EXPEDITED AGILE SYSTEM DESIGN AND REDESIGN APPROACH FOR INDUSTRIAL PUMPS

Transcription

1 Proceedings of NRC/ASME: Symposium on Valve and Pump Testing July 15 18, 2002, Washington DC EXPEDITED AGILE SYSTEM DESIGN AND REDESIGN APPROACH FOR INDUSTRIAL PUMPS Dr. David Japikse Concepts NREC 217 Billings Farm Road White River Jct., VT ABSTRACT L/D hydraulic passage length/ hydraulic diameter A new method for the design of industrial pumps has been gpm gallons per minute created which brings a well-integrated system of advanced KBE technology to the fingertips of all designers throughout the W/ W pump industry, regardless of their specific technical on relational velocities background. The new system, the Expedited Agile System MoT Manager of Technology (EASy! 1 ) integrates the initial meanline design considerations, NPSHR net positive suction head required the details of sophisticated three-dimensional blading design, N s specific speed, US units and the essential characteristics of modern computational fluid N dynamics (CFD) analysis and design procedures. The sm specific speed, metric optimization code, isight 2, is used to develop the specific TEIS two elements in series vane shapes. Furthermore, extensive use is made of historical W/ W blade-to-blade loading coefficient, design and performance data to guide designers in the direction based on relational velocities of proven design technique, and to validate and calibrate the design tools for specific applications. Pumps are extremely important to energy systems worldwide. More than 1% of all the fuel burned in cars and trucks could be completely saved if advanced designs were introduced simply for the water cooling pump on these vehicles. Additionally, over 5% of all electric power is consumed in pumps which, on average, probably can be improved by approximately 10%. These savings are enormous, but the world is routinely producing pumps today which use technology which essentially was developed in the first half of the previous century. The new EASy! suite is a major step in fast-forwarding the design process and providing an opportunity for energy efficient, high performance pump design throughout industry. Good manufacturing methods are, of course, required to produce quality pumps. NOMENCLATURE EASy! Expedited Agile Design System NPSHR net positive suction head required 1 EASy! is a trademark of Concepts ETI, Inc. 2 isight is a trademark of Engineous Software Inc. Knowledge Based Engineering blade-to-blade loading coefficient, based BACKGROUND Industrial pumps play a significant role in the world economy and in the issues of energy conservation and manufacturing production. Over the past 200 years, mankind has evolved a number of diverse pumps which meet endless demands from society including fresh water supply, sewage handling, petrochemical industry demands, automotive cooling, human biological processes (heart pumps, etc.), rocket turbopump propulsion systems, and so forth. Indeed, recent studies indicate that more than 1% of all the fuel burned in cars and trucks could be completely saved if advanced designs were introduced simply for the water cooling pump on these vehicles. Furthermore, it has been known for several decades that over 5% of all electrical power is consumed in pumps. On average, these could probably be improved by approximately 10% (see examples in Section 3). Clearly, these savings are enormous, but the world is routinely producing pumps which today use technology that was essentially developed in the first half of the previous century. This paper illustrates a modern, well-integrated technological system which permits a radical departure from past practices and the easy introduction of advanced designs and redesigned industrial pumps. Copyright 2002 Concepts ETI, Inc. 1

2 The common industrial turbomachinery design approach has been thoroughly reviewed and a number of modern references, including Japikse (1993 and 2002), trace this history. Broadly speaking, the world recognizes the need to do thorough meanline optimization early in a design (or redesign) process, the need for careful geometric layout of impeller, diffuser and volute geometries, and the use of modern and comprehensive computational fluid dynamic (CFD) procedures. All of these techniques have been thoroughly integrated into modern agile systems and expedited so that most of the laborious and tedious protocol procedures of the past have been codified and eliminated from the daily regimen of designers exercises. Indeed, the advances are so significant, that it has been recognized (2000) that the modern engineer must no longer spend his or her time manipulating computer codes and databases, but rather must concentrate on being a Manager of Technology (MoT). Further comments from these important steps are now provided. Meanline design optimization assures the designer of achieving an optimum velocity triangle at each important station in an advanced turbomachine and, ultimately, a realistic performance level. It is based on empirical information and provides stabilization to the basic design process by allowing the designer to review past design experience and, therefore, assure him or her of remaining within rational bounds in the design. Properly done, it allows one to seek optimum velocity triangles subject to known design criteria so that reasonable choices can be made for subsequent optimization using more powerful design tools (through flow and boundary layer calculation methods and CFD techniques). At this level, one uses an extensive empirical database to search past design experiences and then to project what further potential gains might be made. This step is illustrated in Section 2 below. When a proper meanline design optimization is complete, then the designer creates a prototype three-dimensional configuration and begins to evaluate it in great detail. Today, this can be done in seconds and requires very little designer manipulation. The process then of optimization of the shape requires careful examination of surface velocities and pressures throughout the impeller (and other elements) with careful attention to minimize secondary flows, leakage effects, incidence effects, and all other loss-producing mechanisms. Even in the recent past, of just a year or two ago, designers would spend days and sometimes weeks searching for these optimum conditions. Indeed, this historical process was so complex that there was a parallel effort, historically speaking, of using inverse design methods which would seek a geometry which would guarantee a certain velocity or pressure distribution inside a passage. The weakness of the inverse design process was recognized as being the ever-changing perception of what is, in fact, a true or desirable optimum velocity or pressure distribution within a passage. Today, all of this has changed because the direct process, where one postulates a geometry and evaluates flow characteristics, is now so fast and flexible, that inverse methods really cannot hope to compete. The use of expert optimization systems to guide the use of the agile tools is illustrated in Sections 2 and 3 of this paper. In short, the commercial designer is now in the control seat, figuratively speaking, of a very powerful modern designerdeveloped system focusing the total design process. Computational fluid dynamics, or CFD, has a variety of definitions, but usually focuses on the solution of the 3D Navier Stokes equations for fluid mechanics. At this tool level, we hope to effect the most profound calculations possible for fluid mechanics and to carefully assess the internal losses and characteristics of a given pump. Today, these methods are not only widely available, but have been made extremely fast, very robust, and quite accurate. Nonetheless, the user should realize that the method still has an empirical foundation: an empirical turbulence model (which presently is devoid of transition, curvature, and strong rotation effects in most published cases) plus gridding approximations and numerical characteristics. Each effect has an empirical or experimental character to it. The methods illustrated in this paper, are believed to be second to none in terms of speed, ease of use, and accuracy. To illustrate this point, two examples of validation for the CFD method of this study are shown here. Figure 1 shows an independent company s evaluation of the Pushbutton CFD 3 code used also in the present study. This pump manufacturer has validated the Pushbutton CFD tool over a wide range of designs to see how well this CFD models the head coefficient compared with their measurements. Acceptable criteria lines, the solid black lines in Figure 1, are set at approximately ±2% to 4% over a wide range of specific speed (European definition). At exceedingly low levels of specific speed, larger errors are permitted. In general, the present tool meets a ±2% to ±4% accuracy over a very important industrial range. Probably, this range can even be reduced in future efforts. A second example is shown in Figures 2a and 2b showing the CFD calculation of head and efficiency for a robust axial flow industrial pump that is characterized by very shallow inlet and outlet blade angles and Acceptable range of designs Figure 1. Comparison of predicted head rise coefficient from Pushbutton CFD with industrial pump data, courtesy Egger Corporation. 3 Pushbutton CFD is a registered trademark of Concepts ETI, Inc. 2

3 substantial inlet and outlet blade thickness. This validation shows the inherent integrity of the Pushbutton CFD tool which is used herein. More examples are given below. The cases which follow include one new design and two redesign cases for the pump industry. Details of the methods utilized are presented in companion papers (Japikse, 1998, 2000, 2002) and in the modern textbook by Japikse (1997). This paper presents the basic overview of how these design methods are working and particularly concentrates on illustrating the modern attributes of the Expedited Agile System (EASy!) as now utilized by a number of lead pump organizations around the world. The implementation of isight is shown here for the first time with EASy! Inherent in this presentation is an emphasis on several key advances which will be widely used in the next few years and are worthy of careful review at the present time. Impeller Total Head - No Bend (Feet) Volume Flow (GPM) Lab Data (Arith.) CFD at TE+30 Figure 2a. Axial pump head rise by Pushbutton CFD and client measured data. ORIGINAL DESIGN CASE The specifications for our illustration of modern design are as follows: a. Flow rate equals: 300 gpm b. Rotational speed: 3,600 rpm c. total dynamic head: 100 ft. d. NPSHR: 14 ft. For this design, an impeller with a 25 exit vane angle with seven vanes, being three-dimensional in form (covered), and with an exit volute shall be employed. The inlet pipe diameter will be three inches and the exit pipe diameter will be two inches or larger. The first step in the design is to find the optimum eye dimensions and the appropriate diameter and tip depth for the impeller in order to assure the design criteria are met. To do this, all pertinent design principles, as found in common text books or modern technical papers, have been codified in a graphically usable interface (GUI) so that all design parameters can be quickly and appropriately set. Figure 3 shows one of several early steps wherein the exit vane angle is set at 25, exit blade thickness is set, and the vane count of seven is employed. Only seven or eight graphical images such as Figure 3 are required to set all the appropriate parameters for the meanline design optimization. As mentioned above, however, there is the important issue of choosing proper and valid performance modeling for the meanline design optimization based on past pump design and development experience, while allowing for some future growth potential. Figure 4 illustrates this process. Efficiency (%) Volume Flow (GPM) Lab Data Rotor Eff. (Arith.) CFD at TE+30 Figure 2b. Axial pump rotor efficiency from Pushbutton CFD and client measured data. A prototype first design has been automatically generated, as displayed in the lower right-hand corner of Figure 4, and then a complete database search of past similar pumps is reported in the inset Figure 3. Sample GUI frame to define new pump design variables. 3

4 Figure 4. Re-evaluation (second or later iteration) of expected machine performance by interactively accessing a comprehensive performance database of past design experience (shown as inset). illustration of Figure 4. A list of important design variables is shown in the left-hand corner of this inset figure, which is notable in that it includes the specific speed of the pump, the machine Reynolds number for the pump, the rotation number for the pump, and a wide variety of important geometric parameters including inlet blade blockage, passage L/D hydraulic, aspect ratio, and other key parameters such as clearance and blade backsweep. This is the first turbomachinery design system ever to incorporate the important rotation number; this is only possible due to the fact that an exhaustive study of a wide range of past designs and data has been undertaken which has revealed the importance of this parameter. Other important geometric parameters have been discovered recently and included in this database approach as suggested by the inset. The tables on the righthand side of the inset correspond to searches through the empirical database and establish the number of matches according to the pertinent design variables of this particular study. Indeed, it is possible to find several cases which match five of the particular independent variables for this design study and, as a result, certain modeling parameters, which control the impeller internal losses, exit slip or deviation, and secondary flow development can be specified with some confidence. After the designer quickly reviews (remember, the designer is now the MoT or Manager of Technology) the information available, the preferred case is clicked upon, modified if necessary, and a new design optimization is launched based on the then current resulting data. The resulting meanline optimization is shown in Figure 5. In this case, a good head rise characteristic is shown in the lower left-hand corner and a nice efficiency characteristic is shown in the upper right-hand corner. The power required at the design point is approximately 9 hp (upper left corner) and the NPSH characteristic, corresponding approximately to 3% head breakdown, is shown in the middle inset. The design goal of 14 ft. of NPSHR at the design flow of 300 gpm has been met; as is well known, this pump should operate with less NPSHR required at lower flows and significantly higher levels at higher flows as the inset shows. Consequently, a meanline design optimization has been effected and this step now requires minutes rather than hours or days as in the recent past. Figure 5. Optimized meanline pump stage design showing power, efficiency, head, meridional view and, as inset, the NPSHR (in feet). 4

5 Figure 6. Automatically generated first trial geometry of the entire stage. Upon completing the meanline design optimization, and, therefore, establishing a reasonable base of confidence that a good design can be effected, we shift over to the full blading representation. A first view of the entire pump stage is automatically generated (in a matter of seconds) as displayed in Figure 6. If this initial geometry were built, it would function near the design intent parameters, but could have some minor mechanical defects and certainly would not be at the optimum efficiency. the incidence shown in the center of the bottom panel is far excessive, and the very high level of incidence at the hub will engender substantial losses as would the negative incidence at the shroud. Such bad distributions of incidence are typical of many historical designs where rapid, stable, and accurate through-flow codes were not available. Design improvement at this level need not be carried out with CFD but can be done very rapidly and effectively with good through-flow codes. CFD is well applied later in the design process. If we were to conduct a CFD calculation of this first generation prototype design, we would see that an impeller passage hydrodynamic efficiency of about 96.2% is to be expected. Clearly, we would hope that we can improve this somewhat through subsequent design optimization. The basic impeller vane design is described by a hub passage contour, a shroud contour, a hub blade angle (beta) distribution, a shroud beta distribution, and hub and shroud Figure 7 shows the smooth calculation of the streamtube locations in the upper left-hand corner and starting values of the relative velocity distributions for the shroud and the hub in the middle top panel. These distributions are far from optimum as the blade is not yet shaped properly to close the velocity distribution at exit. In addition, the blade-to-blade loading of the hub is very high, as shown in the upper right-hand corner, far above the common 0.7 value, which is the beginning of the yellow band. Additionally, FiFigure First First pass pass streamline streamline calculations calculations showing showing 1) 1) incidence, incidence, 2) 2) surface surface velocities, velocities, 3) 3) blade-to-blade loadings loadings ( W ( W/ W/ W ), ), and and 4) 4) computed computed streamline streamline shapes. shapes. Items Items need need imimprovement through through subsequent subsequent blade blade design design exercises. exercises. 5

6 thickness distributions. These six distributions uniquely specify the entire shape of the impeller vane, unless we wish to use bowed blading. If the latter is desired, then more surfaces are to be described (a straight-forward process today). The six distributions give us considerable flexibility in modern design. Indeed, many design attributes have been significantly underutilized in past design work. For example, hub thickness is a tremendous asset for controlling hub loading. This has not been used greatly in the past because highly flexible and convenient tools have rarely been available to the pump designer. Aerospace engineers have long utilized powerful design tools which allow them to custom craft a nearly ideal set of distributions (all six parameters). In so doing, a carefully manipulated hub and shroud shape, utilizing variations in hub and shroud thickness distributions in creative ways, plus carefully manipulated blade angle distributions, leads to a good final product. Much skill was accumulated through this process, and for decades, this was the pride of the experienced designer. From approximately 1950 to the early 1990s, it was recognized that industry was compelled to hire only the experienced designer who had, typically, ten or more years of design experience. The author and many colleagues went through this process and acquired this historical design experience. With the introduction of several special contributions to turbomachinery technology, it has become possible to capture this past design experience and make it easy to use, even for the novice (but this will require certain issues of responsibility for the industry). Both by using expert or system Knowledge Based Engineering (KBE techniques as now being researched and developed at Cambridge, Oxford, and other leading institutions) and by capturing the readily available current design approaches using optimization procedures such as isight from Engineous, it is possible to skillfully control these six important design parameters with common and automated procedures for half of the variables. The user controls the other set at this time. It is to be expected, in the next months or couple years, that most of the design variables can be taken over with modern optimization search procedures. The speed today is commendable, as the stability of the processes has now been 1 established. and the design process, and they are extremely fast and have been reliable in past design experience. These methods allow us to establish surface relative velocities and then we find the appropriate levels of blade loading at the shroud and the hub. An important blade loading coefficient is the difference in relative velocities between the two sides (suction and pressure sides) divided by an average value. This parameter, WW /, is the well-known blade-to-blade loading coefficient and is invariably kept below a level of approximately 0.7, or sometimes slightly higher or lower values are employed. Experienced designers who have used through flow methods are very comfortable with this process. Streamline curvature calculations, such as shown in Figure 7, are very fast in establishing these values and can be run in approximately one second in modern, well-validated systems. By setting an appropriate target function (see the dotted lines in Figure 8 for the blade-to-blade loading coefficient and the pressure recovery coefficient), it is possible to let a mathematical optimization search technique, such as the isight suite of algorithms from Engineous, to take over the tedious manipulation process of some of the historical design protocols. For this study, we allow isight to search for the hub and shroud blade angles, as well as the hub contour. To do this, we must start with reasonable shroud and hub contours, sensible starting blade angles, and rational levels of hub and shroud thickness. These were automatically generated, as shown in Figure 8. isight uses a variety of techniques, but this design was started with approximately 250 iterations of simulated annealing followed by several hundred iterations of the method of steepest ascent (hill climbing). Instead of manually adjust- Consequently, we can proceed from the illustration shown in Figure 8 and a set of target objective functions as shown in Figure 8 for the blade-to-blade loading coefficients and the key suction surface diffusion characteristics as utilized by experienced designers. Historically, the powerful through flow (or quasi-3d) techniques have been used to capture most of the important physics Figure 8. Details of the first design pass showing excessive aft-end loading [1] and poor incidence [2]. isight may be used to automatically redesign the vanes to cure these defects. 2 6

7 ing each of the Bezier points which are used to describe the hub and shroud beta distribution and the hub contour shape, they were manipulated by isight until an optimum set of loadings was achieved. The final results are shown in Figure 9. The CFD calculation is initiated by clicking an appropriate launch button and all problems of appropriate gridding, solver settings, and run-time issues are automatically set by standard and validated defaults. On rare occasions, a designer might choose to change these defaults, but for most design work they are appropriate and launch a good, repeatable, and accurate CFD calculation. Normally, the designer would reflect on how long he wants the calculation to run and will choose the number of iterative steps. The CFD calculations usually require approximately ten minutes on a modern PC, sometimes less, and sometimes more. For this case, the results are shown in Figure 10. By carefully checking the post-processed output (done automatically), we can establish that this impeller is expected to have a hydraulic passage efficiency of about 97.0%. The extensive validation work carried out for this code indicates a two sigma confidence that this should be accurate to ±2 points and a one sigma confidence that it is probably accurate to ±1 point of efficiency. This assumes, of course, that this stage is built exactly as designed (a process which sometimes is missed in industrial practice). In addition to the impeller optimization, an appropriate volute has been laid out as shown in Figure 6. If desired, CFD can be run automatically on this volute using the FLUENT 4 code with a very complex grid system, which is appropriate to a problem of this type. This is a longer computer run, frequently requiring approximately two hours. In this case, no advantage can be taken of the axisymmetric character of the flow field, and a much more complex gridding and computational process is required. Likewise, if desired, the impeller calculations just presented can also be run in FLUENT in order to have a second, independent, Figure 9. Improved loadings and incidence for a suitable design. CFD shows a gain from 96% to 97% for passage hydraulic efficiency with this isight vane shaping. Figure 10. Pushbutton CFD evaluation of the revised isight-based vane design. Included are fully automatic pre- and post-processing and very short computer time. 4 FLUENT is a trademark of Fluent Incorporated 7

8 X check of the expected performance. This opens additional possibilities for a diverse parameterization of the flow model. The expectations for this stage have been outlined above, and the design process can be carried out in a matter of a few hours. The role of the modern engineer is that of the MoT (Manager of Technology), and the designer is no longer required to serve as a trained monkey to manipulate numbers in a highly repetitive process. Instead, the designer is expected to be well-trained at the university in basic fluid mechanics and have a reasonable familiarity of common PC-based operating systems. Technology behind the system is easy to acquire, study, and to master through published documents such as the corresponding textbook (Japikse, 1997). Such materials are in use in colleges and universities around the world today. REDESIGN EXAMPLES The design process is important in its own right, but redesign is frequently the more common process used in industry. A crude estimate of the number of pumps in use in the world typically yields a value of approximately four or five billion. With such an installed base of pumps available, it is clear that most of the important improvements in our world of pumps must come through modifications of existing pumps. Consequently, two significant pumps, designed by the author and associates over the past two decades, are studied below. Each was a high-performance stage at the time the work was conducted; in this investigation, we illustrate how further improvements can be made. Best Model Fit: ηa = 038. ; ηb = 0.0; DRstall = 13. ; δ2 p = 20. ; χ = 012. Figure 12a. Pump head (ft) versus flow using the TEIS model for a small, high speed (8,000 rpm), N s = 3,500 centrifugal pump. h R (rotor) and ht stage (total) Flow (GPM) Data (Rotor Eff.) Data (Stage Eff.) Best Model Fit (Rotor Eff.) Best Model Fit (Stage Eff.) Figure 12b. Pump efficiency versus flow showing model and data. a. An Interesting Minipump A particular pump shown here, see Figure 11, for redesign is also displayed in the reference textbook, Japikse 1997, as case CPN5/CP9, pages 2-71 and The measured impeller performance in terms of head and efficiency is shown here in Figures 12a and 12b, while Figures 13a and 13b show the measured diffusion and recirculation losses, with appropriate DR Data Best Model Fit 1 Best Model Fit: ηa = 038. ; ηb = 0.0; DRstall = DR2I Figure 13a. Impeller diffusion ratio DR 2i data using the primitive TEIS model φ Best Model Fit: ηd = 007. ; = 104. ; η = 03. ; φd φ φ = 063. ; ηy = 036. ; = 17. φ φ X Y 0.3 D Recirc Data Best Model Fit Figure 11. Photograph of the high speed (8000 rpm) mini pump which is evaluated for redesign Figure 13b. Modeled (matched) and deduced recirculation losses, impeller CPN5, i.e., CP9. Flow 8

9 SOA for Mini Pump Figure 14. Efficiency as a function of specific speed and capacity based on 528 test points. Note N s = N sm (Jekat 1986). meanline model matching included. Additionally, the performance of this pump, including an appropriate estimate for a volute (not included in the original exercise), is displayed in Figure 14, which is one of several industrial standards for expected performance of a pump of this type. This pump was designed in approximately 1985 and reached the state of the art even though it was a small high-speed pump. The question is, however, can this design be improved further? We approach this design issue from the redesign point of view. The redesign begins by using the existing geometry in the meanline code and searching through for higher possible levels of performance. Figure 15 shows the results of a review of the original design. The reader may question where such information could come from. In this case, we allow the data from the world of compressors to be commingled with the data from the world of pumps in order to search the possibility of higher performance levels. When this is done, higher levels of the modeling parameters are suggested, indicating that things may have been learned in the compressor industry which could provide guidance for improved pump design. When these are utilized, then an improved possible pump configuration results as shown in Figure 15. Now it remains to be seen whether appropriate vane shapes can be developed, subject to the same design parameters, to match this expected improved efficiency. Consequently, we use the isight-driven blade manipulation, with the appropriate designer (Manager of Technology MoT) guiding the process to look for all possible avenues gained. The above optimization process with isight yielded a design configuration as shown in Figure 16. The shape is encouraging and a comparison with the original design suggests the possibility of real performance gain. This must be confirmed by using CFD, see below, and eventually by building the modified design. CFD validation calculations for the original pump are shown in Figures 17 and 18. Good modeling has been achieved. The resulting CFD calculations, using the Pushbutton CFD code, indeed, show improvement in efficiency (0.87 to 0.91). The baseline results were shown previously in Figures 12 and 13; the improved configuration is now reflected in the results of Figure 19. The gain is well worth pursuing for further product improvements. b. Improved Turbine Pump Just as the minipump was examined in detail in the preceding section, likewise, an important turbine pump can be examined carefully. This pump was shown on the cover of the textbook referenced above (see Figure 20), and a few characteristics for this pump are shown on pages 7-66, 7-68, and An example of the measured performance for this pump impeller is shown in Figures 21 and 22 with best modeling shown in the figures. Following the procedures of the preceding examples, a thorough exercise was conducted to see if further improvements could be made in this pump. Indeed, this has been possible but only slightly (96.4 versus 96.2). The resulting configuration, with overlays, is shown in 9

10 Figure 15. Recent evaluation of original (ca. 1985) Mini Pump. Notice the poor loading diagrams of the lower LHS two panels and the poor incidence distribution. Figure 16. isight-driven Mini Pump redesign. The loading and incidence is sensibly developed. CFD evaluates a passage hydraulic efficiency gain from 87.8 to

11 TOTAL DYNAMIC HEAD CPN 5 Mini Pump NEW SOLVER Total Dynamic Head (m) g CFD impeller head TE+1 measured impeller head based on average exit conditions Mass Flow (kg/s) Figure 17. Pushbutton CFD impeller head modeling compared to measured data. 1 EFFICIENCY CPN 5-Mini Pump NEW SOLVER 0.9 Efficiency, Total-to-Total measured impeller efficiency (average exit conditions) g CFD impeller efficiency (average exit conditions) measured stage efficiency CFD stage efficiency Mass Flow (kg/s) Figure 18. Pushbutton CFD efficiency modeling compared to measured data. 11

12 Figure 19. CFD comparison of original mini-pump, LHS, and revised, using isight, on the RHS. Notice the far superior pressure at discharge from the revised case. Note: yellow green p ~ 210 kpa; red p ~ 270 kpa 02A 02A Figure 20. Vertical turbine pump design from Concepts ETI, Inc. ca

13 TOTAL DYNAMIC HEAD CPN 7-Stage NEW SOLVER Total Dynamic Head (m) CFD g Head rise at TE+1 PUMPAL Head P02A Mass Flow (kg/s) 35 Figure 21. Impeller head rise modeling with Pushbutton CFD compared to measured data. EFFICIENCY CPN 7-Stage NEW SOLVER Efficiency, Total-to-Total measured data on average basis g Pushbutton CFD efficiency modeling Mass Flow (kg/s) Figure 22. Impeller passage hydraulic efficiency modeled with Pushbutton CFD compared to measured data. 13

14 Figure 23. Evaluation through Pushbutton CFD is likewise illustrated in Figure 24. We conclude that improvements, even in this high performance pump, can be obtained but are not economically practical. The resulting data are shown on a state-of-the-art standard pump performance reference curve (see Figure 14) and, once again, we observe that important improvements in expected performance of the centrifugal pump lie within our reach. Figure 23. isight-based impeller vane redesign of high performance pump. Pushbutton CFD confirmed a small redesign gain. Figure 24. Pushbutton CFD evaluation of redesigned pump vanes. The code is fast, accurate, and easy to use with full pre- and post-processing and validation. SUMMARY, CONCLUSIONS, AND EXPECTATIONS This paper has illustrated aspects of the expedited agile system as applied to an original design and to pump redesign problems. Each exercise can be conducted within a matter of hours and, potentially, be accelerated even further. The tools operate seamlessly from beginning to end and eliminate most of the historical operator code and database manipulations of the past. The design process is comparably easy to learn and can be applied quickly to a very wide range of design problems. In fact, it has been tested on over 20 different common pump design problems before this paper was released. The base of validation grows daily and many aspects of the design process have already been subject to a thorough validation process against measured laboratory data. The reader may appropriately conclude that tools are now available to facilitate a significant upgrading of many basic pump designs as used throughout the world. A new chapter of energy conservation can now be written and designs with truly optimized performance can be effected quickly by competent engineers participating in a well organized, properly understood, semi-automatic design process. Today s engineers, or MoTs (Managers of Technology), must be cognizant of basic fluid mechanics, including turbomachinery design principles and viscous flow phenomena (including boundary layer physics) in order to properly 14

15 monitor and control the design process. However, such an engineer now has at his or her fingertips, a highly flexible, well validated, expedited agile system. ACKNOWLEDGMENTS The author acknowledges the contributions of a large team of specialists who have developed this software, tested it thoroughly for application, and used it routinely in daily design practice. Specifically, Mr. Mark Anderson has directed the Concepts NREC software development team through all of its recent expedited agile system work, Dr. Fahua Gu has upgraded the historical Dawes-based CFD code, Mr. Alex Plomp has upgraded the blading process substantially to facilitate the use of isight, and a dedicated team of software development specialists has supported these three technology leaders. On the application side, Mr. Nick D Orsi, Mr. Tsukasa Yoshinaka, and Dr. Colin Osborne have consistently led an advanced design team which has utilized the code procedures for many years in designing a wide variety of industrial stages, both compressors and pumps. Finally, a team of diverse individuals, including application engineers, software development engineers, and sales and marketing specialists (some of whom were engineers and some were not) have tested the software to determine the ease of use and guided further improvements from their input. The contributions of all of these individuals, plus our historical customer base (which has frequently given substantial input and feedback for additional development) is gratefully acknowledged and deeply appreciated. REFERENCES Japikse, D., and Olsofka, F. A., Agile Engineering: A Look at Tomorrow, Keynote address for the ASME Rotating Machinery Conference and Exposition, November ; appeared as Agile Engineering Accelerates Design, Mechanical Engineering, Vol. 115, No. 11, November 1993, pp (An expanded version of this paper was included in Rotating machinery: problems & solutions, ROCON 93 Proceedings, November 1993.) Japikse, D., Centrifugal Pump Design and Performance, Concepts ETI, Inc., Wilder, VT, Japikse, D., Design System Development for Turbomachinery (Turbopump) Designs 1998 and a Decade Beyond, Presented at the Joint Army-Navy-NASA-Air Force (JANNAF) Conference, Cleveland, OH, July 15-17, Japikse, D., Precision Turbomachinery Products in the Agile Engineering Era, presented at the ISROMC-8 for the Pacific Center of Thermal-Fluids Engineering, Honolulu, Hawaii, March 26-30, Japikse, D., Developments in Agile Engineering for Turbomachinery, presented at the 9 th International Symposium on Transport Phenomena and Dynamics of Rotating Machinery, February 10-14, 2002, Honolulu, Hawaii; also presented at the 2002 Fluids Engineering Division Summer Meeting, Montreal, Quebec, Canada, July 14, Jekat, W. K., Centrifugal Pump Theory, in Karassik I. J., Krutzsch, W. C., Fraser, W. H., Messina J. P., Pump Handbook, McGraw-Hill Book Company, New York,