Haywood (1980), so for unit mass flow the mininxum ‘work input to produce the compression from 1-27 is Wig = (tar ~To82y) ~ (ha Tos) Here Ty is the temperature of the surrounding environment to which heat can be rejected. This expression is valid for real gases as well as ideal ones but if the gas is ideal then jt’can easily be shown that the entropy change is Sor—5; = —RIn (p2/p,), where R is the gas constant, and hyp=hy. A formal statement of the isothermal efficiency is therefore Muotems = Leesan) — hs = Tos) 1.43) Macteat a a ‘ia ae” “ato aeEfficiency 43 Entropy $ Fig. 1.20 A schematic temperature entropy diagram showing an irreversible adiabatic process 1-2 (broken line), an adiabatic process with intercooling at pressure pi. and ‘an isothermal compression 1-27" which for an ideal gas reduces to Efficiency in terms of pressure loss It is sometimes convenient to think of efficiency as the ratio of the actual pressure rise to that which would be produced with the same work input if the process were ideal. Restricting aitention to ideal gases under adiabatic con- ditions the same work input corresponds to equal temperature rise, The com- parison is illustrated in Fig. 1,21 witere py is the pressure actually produced whereas py is the ideal pressure rise for the same work input, The efficiency may then be written AP, P21 (Lady APigea — Pai — PL Entropy & Fig. 1.21 A schematic temperature entropy diagram for a compression process with pressure loss py; ~ p = Apy,44 Usetul basic ideas If the pressure losses in the machine can be estimated then p2=p2)—EAp1, where Lp, is the sum of the stagnation pressure losses through the machine For a perfeet gas the pressure ratios may be replaced by temperature ratios so that = JT If the temperature rise is much less than the inlet or outlet temperature, which it will be for most axial stages and for many entire low-speed compressors, | then this equation may be expanded by the binomial theorem to give Ts ~ Tr yo Th which is the form for the isentropic efficiency, equation 1.39, in the case of an ideal gas, Giving efficiency in terms of the pressure loss is convenient when the temperature changes arc vory small, for example in low-speed machines used for research, or when the correlation for lass is given in terms of pressure Joss, In the next chapter the loss correlation will be used in this way to estimate hypothetical stage efficiency. : The loss is given for blade rows in terms of the relative flow velocities in the appropriate frame of reference fixed to the blade row. Care is needed because the stagnation pressures (and the loss in stagnation pressure) change when the frame of reference is changed. ‘This difficulty can be avoided by working with the entropy, since entropy is equal by definition for stagnation and static conditions and is therefore independent of the frame of reference adopted. Use is made of TyAs = Alty—Apyloo, where A denotes the changes because of loss. By definition the loss is the drop of stagnation pressure at a fixed value of stagnation enthalpy so that Afig=0 and the entropy can be related directly to the loss in stagnation pressure. The rise in enteopy depends on the local value of stagnation temperature Ty which is varying throughout the compression process, and for complete consistency the entropy change should be obtained from an integration, Since the object of converting to entropy rise from stagnation pressure loss ia to allow a proper summation to bo made this does not matter and if the value of Apg is given at the blade row or impeller trailing edge it can be converted to entropy loss using the local value of temperature, A more fundamental problom atises if the temperature is very q 4 eee ee