Proceedings of FEDSM 98 1998 ASME Fluids Engineering Division Summer Meeting June 1-5, 1998, Washington, DC FEDSM98-4988 NUMERICAL AND EXPERIMENTAL FLOW ANALYSIS OF A CRYOGENIC POWER RECOVERY TURBINE Nick Baines Concepts ETI, Inc., 8 Waverley Avenue, Kidlington, Oxon OX5 ND, UK Tel: +44 ()1865 371791 Fax: +44 ()1885 371791 e-mail: nbaines@conceptseti.com Hans E. Kimmel, Ebara International Corporation Cryodynamics Division, 35 Salomon Circle, Sparks, Nevada 89434, USA Tel: 1 7 356 796 Fax: 1 7 356 884 e-mail: hkimmel@ebaraintl.com ABSTRACT In many cryogenic processes, liquids are often expanded from a high pressure to a lower pressure level. Traditionally the expansion has been achieved by reducing the pressure across a Joule-Thompson valve in which kinetic energy in the fluid is mostly transferred into heat. A cryogenic turbine has been developed to expand the liquefied gas in a hydraulic turbine generator, recovering the fluid energy as electrical power and increasing the productivity of the process plant. Since cryogenic processes have changes in volumetric flow and differential pressure, variable speed hydraulic turbines are ideal expansion machines to accomplish high efficiencies for power recovery. An extensive CFD analysis combined with test verification has been performed to optimise the efficiency of the cryogenic turbine. INTRODUCTION Natural gas liquefaction plants are very energy intensive, and the reduction in operating costs is an essential part of any plant optimization study. The conventional liquefaction process involves the expansion of high pressure LNG through a Joule- Thompson valve. This is essentially a constant-enthalpy expansion process. Alternatively, the Joule-Thompson valve can be replaced with a cryogenic expander turbine to recover power during the expansion. If the turbine is linked to an electrical generator, the overall efficiency of the plant is improved (Cengel and Kimmel 1997). Offsetting this advantage is the additional cost and complexity of the turbine-generator. To make the concept viable and to maximize the return for the plant operator, it is essential to achieve a high level of efficiency of the expander. In this paper, the design and development of a turbine expander is described. CFD was used extensively in this process, and was also used to predict the turbine efficiency in advance of prototype construction and testing. The prototype turbine was tested in a water flow rig, and in this way the performance could be checked directly against the CFD predictions. APPROACH The initial design concept was for a multi-stage radial turbine. Multiple stages were necessary to expand the working fluid through the required head drop, and a radial turbine offers the advantage of a higher work output/stage than an axial stage of comparable dimensions (Japikse and Baines 1994). Thus, a single stage will comprise a nozzle row, accelerating the fluid into the rotor, through which it travels radially inwards. The rotor is then followed by an S-shaped return bend to return the fluid to the correct radius for the following stage. This presents a complex flow path, and one which is difficult to analyze. Fortunately, appropriate computer-based analysis and design tools are now available. The basic design information was used to construct a computer model of the turbine stage for analysis. The tool used for this purpose was CCAD, an interactive and highly flexible geometry generator based on Bezier-Bernstein polynomials to represent the hub and shroud lines and the blade angle variation (see Casey 1983). CCAD also enables preliminary estimates to be made of a blade row performance by means of an inviscid, 1 Copyright 1998 by ASME
Condition Table 1. Summary of performance. Preliminary design Modified design Water test, full speed Water test, scaled speed Gross torque (Nm) 15.36 111.81 365.38 494.97 Gross power (kw) 33.83 35.66 743.6 67.38 Stage efficiency.819.894.86.836 Rotor efficiency.853.936.918.98 Head drop: stage (m) 359.15 314.59 34.5 66.74 Head drop: nozzle (m) 8.65 9.5 13.89 4.11 Head drop: rotor (m) 344.89 3.5 34.79 61.4 Head drop: return bend (m) 5.6 4.83 5.85 1.3 Nozzle total pressure loss coefficient.88.88.14.7 Return bend total pressure loss coefficient 1.13 1.13 1.6 1.634 two-dimensional streamline curvature analysis. This feature was used to guide the redesign of the rotor, but the final analysis and assessment was made using FINE/TURBO, a viscous threedimensional CFD analysis code. Meshes of 19, and 15, nodes were used. The coarser mesh was used initially, and the fine mesh was used for the rotor redesign. The properties of the working fluid were those of methane, which closely approximate to those of LNG. The CFD analysis enables a complete assessment to be made of the flow field through each component of the turbine. Additionally, by mass-averaging the properties on planes at inlet and exit of each component, the performance can be estimated. The gross work output of the rotor is calculated from the change in the product of radius and tangential component of velocity (rc θ ) across the rotor, according to the Euler turbomachine equation (see Japikse and Baines 1994). ASSESSMENT OF EXISTING DESIGN The preliminary design was based on an existing pump. The nozzle blade design was developed from that of the pump diffuser vane, reversed and with a semicircular leading edge added. Initially it was hoped that the pump runner could be used unmodified. The first task was to evaluate this turbine configuration. The performance of the turbine stage at the design point is listed in Table 1. The stage efficiency is relatively low at.819, and the comparison between the stage efficiency (based on the head drop available for the complete turbine stage) and the rotor efficiency (based on the head drop available for the rotor alone) shows that this is largely accountable to the rotor itself. In fact the nozzle, has a total pressure loss coefficient (total pressure loss/exit dynamic pressure) of.88 and a deviation angle (exit blade angle exit flow angle) of only.3, both quite low values for a nozzle with such a high degree of turning. The total pressure loss coefficient of the return bend is 1.1 which is a relatively high value, although its impact on the stage performance is reduced by the low velocities through the bend itself. For comparison, the actual total pressure loss in the nozzle is predicted to be.359 MPa at the design condition, and that of the return bend is.33 MPa. An examination of the CFD results showed that there are several poor features in the flow field. Near the leading edge of the rotor the flow separates on the outside of the bend where the flow suddenly contracts (Fig. 1). There is also evidence in this region of a suction surface separation, presumably the result of incidence. Near the trailing edge there is evidence of a large separation on the pressure side of the blade where the flow reverses (Fig. ). On a mass-average basis, the rotor incidence is +7.7. Radial-inflow turbines tend to be particularly sensitive to positive incidence, as can be seen in the regions of separated flow near the leading edge. The deviation angle is 4., which is very large, but the flow distribution is so non-uniform at the trailing edge that this single figure actually has very little meaning. In the return bend, Fig. 3 shows that a very large vortex forms in the outer region of the bend between struts. The effect of the struts themselves is not to eliminate this vortex, but to move it beyond the trailing edge. A separated region also exists on the inside of the outermost bend, but this appears to reattach before the leading edge of the next stage nozzle, and although it is itself a source of loss, it does not appear to interfere excessively with the next stage. ROTOR REDESIGN The rotor was redesigned in several stages in order to reduce or eliminate the poor flow features which were identified in the previous analysis. In summary, the changes were: 1. Increase the blade number from 6 to 1 in order to reduce the loading per blade (a small blade number is often selected for pumps in order to increase the stable operating range; stability, however, is not a problem for turbines).. Modify the hub and shroud contours to reduce the curvature and improve the annulus area distribution. Copyright 1998 by ASME
Figure 1. Preliminary design runner velocity vectors and streamlines. Mid-passage. Figure. Preliminary design runner velocity vectors and streamlines. Mid-span. 8 7 6 R (in) 5 4 3 Figure 3. Preliminary design return bend velocity vectors and streamlines. 3. Modify the blade angles to increase the turning and reduce the passage area going from the inlet to the exit. 4. Eliminate the sudden contraction at inlet to the rotor by changing the radii of the bend just upstream in order to blend smoothly with the rotor inlet. Figure 4 shows a comparison of the original and new meridional profiles. It is apparent that the changes to the rotor imply some significant re-shaping of the return bend also, and that considerably more diffusion must occur in this component. The curvature of the hub and shroud lines is an important parameter, influencing as it does the streamline curvature and hence the 4 5 6 7 8 Z (in) Figure 4. Meridional section of preliminary and redesigned runners. normal pressure gradient which tends to generate secondary flow. Figure 5 illustrates that the changes in curvature on both surfaces are now much more gradual. The changes in blade angle are shown in Fig. 6, and the comparison of passage area is shown in Fig. 7. In a turbine a reducing passage area in the streamwise direction is sought, in order to accelerate the flow through the passage, and it can be seen that the combination of modifications has enables this to be achieved. The CFD analysis shows much better results than previously. On a mass-averaged basis, the rotor efficiency is now.936 at the design point, which is an improvement of eight 3 Copyright 1998 by ASME
1.5-1 1. shroud hub -3 hub shroud hub Curvature (1/in).5 shroud hub Blade angle (deg) -5 shroud. 4 6 8 1-7 -.5 Figure 5. Meridional plane curvature of preliminary and redesigned runners. -9 4 6 8 1 Figure 6. Blade angle variation of preliminary and redesigned runners. 5 Passage area (in ) 15 1 4 6 8 1 Figure 7. Passage areas of preliminary and redesigned runners. points over the preliminary design. The velocity vector and streamline distributions are shown in Figs. 8 and 9. The meridional plane flow (Fig. 8) is now very smooth and regular. On the blade-to-blade plane (Fig. 9) the pressure side of the blade is not quite optimum and there is still a small separation bubble there. Also the curvature on the shroud is still somewhat too high near the trailing edge, and this leads to the small recirculation zone there. However, these are relatively small effects, and the changes noted here have clearly addressed the major shortcomings of the preliminary design. Figure 8. ed runner velocity vectors and streamlines. Mid-passage. RETURN BEND REDESIGN The return bend was also redesigned, both to match the redesigned rotor and to improve its own performance. It was recognized that any significant improvement must rely on increasing the axial length of the return bend, and it was agreed that an increase in overall stage length of 1% was permissible (the axial length of the turbine is ultimately limited by rotordynamic considerations). This allowed the return bend itself to be increased by approximately % in axial length. As with the rotor, a series of modifications to the return bend hub 4 Copyright 1998 by ASME
been considerably improved. A small recirculation is still evident on the shroud surface of the second bend. Further development would require additional increase of the radii of curvature of the bends, and thus in the axial length of the stage. Based on the CFD analysis, the total pressure loss coefficient is now.99. This is almost the same as the previous value, but this is because both the total pressure loss and the exit dynamic pressure have both been reduced, the former by a more smooth flow field and the latter by more effective diffusion. A better comparison is the absolute value of the total pressure loss, which is.1963 MPa at the design point, as compared with.33 MPa for the original design. This is an improvement of 16%. In its effect on stage performance, this improvement is worth approximately % of stage efficiency. Figure 9. ed runner velocity vectors and streamlines. Mid-span. and shroud lines was made to provide a smooth increase in passage area in order to diffuse the flow (mindful of the likely limits of stable diffusion, see Japikse (1984)), without introducing excessive curvature. The new return bend profile is shown together with the original for comparison in Fig. 1. The hub and shroud line curvatures are in Fig. 11, and Fig. 1 shows the variation of passage area with meridional distance. The streamlines and velocity vectors predicted by CFD analysis are shown in a meridional section in Fig. 13, and as projections of the midspan surface in Fig. 14. It is apparent that the flow field has 9.6 PERFORMANCE PREDICTION The performance prediction for the redesign stage at the design point is shown in Table 1 for comparison the results of the preliminary design As a result of redesigning the runner and return bend, the stage efficiency has been increased by eight points, to.894. Most of this improvement has been due to modifications to the rotor, but the changes to the return bend also make a significant contribution. VALIDATION The final stage of the redesign was to validate the analysis by testing a prototype turbine. Full-size testing with LNG was not feasible because of test facility limitations. Furthermore, there was no cryogenic torquemeter available to measure the shaft torque, which is necessary to assess the power and efficiency. Accordingly, a water test was undertaken. 7.9 R (in) 6.3 4.7 3.1 1.4..8 5.6 8.4 11.1 13.9 Z (in) Figure 1. Meridional section of redesigned turbine. 5 Copyright 1998 by ASME
. (a) Bend to strut 15 (a) Bend to strut Curvature (1/in) Curvature (1/in) Curvature (1/in) 1.5 1..5. -.5 4 - -4-6.5. -.5-1. shroud hub shroud hub (b) Strut shroud shroud hub hub (c) Bend to exit shroud hub shroud Passage area (in ) Passage area (in ) Passage area (in ) 1 5 15 1 5 1 8 6 4 (b) Strut (c) Bend to exit hub -1.5 4 6 8 1 Figure 11. Meridional curvature of return bend. Application of simple scaling rules lead to the conclusion that the prototype could be tested at full size in the available facility, and the speed and flow rate which are representative of actual operating conditions were determined. However, the water test Reynolds number is.8 1 7, as compared with 1.53 1 8 for LNG. This is a large difference and might be expected to have a significant influence on the efficiency. It was therefore decided to repeat the CFD analysis at the water test conditions and using the properties of water. This was done in two stages. Initially the CFD analysis was repeated using the properties of water at the full speed 4 6 8 1 Figure 1. Passage area of return bend. condition. A further analysis was then done at the scaled speed condition. The results of the former analysis of the rotor are shown in Figs. 15 and 16. Comparing these with the equivalent results for methane, Figs. 8 and 9, indicates that there is very little change to the flowfield in the meridional plane. In the blade-to-blade plane the effects of incidence are, if anything, slightly smaller, but there is slightly more underturning and the mass-averaged results in Table 1 show that the rotor efficiency is slightly worse as a result. Similar results at the reduced speed condition are shown in Figs. 17 and 18. By correct application of the scaling rules, the 6 Copyright 1998 by ASME
Figure 13. ed return bend vectors and streamlines. Mid-passage. Figure 14. ed return bend vectors and streamlines. Shroud surface. Figure 15. Water flow at full speed condition. Velocity vectors and streamlines. Mid-passage. flow rate is reduced simultaneously with the speed (see Table 1) and with it the meridional component of velocity. Therefore the correct incidence angle is preserved at the lower speed, as can be seen in Fig. 18. Once again, some underturning is apparent near the trailing edge and the rotor efficiency is reduced by another 1% relative to full speed with water. The stage efficiency is predicted to fall from.894 with methane, to.86 at full speed with water, and.836 at scaled speed with water. In general, the water flow predictions are quite similar to those of LNG. The principal hydrodynamic effect of the change in working fluid is to modify the Reynolds number, and while this undoubtedly has an effect on boundary layer growth and Figure 16. Water flow at full speed condition. Velocity vectors and streamlines. Mid-span. secondary flow generation, these changes do not have any major influence on the qualitative flow field. The water test conditions also give rise to rather more flow disturbance in the nozzle and the return bend, and this is reflected in the higher total pressure losses for these components. The head and efficiency of the turbine on water test are plotted in Fig. 19 as functions of flow rate at the scaled design speed. As measured, the maximum efficiency is 8%, as compared with the predicted value of 83.6%. However, the CFD analysis does not account for the leakage flow through the wear ring. The wear ring clearance was set larger for the water test than it would be in practice, because there is no lubrication 7 Copyright 1998 by ASME
Figure 17. Water flow at scaled speed condition. Velocity vectors and streamlines. Mid-passage. Figure 18. Water flow at scaled speed condition. Velocity vectors and streamlines. Mid-span. Head (m) 7 65 6 55 5 45 4 35 3 5 15 1 5-5 BEP Predicted Tested Efficiency % 83.6 8.36 Flow (m 3 /min) 7.574 7.574 Speed (RPM) 13 13 Head (m) 66.74 61.55 Efficiency Zero Leakage Flow Adjusted Predicted BEP Efficiency -1-4..5 1. 1.5..5 3. 3.5 4. 4.5 5. 5.5 6. 6.5 7. 7.5 8. Total Flow (m 3 /min) Figure 19. Water test results. Head 1 8 6 4 - Turbine efficiency (%) as there is with methane or LNG. The leakage flow was calculated from the seal ring flow area and head drop, and then the measured efficiency could be corrected to zero leakage. The increase in efficiency which results from adding the leakage flow is.4%. This brings the corrected maximum efficiency on test to 8.4%, which is just slightly below the predicted value. Figure 19 also shows that the corrected head is very close to the predicted value. CONCLUSIONS This project has demonstrated the value of CFD as a tool for both design and performance prediction of cryogenic rotating machines. The preliminary design of expander turbine was based on running a pump backwards, but this concept was predicted to have a poor efficiency of only 81.9%. Using CFD analysis, the nozzle, runner, and return bend were all redesigned 8 Copyright 1998 by ASME
to provide a smooth change of annulus area, to minimize streamline curvature, and to match correctly the various components. By this means, the predicted efficiency was increased to 89.4%. The redesign was validated by water testing. In spite of careful application of the scaling rules to the water test conditions, the Reynolds numbers of the water test condition and the actual design point using LNG are quite different, leading to expectations that the efficiency would change. A further CFD analysis at the water test condition lead to a predicted value of 83.6%. On test, very good agreement was achieved. The measured efficiency, when corrected for seal ring leakage, was 8.4% at the BEP. REFERENCES Casey M. V. 1983 A computational geometry for the blades and internal flow channels of centrifugal compressors, ASME Journal of Engineering for Power Vol. 15, pp. 88-95. Cengel Y. A., and Kimmel H 1977 Power recovery through thermodynamic expansion of liquid methane. Proceedings of the American Power Conference Vol. 59-II, Chicago. Japikse D. 1984 Turbomachinery Diffuser Design Technology, Concepts ETI, Inc., Wilder, Vermont. Japikse D., and Baines N. C. 1994 Introduction to Turbomachinery, Concepts ETI, Inc., Wilder, Vermont. 9 Copyright 1998 by ASME