431

Turbine performance studies for automotive turbochargers. Part 1: steady analysis

A. Romagnoli and R.F. Martinez-Botas*

Department of Mechanical Engineering Imperial College London

SW7 2AZ Exhibition Road London, UK

*Corresponding author

S. Rajoo Dept. Automotive Engineering

Faculty of Mechanical Engineering Universiti Teknologi Malaysia

81310 Johor - Malaysia SYNOPSIS This paper presents the results from an experimental investigation

conducted on different turbine designs for an automotive turbocharger. The design progression was based on a commercial nozzleless unit that was modified into a variable geometry single and twin entry turbine. The main geometrical parameters were kept constant for all the configurations and the turbine was tested under steady and pulsating flow conditions (pulsating findings are presented in an accompanying paper).

A significant depreciation in efficiency was measured between the single and twin entry configuration due to the mixing effects. The nozzleless unit provides the best compromise in terms of performance at different speeds.

The twin entry turbine was also tested under partial and unequal admissions. Based on the test results, a method to determine the swallowing capacity under partial admission given the full admission map is presented. The test results also showed that the turbine swallowing capacity under unequal admission is linked to the full admission case.

NOMENCLATURE

A Area, [m2] C Absolute velocity, [m/s] P Pressure, [Pa] SE Single entry

10.1243/17547164C0012010034

432

T Temperature, [K] TE Twin entry VR Velocity ratio

Mass flow rate, [Kg/s] r Radius, [m] Absolute flow angle, [deg] Density, [Kg/m3] Subscript Azimuth angle 0 Total condition 1 Inner limb/Inlet 2 Outer limb/Outlet atm Atmospheric ex Exit is Isentropic pa Partial admission r Rotor te Twin entry un Unequal admission

INTRODUCTION Increasing limitations on exhaust emissions and the need to reduce fuel consumption have encouraged extensive use of turbochargers in the automotive sector. The turbocharger turbines comprise two main elements: the wheel and the stator (or volute). The wheel can be either radial or mixed flow and its function is to extract work from the exhaust gases. The main function of the stator is to accelerate and distribute uniformly the flow around the wheel. The most common configurations for the stator are nozzleless, nozzled and twin entry. In a nozzleless turbine, the volute is solely responsible for providing an adequate swirl and consequently set the inlet flow angle for the rotor. In a nozzled turbine, the volute is complimented with the downstream nozzle ring in setting the flow characteristics into the rotor. In this case, the volute is designed to provide uniform flow into the nozzle. Additionally, in a variable geometry stator the volute-nozzle coupling provides additional flexibility in adapting to the incoming flow [1].

10.1243/17547164C0012010034

433

Twin entry turbines are adopted for using the energy of pulsating exhaust gases. Two banks of exhaust manifolds feed each entry of the turbine so that the windmill is minimized as the mass flow drops to zero. Spence et al. [2] carried out a performance investigation on an equivalent swallowing capacity basis for three different nozzleless and nozzled stators. The nozzleless configuration revealed a better performance than the nozzled one. This was attributed to different levels of roughness between the stators. In the case by Baines and Lavy [3] the highest efficiency was measured for the nozzled configuration. Capobianco and Gambarotta [4] instead compared a single entry to a twin entry turbine and found that, under full admission, the efficiency of the twin entry turbine under is about 7% less.

The study reported here shows the performance for a design progression from single to twin turbine. An evaluation of the turbine efficiency and swallowing capacity was carried out on the basis of the experimental results.

EXPERIMENTAL FACILITY The automotive research group at Imperial College has actively been involved in the development and understanding of pulsed flow turbines since the 1980's. The turbocharger test facility was originally developed by Dale and Watson [5] and extended by Baines et al. [6][7]. The experimental facility available at Imperial College London is a simulated reciprocating engine test bed for turbocharger research. The facility can perform steady state testing in single and twin entry turbines; it is also capable of carrying out unsteady tests [8][9]. The recent installation of an eddy current dynamometer enables turbine testing within a large velocity ratio range [10][7]. The test-rig is supplied by screw-type compressors, capable to delivering air up to 1.2 kg/s mass flow rate at a maximum pressure of 5 bars (absolute). For unsteady testing, an air pulse generator is employed. It simulates experimentally the engine exhaust gas pulsations by means of a set of counter rotating plates with appropriately designed cut-outs. A variable speed D.C. motor controls the rotating frequency of the chopper plates; hence the frequency of the pulsation can be set [1].

10.1243/17547164C0012010034

434

TURBINE VOLUTE DESIGN The turbine used in the current study is a mixed flow variable geometry designed at Imperial College [1]. The design was based on a commercial nozzleless unit (HOLSET H3B) and it aims for increased flexibility in the operating envelope of the turbine. The turbine volute was manufactured in two halves allowing a nozzle-ring and the divider to be inserted. Such an arrangement of the turbine volute allows change between four turbine configurations: single entry nozzleless, single entry nozzled, twin entry nozzled and twin entry nozzleless. The main geometrical parameters of the turbine arrangements remain the same.

where S is a constant (1)

(2) The volute design was carried out using the well established mean

line analysis method [11][12]. The main assumptions for this approach are: free vortex conditions in the volute and uniform flow distribution around the volute periphery. This approach results in two basic equations (1) & (2). Assuming incompressible flow and rearranging Eq. (1) and (2) one obtains the A/r relation:

The equation above shows that the two critical parameters to be

taken into account in the design of the volute are the cross sectional area (A) and the correspondent centroid radius (r). In order to distribute the mass uniformly around the circumference of the rotor, the ratio between the area and the radius must be a linear function of the azimuth angle. For a given mass flow rate and density, the radial component of the velocity going into the rotor is then fixed by the cross sectional area at the rotor inlet. This means that the volute outlet angle is given by:

10.1243/17547164C0012010034

435

In order to determine the volute exit flow angle, a simple analysis was carried out using Eq. (1) to (4). A similar analysis of the HOLSET 3HB turbine provided an exit flow angle of 68 with respect to the radial direction. Hence a target range of 67 to 72 enveloping the HOLSET H3B value was chosen and the main geometrical parameters (area and radius) calculated accordingly.

A similar approach was followed in the design of the divider for the twin entry turbine. A solid modelling analysis was then carried out in order to determine the best compromise between area available to the flow and strength of the material. Amongst all the possible solutions proposed, a tapered shape divider was chosen. The final design of the single entry turbine together with the turbine including the divider is shown in Figures 1 & 2.

Figure 1: Whole turbine stage Figure 2: Divider fitted in the in single entry turbine

10.1243/17547164C0012010034

436

STEADY TURBINE PERFORMANCE This paper reports performance results from two single entry

configurations (nozzled and nozzleless volutes) previously characterised at Imperial College [1][9][13]. These results are complemented with data recently acquired for a variable geometry twin entry mixed flow turbine. In this manner, a comprehensive comparison of different configurations can be carried out. The static efficiency (given as the ratio between the actual power and the isentropic power) and the pseudo-dimensional mass flow parameter have been plotted against the velocity ratio and the pressure ratio respectively. The definition of these parameters can be found below:

The turbines performance was assessed on an equivalent geometry

basis. The design progression of the volute was aimed to maintain the A/r, the exit flow angle and the shape of the cross-section similar to the base line (HOLSET H3B). The turbine wheel used for all the tests is of a mixed flow nature previously designed at Imperial College by Abidat [14]. For consistency with previously reported results, the wheel is referred to as rotor A. The main geometrical parameters are given in Table 1 and more details can be found in available literature [15][16].

Table 1 One-dimensional analysis of the advanced rotor A

Rotor Type A Inlet mean diameter (mm) 83.58 Number of blades 12

Exducer hub diameter (mm) 27.07 Exit mean blade angle -52

Inlet blade height (mm) 18.0 Inlet blade angle 20

10.1243/17547164C0012010034

437

Figure 3: SE at 80% speed-Nozzled (60) vs. Nozzleless: a- ts vs. VR & b-MFP vs. PR

10.1243/17547164C0012010034

438

Figure 4: SE at 50% speed-Nozzled (60) vs. Nozzleless: a- ts vs. VR & b-MFP vs. PR

Figures 3 & 4 show the performance parameters for a nozzled

turbine at 600 vane angle (corresponding to the optimum vane angle for this turbine) and a nozzleless equivalent (base line). The figures show results for 50% and 80% of the design speed testing which corresponds to 32000rpm and 48000rpm respectively. For the nozzleless mixed flow turbine, the peak total to static isentropic

10.1243/17547164C0012010034

439

efficiency was 0.77 displaying a slight drop of 3 percentage points at low velocity ratios (50%). Overall, the turbine exhibits features common to mixed flow turbines [17]. The efficiency curves remains fairly flat for velocity ratios near to peak. This is even more evident at 50% speed where the efficiency drop, even at higher velocity ratios, is

Figure 5: SE/TE at 80% speed-Different configurations: a- ts vs. VR

& b-MFP vs. PR

10.1243/17547164C0012010034

440

small. In the nozzle configuration, the peak efficiency was 0.81 at 80% speed (vane angle of 60). This represents an improvement of 4 percentage points at all low velocity ratios when compared to the nozzleless case. The same improvement is not visible at the low speed condition (50% speed) for which similar efficiencies were found.

Figure 6: SE/TE at 50% speed-Different configurations: a- ts vs. VR

& b-Mass vs. PR

10.1243/17547164C0012010034

441

Interestingly, a significant shift of the efficiency curve towards low velocity ratio can be seen in this case: the peak efficiency velocity ratio shifts from 0.70 to 0.64. Such a shift is significant for energy extraction, as in the real pulsating condition of the turbine high efficiency at high pressure ratio (low velocity ratio) is desired. One would thus presume that such a performance curve would lead to greater pulse flow performance [15][16][18]. At the present no data are available on support of this assumption. However a preliminary comparison between the pulsating flow performance for single entry nozzled and nozzleless turbine is provided in Part Two of this paper [19].

In order to test a twin entry turbine, a divider was inserted within the volute. The design took care to find the best compromise between the strength of the divider and the area available to the flow. The main geometrical parameters are the same as those of the nozzled turbine except for the A/r that was reduced by 6%. Twin entry turbines are usually adopted to isolate the gas flow from each separate bank of manifolds. The turbine works under unequal and/or partial admission conditions for most of its operation; consequently, full admission does not replicate the working conditions of the turbine under normal engine operating conditions. Nevertheless, turbine maps are usually available only for full admission conditions; therefore it seems logical to report tests under full admission conditions for the twin entry geometry.

Figures 5 & 6 report the turbine efficiency under full admission at 80% and 50% turbine speed. The presence of the divider has a small detrimental effect on turbine efficiency; the peak efficiency was found to be 0.79, which is slightly lower than that measured in single entry. At high velocity ratios the turbine performance shows an improvement of few percentage points with respect to the single entry configuration at both 50% and 80% speeds.

In order to characterize the effects of the vane angle, a set of tests were carried out in the 70 and 40 vane angle range for both types of entries (single and twin). The results are reported in Figures 7 & 8. A slightly lower swallowing capacity was measured for the twin entry turbine in comparison to the single entry case for both 70 and 40 vane angle. This can be explained as an effect of the divider within the volute that reduces the area available to the flow. At 70 vane angle, no significant difference in the peak efficiency value between the single and twin entry configuration was measured. As the velocity

10.1243/17547164C0012010034

442

ratio increases, the twin entry turbine performs better than the single entry. The same conclusions can be reached for the 60 vane angle (refer to Figure 5 & 6) but at 40 the results are different, showing a

Figure 7: SE/TE at 80% speed - 70 vane angle: a- ts vs. VR & b-MFP vs. PR

5% efficiency drop over the whole range of velocity ratios. The mixing effect between the flows out of two limbs is considerable and should be taken into account in design-phase.

10.1243/17547164C0012010034

443

Figure 8: SE/TE at 80% speed - 40 vane angle: a- ts vs. VR & b-MFP vs. PR

Figure 9 reports the single and twin entry performance parameters

for 40, 60 and 70 vane angles at 80% speed. The single entry configuration exhibits the best overall performance with a peak of 81% (60 vane angle 80% speed). At 40 vane angle an efficiency drop of almost 25% and 20% was measured for the twin and single

10.1243/17547164C0012010034

444

entry configuration in respect to the peak efficiency point (60 vane angle 80% speed). At 70 the peak efficiency for both the single and twin entry turbine shows a shift in the velocity ratio down to 0.57. On the mass flow side, the swallowing capacity at 70 is much lower than that measured at 60 and 40 vane angle and, for mid range pressure ratios, it is almost half than that measured at 40.

Figure 9: SE/TE at 80% speed - Different vane angles: a- ts vs. VR &

b-MFP vs. PR

10.1243/17547164C0012010034

445

Figure 10: SE/TE at 80% speed-Same swallowing capacity: a- ts vs. VR b-MFP vs. PR

10.1243/17547164C0012010034

446

Figure 11: TE partial admission 80% speed - 60 Vane angle: a- ts vs. VR b-MFP vs. PR

10.1243/17547164C0012010034

447

Figure 12: TE partial admission 50% speed - 60 Vane angle: a- ts vs. VR b-MFP vs. PR

In order to complete the analysis an assessment of turbine

efficiency on an equivalent swallowing capacity basis was also carried at 80% speed. In order to let more mass flow to go through, the vanes were open at 40. As the vanes open, they produce a deviation from the optimum incidence condition leading to an increase in losses. This is clearly visible in Figure 10 where there is a detrimental effect of vane opening to meet the same swallowing capacity.

10.1243/17547164C0012010034

448

As it was previously stated, the twin entry turbocharger always exhibits an imbalance of flow conditions between the two entries, caused by the manifold arrangement; the twin entry turbine was also tested under partial admission conditions for the optimum (60) vane angle. The importance of partial admission conditions becomes apparent when evaluating aerodynamic losses and translating these into the real pulsating operation of the turbocharger.

Figures 11 & 12 report the turbine efficiency at 80% and 50% speed under full and partial admission condition; where partial is meant to be the condition in which one entry does not flow at all while the other flows (labelled as outer open in the figure). These tests are repeated by reversing entry that flows (labelled as inner open on the figures), one can thus see the effect of the chosen entry to flow. Independent the entry that has flow, a large fall in efficiency is measured, the drop is as large as 20 percentage points at 80% speed and 24 percentage points at 50% speed. The peak efficiency point for either inner or outer limb fully open is the same at both speeds, which is 61% and 53% for 80% and 50% speed respectively. Nevertheless the inner limb seems to perform better than the outer limb; note the mass flow in both limbs is the same. This is not a new finding, it is the consequence of the different paths taken by the flow for a given shroud curvature.

Figure 13: Mass flow prediction under partial admission - 50% speed - 60 vane angle

10.1243/17547164C0012010034

449

Figure 14: Mass flow prediction under partial admission - 80% speed - 60 vane angle

As previously mentioned, turbine maps are usually provided under full admission conditions and so far no method has been proposed for evaluating the mass flow parameter under partial admission given a full admission mass flow curve. Here a simple method is proposed. Figures 13 & 14 report the swallowing capacity for the twin entry turbine under full admission conditions for both 50% and 80% speeds. It can be seen that by simply halving the mass flow measured under full admission, the corresponding partial admission is not reached. In fact by following this simple approach, one would be treating the turbine as a single entry turbine with half the passage area and thus, it does not take account for the interaction existing between the two limbs. Furthermore, even though in partial-admission conditions no air flows at the inlet of one of the entries, stagnant air at atmospheric pressure is still present within in the non-flow limb and leakages into that entry from the flowing side can occur. The general expression for the mass flow parameter is given by Eq. (7). For a twin entry turbine the mass flow parameter is calculated as a mass-averaged mass flow parameter where an area averaged pressure is used:

10.1243/17547164C0012010034

450

By blanking off entry 2 (thus = 0) the total mass flow rate

= , leading to:

The total pressure P02 is near atmospheric as there is no dynamic head to be taken into account and neglecting the centrifugal head imposed by the rotor (which is small in any case). The mass flow parameter calculated by mean of Eq. (10) is given in Figures 13 & 14. The mass flow prediction from the full entry maps is much improved for both speeds. At high pressure ratios the newly predicted mass flow matches that measured experimentally while at low pressure ratios the simple approach are less accurate. This can be explained if we consider that at low pressure ratios, the total pressure in the flow-limb is similar to atmospheric and hence the denominator of Eq. (10) does not change significantly when the term P02/2 is added.

The above discussion centred on what is commonly called partial admission where one port is completely closed; here we report results from the partial and full flow cased. These are labelled as unequal admission cases and one of the questions relates to how the steady tests under unequal admission should be carried out. For the purpose of this research, the unequal admission condition was obtained by keeping constant the pressure ratio in one limb and let the other change in order to match the selected speed (seen in the line labels of Figure 15 as Outer followed by the pressure ration kept constant in that entry). Other approaches are possible such as reported by Capobianco and Gambarotta [4] that performed a set of unequal cases by keeping the mass flow ratio between the two limbs constant.

10.1243/17547164C0012010034

451

Figure 15: Twin entry - Mass flow in unequal admission - 80% speed - 60 vane angle

10.1243/17547164C0012010034

452

Figure 16: Twin entry - Mass flow in unequal admission - 50% speed

- 60 vane angle

Here the tests were carried at 80% and 50% speed and the results shown in Figures 15 & 16. The trends are similar to those measured under full admission; this is more evident as the pressure ratio and the turbine speed increase. For instance at 80% and pressure ratio 1.9 the mass flow curve follows very closely the full admission curve. As already mentioned, the unequal admission condition is somewhat difficult to analyse and published research does not provide much

10.1243/17547164C0012010034

453

insight. Given the similarity between the mass flow curves, we tried to understand whether or not a common correlation exists between the full and unequal admission condition. Given the mass flow curves in Figures 15 & 16, in order to proceed with the analysis, the ratio between the unequal admission flow and that in the flow-limb was calculated. The same was done for the pressure ratios.

Figure 17: Swallowing capacity under unequal admission Different test conditions

The results of such an approach are shown in Figure 17, this figure

includes all the test conditions shown in Figures 15 & 16. It was found that all the points collapse into a single curve following an exponential trend. This suggests that a unique correlation links the mass flows between the two limbs. In fact if we develop the expressions for the Expansion Ratio and the Mass Flow Parameter Ratio we obtain Eq. (11) & (12). These equations show that the ratio of Expansion Ratio corresponds to the pressure ratio between the inner and outer limbs and that Mass Flow Parameter Ratio is a function of both K and the mass flow ratio in both limbs.

10.1243/17547164C0012010034

454

Figure 17 shows that Eq. (12) follows an exponential trend and hence for a given pressure ratio and mass flow in one limb, the mass flow in the other limb is uniquely defined. In other words, since the full admission maps are usually available, given the mass flow and the pressure ratio in one limb, it is possible to define the mass flow and pressure ratio in the other limb such that their combination corresponds to any point of the full admission curve. This could be significant as it could contribute to developing a tool to generate unequal admission maps from a full admission map. However, it is still necessary to carry out limited testing to ascertain the constants A and B. The collection of a sufficient large database will eventually indicate if there are some commonalities that can be taken into account in order to find what are the parameters affecting A and B.

CONCLUSIONS The current paper discusses the progressive evaluation of different turbine designs from nozzleless to twin entry. The base turbine design is a commercial nozzleless unit, from which the single entry nozzled turbine was designed and progressively redesigned to twin entry. The evaluation shown in the paper covers the steady flow conditions with discussion on full, partial and unequal admissions.

The nozzled single entry turbine (a with a 60 degree vane angel) was found to perform better than the corresponding nozzleless configuration. In twin entry mode the insertion of the divider was found not to be detrimental to the overall performance; it showed improvement at high velocity ratios.

As the vanes open (40), a large drop in efficiency occurred between the single and twin entry turbine. The mixing effects were evaluated to account for as much as 5% in efficiency loss over the whole range of velocity ratios. On the other hand, as the vanes close (70), no difference in efficiency was measured between the single and twin entry configurations. The peak efficiency is slightly lower than that measured at 60 vane angle with a significant shift in the velocity ratio from 0.67 to 0.57.

10.1243/17547164C0012010034

455

The efficiency of the turbines at equivalent swallowing capacity under steady flow was found to be decreasing from nozzleless to twin entry design configurations. At velocity ratio of 0.67, the single entry nozzled turbine efficiency is 16% lower than the nozzleless, meanwhile the twin entry turbine is 21% lower than the single entry.

The paper also presents a method to calculate the partial admission swallowing capacity of a twin entry turbine from the full admission map. This is particularly beneficial as an engine developer will only generally have the full admission map in a twin entry turbine, and it is accepted that the turbine operates in partial admission in most cases. The test results also revealed that the mass flows between limbs are correlated by an exponential manner.

ACKNOWLEDGEMENTS

The authors would like to acknowledge Ricardo plc, Ford Motor Company Ltd and University of Brighton. This consortium along with Imperial College are part of funded program (TSB) named VERTIGO (Virtual Emission Research Tools and Integration).

Authors 2010 REFERENCES 1 Srithar, R., Steady and Pulsating Performance of a Variable

Geometry Mixed Flow Turbocharger Turbine, Thesis (PhD). Imperial College of Science, Technology, and Medicine, University of London, England, 2007.

2 Spence, S.W.T., Rosborough, R.S.E., Artt. D., McCullogh, G., A

direct performance comparison of vaned and vaneless stators for radial turbines, Journal of Turbomachinery, ASME Vol.129, pp. 53, January 2007.

3 Baines, N. C., Lavy, M., Flows in Vaned and Vaneless Stators of

Radial Inflow Turbocharger Turbines, Institution of Mechanical Engineers Turbochargers and Turbocharging Conference, Paper No. C405/005, pp. 712, 1990.

10.1243/17547164C0012010034

456

4 Capobianco, M., Gambarotta, A., Performance of a twin-entry automotive turbocharger turbine. ASME Paper 93- ICE-2, 1993.

5 Dale, A., and Watson, D., Vaneless radial turbocharger turbine performance, 110/86. Turbocharging and turbochargers, IMechE, 3rd:65:76, 1986.

6 Baines, N.C., Hajilouy-Benisi, A., and Yeo, J.H., The pulse flow performance and modelling of radial inflow turbines, c484/006/94. IMechE, pages 209:219, 1994.

7 Szymko, S., The Development of an Eddy Current Dynamometer forEvaluation of Steady and Pulsating Turbocharger Turbine Performance, Thesis (Ph.D.), Imperial College London, UK, 2006.

8 Arcoumanis, C., Hakeem, I., Khezzar, L. and Martinez-Botas, R.F. Performance of a Mixed Flow Turbocharger Turbine Under Pulsating Flow Conditions, Transc ASME 95-GT-210, 1995.

9 Karamanis, N., Martinez-Botas, R.F., Mixed-Flow Turbines for

Automotive Turbochargers: Steady and unsteady Performance, IMechE Int. J. Engine Research, Vol 3 No.3, 2002.

10 Szymko, S., Martinez-Botas, R.F., Pullen, K. R., McGlashan, N.R.

and Chen, H., A High-Speed, Permanent Magnet Eddy-Current Dynamometer for Turbocharger Research 7th Int. Conf on Turbochargers and Turbocharging, Proc. of the IMechE, paper C602-026, 2002.

11 Watson, N., Janota, M.S., Turbocharging the Internal Combustion Engine, London: The Maxmillan Press Ltd, 1982.

12 Japikse, D., and Baines, N.C., Introduction to Turbomachinery,

Concept ETI Inc., USA and Oxford University Press, Oxford, 1994.

13 Szymko, S., Martinez-Botas, R.F. and Pullen, K.R. Experimental Evaluation of Turbocharger Turbine Performance under Pulsating

10.1243/17547164C0012010034

457

Flow Conditions, Proc. of ASME Turbo Expo, GT 2005-68878, 2005.

14 Abidat, M., Design and Testing of a Highly Loaded Mixed Flow Turbine, Thesis (PhD.), Imperial College of Science, Technology, and Medicine, University of London, England, 1991.

15 Karamanis, N., Inlet and exit flow characteristics of mixed flow

turbines in advanced automotive turbocharging, Thesis (Ph.D.), Imperial College London, UK, 2000.

16 Hakeem, I., Steady and Unsteady Performance of Mixed Flow Turbines forAutomotive Turbochargers, Thesis (Ph.D.), Imperial College of Science, Technology, and Medicine, University of London, England, 1995.

17 Srithar, R., Martinez-Botas, R.F., Mixed flow turbine research: a review, Journal of turbomachinery, ASME Vol.130, Issue 4, October 2008.

18 Palfreyman, D., Aerodynamics of a mixed flow turbocharger turbine under steady and pulse flow conditions: A numerical study, Thesis (Ph,D.), Imperial College London, UK, 2004.

19 Romagnoli, A., Rajoo, S. and Martinez-Botas, R.F., Turbine performance studies for automotive turbochargers Part two: unsteady analysis, 9th Int. Conf. on Turbocharging and Turbochargers, Instn. of Mech. Engrs, London, 2009.