Development of a high performance centrifugal compressor using a 3D inverse design technique
M. Zangeneh Mechanical Engineering Department University College London UK B. Nikpour Cummins Turbo Technologies Ltd Huddersfield, UK Hiroyoshi Watanabe 1 Advanced Design Technology Ltd London, UK
Abstract Diesel engines developed to meet future US and European emissions regulations place an unprecedented demand on compressor performance in terms of pressure ratio and flow range capability. Additionally the direct impact compressor efficiency has on engine fuel consumption means that the new generation of centrifugal compressors designed for Turbochargers will have to meet very demanding efficiency targets. Cummins Turbo Technologies have been exploring the application of inverse design techniques for improved compressor efficiency and pressure ratio. This paper describes the application of 3D inverse design method TURBOdesign-1 to a centrifugal compressor wheel. The baseline compressor is a production impeller with straight filament blades. In this paper the main features of the inverse design method and its design parameters are described. The impeller computed by the inverse design method has 3D geometry and has a pitchwise splitter location that is automatically optimized across the span. The structural and vibration characteristic of the new impeller are computed and compared with the conventional impeller by using 3D Finite Element Analysis. The flow through the inverse designed and conventional impellers are analysed by using a 3D CFD code both for impeller only and whole stage configurations. Furthermore comparison of the measured performance map of the inverse designed and conventional stages are presented. The results confirm that the inverse designed impeller results in improved stage efficiency of up to 3 points and improvements in pressure ratio of up to 4.5% while maintaining or slightly improving the compressor’s stable operating range.
1
On secondment from Ebara Corporation Japan, Fluid Ma chinery & Systems Company 11-1 Haneda Asahi-cho, Ohta-ku, Toky o 144-8510, Japan. 135
10.1243/17547164C0012010010 Notation BSFC CFD
Break Specific Fuel Consumption Computational Fluid Dynamics
∂r V θ / ∂m
Blade loading
m
r V θ T0,in T0,out ΠC ηcomp κ
Percentage meridional distance Pitchwise average swirl velocity Compressor inlet temperature (stagnation) Compressor outlet temperature (stagnation) Compressor pressure ratio (total/ total) Compressor efficiency Ratio of specific heat capacities
Introduction One of the key challenges for the automotive industry is improvement of engine fuel economy with the aim of both reducing global CO2 emissions and making best use of the finite oil resources. It is well understood how a Turbocharged diesel engine’s fuel consumption is directly affected by the turbocharger efficiency through the engine’s pumping cycle work. The more efficient the turbomachinery, the lower the work done by the engine during the pumping cycle, resulting in better engine BSFC. Turbocharger manufacturers have undertaken considerable amount of research and development combining application of computational fluid dynamics and experimental measurements to improve the aerodynamic efficiency of centrifugal compressors used in turbochargers. This has been done under the constraints of wide operating range, high reliability and structural integrity. Turbocharger impellers have to operate at relatively high pressure ratios, and hence tip speeds, and until recently most turbocharger manufacturers were constraining their designs to blades with straight filaments. However, it has become clear recently that further improvement in turbocharger performance is difficult to achieve by using straight filament geometry for the impeller, and some manufacturers have considered the development of 3D geometries by using conventional design methods in which the blade geometry is modified iteratively by using some CAD representation and then analysed by CFD. The ‘direct’ design approach based on iterative modification of the blade shape results in a time consuming design process, especially if the various multi-objective requirements of the impeller design for turbocharger applications are to be met. These difficulties are compounded for impellers with splitter blades, which are commonly used in turbocharger applications. In conventional design practice, it is customary to use the same blade profile on the splitter blade as on the main blades with splitter camber line being placed at midpitch between the two main blades. There is, however, considerable evidence that this practice leads to poor performance of the splitter blade as a result of the mismatch between the splitter leading edge blade angle and the local flow angles. In recent years there have been some attempts by designers to modify the splitter leading edge slightly to improve the matching of the splitter to the local flow. An example of this is the work of Drtina et al [1], who showed that by modifying the pitchwise location of the leading edge of a splitter blade they could improve the pressure recovery of a vaned diffuser. However, in the design of 3D blades it is rather difficult to arrive at an optimum pitchwise distribution of the splitter across the span. 136
10.1243/17547164C0012010010
An alternative approach for the design of centrifugal impellers with 3D geometry is to use a 3D inverse design method such as TURBOdesign-1[2]. In this inverse design method the blade geometry is computed for a specified distribution of blade loading ( ∂r V θ / ∂m ), which is the meridional derivative of the tangentially mean swirl velocity and is directly related to the blade bound circulation 2π r V θ . In this method, in addition to the blade loading the normal thickness distribution is specified. The basic theory behind this approach is also presented in Zangeneh [3] and its extension to design of blades with splitters is presented in Zangeneh [4]. The inverse design method TURBOdesign-1 allows the designer to design the blade geometry on the full and splitter blades subject to a specified distribution of blade loading and therefore automatically optimizes the pitchwise location of the splitter blade across the span. The method has already been applied to the design of industrial centrifugal compressor impellers, where simple design guide-lines have been developed for suppression of secondary flows and the resulting jet/wake flow effects at the impeller exit (see Zangeneh et al, [5] and [6]). This method has also been applied to the design of compact high performance vaned diffusers for centrifugal compressors (see Zangeneh et al, [7]) and industrial compressors with splitters ( see [8] and [9]. In this paper, this inverse design approach is applied to the design of a medium sized turbocharger for heavy duty diesel engine application. The baseline design is a production compressor impeller which has optimum geometry based on straight filament design. DESIGN OBJECTIVES AND INVERSE DESIGN OF THE IMPELLER The baseline impeller has 7 full blades and 7 splitters with nominal design shaft speed of 101,721 RPM. The nominal pressure ratio at the design point is 3.5. The main objectives of redesign were to improve the impeller efficiency and pressure ratio without any adverse effects on the impeller stable operating range. In the redesign of the new impeller no attempt is made to re-design the meridional geometry and the same meridional geometry as the conventional impeller is used in the design. The meridional geometry used is shown in Fig.1. The splitter leading edge has a cut back configuration typical of those used for high speed applications. In addition to meridional shape the same normal blade thickness, as the conventional impeller was used in the design of the new impeller.
Fig 1 – Meridional Geometry of the Impeller
137
10.1243/17547164C0012010010
Fig. 2 – Specified Blade loading Once the meridional geometry is fixed, the main choice of design parameter in TURBOdesign-1 which controls the flow field in the impeller is the blade loading distribution. As noted earlier the blade loading is the meridional derivative of the circumferentially averaged swirl velocity and is directly related to the jump in pressure across the blade. The integral of blade loading curve with respect to the meridional distance along each streamline is equal to the change in rVθ between leading and trailing edge at that streamline and hence is equal to the specific work on that section. In TURBOdesign-1, the blade loading on the full and splitter blades can be specified independently as shown in Fig. 2. The blade loading specified has an aft-loaded distribution at both hub and shroud because in this impeller, the effect of inducer shock dominates losses in the impeller. Also relatively large incidence has been specified in the design to ensure the required choke margin. An important input specification in TURBOdesign-1 is the r V θ ratio which is defined as:
r V ratio = (∆r V θ ) splitter /(∆r V θ )total θ where
( ∆r V θ ) splitter is the change in specific work on the splitter blade and
( ∆r V θ ) total is the change in specific work in the overall impeller. Normally the r V θ ratio is based on the ratio of the meridional length of the splitter (at the shroud) over the overall meridional length. This type of definition of r V θ ratio will essentially ensure similar levels of maximum loading between the full and splitter blades. However, it is equally possible to unload the splitter at the expense of full blades or vice versa. This provides additional flexibility in the design of impellers with splitter blades. In this case the r V θ ratio was varied from a value similar to the meridional ratio at the shroud to value slightly lower than the meridional ratio at the hub. The final important input specification in TURBOdesign-1 is the stacking condition, where the spanwise variation of wrap angles at one quasi-orthogonal are specified. TURBOdesign-1 uses the wrap angles specified on the stacking quasi-orthogonal as the initial condition of the hyperbolic partial differential equation used to compute the blade wrap angles. In this case the blades were stacked at the trailing edge plane with a wrap angle stacked in such a way as the shroud leads the hub in the direction of rotation, i.e. similar rake to conventional impeller. The resulting geometry computed by TURBOdesign-1 is shown in Fig. 3 and compared with the
138
10.1243/17547164C0012010010 conventional impeller with straight filaments. The inverse designed geometry is three dimensional and characterised by a bow near the leading edge.
Fig. 3 Geometry of original and inverse designed impellers Structural Analysis of the New Design Structural analysis of the new design was carried out using Ansys Workbench FEA package. Figure 4 shows the 3D model constructed for FEA analysis. This model contains over 400,000 elements capturing all geometry features including, blade root radii and leading / trailing edge treatment. This model was used for both vibration and stress analysis. One of the features of inverse design approach (discussed earlier) is the non-radial nature of the blade filaments, especially in the inducer region of the impeller. This type of inducer design was found to be advantageous for blade natural frequency when compared to standard design. The inverse design blades were found to be ‘stiffer’ in the inducer with the 1 st mode (inducer flap) blade vibration frequency in the region of about 20% higher than conventional design. The stress distribution patterns for the inverse design impeller are compared to the conventional design geometry in figure 5. It is observed that the location of the maximum principal stress has moved from the suction surface root radii near the trailing edge (conventional design) to the inducer of the main blade (inverse design). This pattern is another characteristic of this type of design due to the non radial nature of the blade filaments. In this work it was found that when designing both types of impeller to the same stress limits, the durability of the inverse design is comparable to the standard design.
Fig. 4 Mesh used in structural analysis 139
10.1243/17547164C0012010010
Fig. 5 Comparison of Maximum Principal stresses
Comparison of the Predicted Flow Field Initially CFD computations were performed on the impeller only using a single blade channel configuration. The results of these computations were used to compare the differences in flow fields between the two impellers qualitatively. The comparison was done with the same computational mesh and the same level of tip clearance in both cases. In Fig. 6, comparison of the relative Mach number near the blade tip between the two cases at near peak efficiency point and 100% speed are shown. The results confirm a reduction in the shock strength in the inducer of the inverse designed impeller as well as a change in trajectory of the clearance vortex in the inducer. As a result the clearance vortex no longer intersects the leading edge of splitter in the inverse design impeller as it does in the original impeller
(a) Original Design
b) Inverse design
Fig. 6 Comparison of Relative Mach numbers near the tip
140
10.1243/17547164C0012010010
In addition to impeller only computations, full stage 3D analysis of the inverse and standard impellers were carried out using the commercial CFD code CFX 10. Figure 7, shows the computational mesh used, with pressure variation contours superimposed for one running condition. Structured computational mesh was used for the two impellers and unstructured mesh was used for the vaneless diffuser and the volute. The same vaneless diffuser and volute geometry was used for both impellers. The models contained around 7 million computational cells and were run on a 64 node parallel computational cluster (each operating condition taking around 1 hour in computational time). Both conventional and inverse design models were run at three speed lines and for a range of flow rates representing the broad operational range of the compressor. Results of the full stage CFD analysis are summarised in Figure 8. The results indicate that the inverse designed stage is showing better efficiency and pressure ratio throughout the operational range except at one point near choke and at 100% speed. The drop in predicted efficiency of the inverse designed stage at near choke conditions of 100% speed was not actually observed in actual test results (next section).
Figure 7 – Full stage CFD computational mesh
141
10.1243/17547164C0012010010 Full stage CFD analysis data
y c n 105% e i c 100% i f f 95% E 90% d 85% e s i l 80% a 75% m r 70% o 20 N
40
60
80
100
80
100
Normalised Flow % 5.00
Standard des ign 100% speed
4.50
Inverse design 100% speed
4.00
Standard des ign 80% spee d
3.50
Inverse design 80% spee d
o i t a 3.00 R e r u s 2.50 s e r P
Standard des ign 60% spee d Inverse design 60% spee d
2.00
1.50
1.00 20
40
60
Normalised Flow %
Figure 8 – Results of the full stage CFD analysis
Measurement of Stage Performance Production intent prototype of the inverse designed impeller was manufactured. The inverse design stage used the same vaneless diffuser and volute as the conventional stage (current production hardware). The inverse design stage was then tested in the same test facility as the conventional stage to obtain its performance map. Figure 9, shows the comparison of the performance of the new stage with the conventional production stage. The performance map overplot shows compressor pressure ratio and normalised efficiency plotted against flow rate. Efficiency and flow rate are normalised with respect to the maximum values for the conventional design hardware. We can see that the inverse designed prototype is delivering up to 4.5% better pressure ratio and up to 2.5 to 3% higher efficiency. The efficiency measured in both cases was the total to total efficiency defined as:
⎡ ⎛ ⎜⎝ T o,in * ⎢Πc ⎣ =
η comp
k - 1 ⎞
⎟
k ⎠
⎤
- 1⎥
T o, out − T o,in
⎦
(1)
where T0,in and T0,out are the stagnation temperatures at inlet and exit of the test facilities and Pc is the compressor pressure ratio. The inverse designed stage 142
10.1243/17547164C0012010010 delivers approximately 2% less choke flow to the original stage, which is less than half of the difference in the computed data shown in Fig. 8. Also it seems to have a slightly larger stable operating range at 100% speed as opposed to the conventional stage. However, at part load there is slight shift in the compressor characteristic to higher flow rates as compared to the original stage. The difference observed between the minimum (surge) flow rates of the test data (figure 9) and the computational results of figure 8, is due to the test hardware utilizing super map width enhancement (S-MWE) inducer slots (see Nikpour [10] for details of s-mwe). S-mwe was not modelled in the computational cases because, the added complexity, would have resulted in a significant increase to the mesh size and the computational time.
Fig. 9 Comparison of the measure performance map of inverse designed stage ( Red) and conventional stage ( blue).
Figure 10 is a contour plot showing the measured efficiency improvement of the new inverse design stage relative to the conventional compressor. This plot shows the efficiency improvement is about 0.5% at low flows (near surge) increasing to 2.5 to 3% at higher pressure ratio regions of the performance map. It is also observed that the overall trend of efficiency and pressure ratio improvements are very similar to the full stage CFD results (previous section), except at the 100% 143
10.1243/17547164C0012010010 speed. The reasons for the discrepancy at 100% speed between measurements and full stage CFD is currently being investigated.
Fig. 10 Comparison of the improvement in efficiency of inverse design stage versus the conventional stage over the performance map
Conclusions In this paper the application of 3D inverse design method TURBOdesign-1 to design a centrifugal compressor for heavy duty diesel engine turbocharger application is described. In this method the full and splitter blades are computed for a specified distribution of blade loading and normal thickness. The results of the study confirm that by using an aft-loaded blade loading distribution on the shroud and careful control of the work ratio between the full and splitter blades the shock strength in inducer can be reduced. As a result significant improvement in measured stage performance of the compressor of up to 3 point in efficiency and 4.5% in pressure ratio can be obtained without any effective reduction in stable operating range. The inverse designed impeller has 3D blade shape. 3D Finite Element Analysis was used to investigate the effect of the 3D shape on centrifugal stresses and vibration characteristics of the impeller. The results confirm that 3D shape of the inverse
144
10.1243/17547164C0012010010 designed impeller only changes the location of maximum principal stress on the blade from the root radii region of the suction surface to the mid-span location on the inducer. But maximum values of the stress are the same in the inverse designed impeller as compared to the conventional straight filament designs. Furthermore vibration analysis confirms that the inverse design impeller is stiffer with a 1st mode frequency that is 20% higher than the corresponding conventional straight filament impeller. The results of this study confirm that the 3D inverse design method offers an efficient alternative for designing high performance centrifugal compressors which help to improve turbocharger efficiency and boost as well BSFC of heavy duty diesel engines. Acknowledgements The authors wish to thank Cummins Turbo Technologies Ltd for their support of this work and their permission to publish this paper. © Authors 2010 References
1.
Drtina, P., Dalbert, P. , Rutti, K. and Schachenmann, A., (1993), ‘ Optimisation of a diffuser with splitter by numerical simulation’, ASME paper 93-GT-110. 2. Turobodesign-1 version 3.1, Advanced Design T echnology Ltd, 2009. 3. Zangeneh, M. , 1991, " A compressible three dimensional Blade design method for radial and mixed flow turbomachinery blades", Int J. Numerical Methods in Fluids, Vol. 13, pp. 599-624. 4. Zangeneh, M., 1998, “On 3D inverse design of centrifugal compressor impellers with splitter blades”, ASME Paper 98-GT-507. 5. Zangeneh, M., Goto, A. and Harada, H., 1998, “On the design criteria for suppression of secondary flows in centrifugal and mixed flow impellers”, ASME Journal of Turbomachinery , Vol 120, pp. 723-735. 6. Zangeneh, M., Goto, A. and Harada, H., 1999, “On the role of threedimensional inverse design methods in turbomachinery shape optimization”, Proceedings of IMECHE, J. of Mech. Eng. S ci., Part C, Vol. 213, No. C1, pp. 27-42. 7. Zangeneh, , Vogt, D., Roduner, C., 2002, “Improving a vaned diffuser for a given centrifugal impeller by 3d inverse design”, ASME Paper GT-200230621 8. Zangeneh, M., M., Schleer, M., Ploger, F., Hong, S.S., Roduner, C., Ribi, B. and Abhari, R, 2004, “Investigation of an inversely designed centrifugal compressor stage part 1: Design and numerical verification” Trans. Of ASME, J. of Turbomachinery, Vol 126, pp. 73-81 (2004). 9. Schleer, M., Hong, S.S., Zangeneh, M., Ploger, F., Roduner, C., Ribi, B. and Abhari, R “Investigation of an inversely designed centrifugal compressor stage part 2: experimental investigations”, Trans. Of ASME, J. of Turbomachinery, Vol 126, pp. 82-90 (2004). 10. Nikpour, B., “Turbocharger compressor flow range improvement for future heavy duty Diesel Engines.”, Thermo-and fluid dynamic processes in Diesel Engines conference 2004, ISBN: 84-9705-621-3, pp 125-143
145
