`Transient conjugate heat transfer analysis of a turbocharger A. Dimelow Cummins Turbo Technologies Aerodynamics Dept, England ABSTRACT Computational Fluid Dynamics is routinely used for predicting the aerodynamic performance of the turbine and compressor stage in a turbocharger. With increased interest in a better understanding of thermal exchange across components and enhanced temperature prediction techniques, this paper describes key considerations for using Conjugate Heat Transfer analysis to analyse an entire turbocharger at both steady and transient operating conditions. A close agreement of temperatures at sensitive bearing housing locations was achieved between the model and thermal survey at a steady operating condition and a level of confidence is demonstrated when utilising the analysis approach for temperature predictions under a transient Hot Shutdown Event. Conjugate Heat Transfer analysis was also demonstrated to successfully model an Accelerated Thermal Cycle Test in order to generate thermal loads for a subsequent Damage and Thermo-Mechanical Fatigue analysis. This paper will show the great improvements to the accuracy in life cycle prediction when in comparison to a standard boundary condition approach. NOMENCLATURE CHT Conjugate Heat Transfer CFD Computational Fluid Dynamics DTMF Damage and Thermo-Mechanical Fatigue CTT Cummins Turbo Technologies CAD Computer Aided Design x Distance (m) k Thermal Conductivity (Wm -1 K -1 ) SST Shear Stress Transport RFR Rotating Frame of Reference TIT Turbine Inlet Temperature ( C) TIP Turbine Inlet Pressure (bar) ER Expansion Ratio PR Pressure Ratio CIT Compressor Inlet Temperature ( C) COP Compressor Outlet Pressure (bar) TC Thermocouple HTC Heat Transfer Coefficient (Wm -2 K -1 )
WAT HSE FEA ATCT Wall Adjacent Temperature ( C) Hot Shutdown Event Finite Element Analysis Accelerated Thermal Cycle Test 1 INTRODUCTION Computational Fluid Dynamics (CFD) is routinely used for predicting turbocharger behaviour and generating performance maps. Typically, such analysis is primarily concerned with the aerodynamics of the respective stage and does not consider the effects of thermal exchange with neighbouring components. With advanced modelling capabilities and access to greater computational resource, it is possible to build a comprehensive Conjugate Heat Transfer (CHT) model of an entire turbocharger. This multi-domain model is inclusive of housings, core solid components and stage cavities and provides a tool for predicting component temperatures across various steady and transient operating conditions by simultaneously solving conduction and convection. In addition, a CHT model enables the opportunity to explore thermal effects on turbocharger performance and can be used to generate thermal loads in support of higher resolution subsequent Damage and Thermal Mechanical Fatigue (DTMF) analysis. This paper has outlined the general steps and considerations involved in the building of a CHT model of a turbocharger and describes three unique studies demonstrating not only the recent analysis developments at Cummins Turbo Technologies (CTT) but the broad application of this analysis approach. All three studies were completed alongside physical tests in order to provide a measurement of validation. 2 METHOD 2.1 GENERAL ASSEMBLY PREPARATION The level of geometric complexity needed in the 3D CAD models for all cases was governed by the resolution required to achieve the analysis objectives and the available computing resource. A typical turbocharger may comprise of approximately 80 individual components of varying complexity and in most cases, a level of component de-featuring and omission is required. Geometry modifications included the elimination of very small clearances in order to ensure flush component interfaces, for instance between housing architecture, and the simplification of small component details. This was to ensure the generation of a high enough quality mesh throughout the model. Components considered less significant in terms of overall heat transfer across the turbocharger, for instance, o-rings, probes and the actuator assembly, were removed. The majority of geometry changes were made in the native CAD package but it in some instances it was more practical to utilise the geometry preparation tool, ANSYS Design Modeller. It was ensured that modifications were not made to fluid cavity walls if they were anticipated to affect the aerodynamic performance. For example, any clearance between the turbine housing and the heat shield was not omitted. The fluid cavities were generated and split into the respective stationary and rotating domains in ANSYS Design Modeller. Commented [CSH1 be caps either, ch just the first wor Commented [AD2 Commented [AD3 2.2 GENERAL MESHING CONSIDERATIONS Solid domains were meshed using suitably sized tetrahedral elements. A global element size of 2mm was used but in regions requiring higher mesh resolution, for instance at the wheel/impeller leading/trailing edges, face size values ranging from
0.1mm to 1.5mm were applied. The total mesh count for all solid domains of the turbocharger was ~20million. Fluid domains were meshed using a combination of tetrahedral and prismatic elements. A global element size of 2mm was used but face element sizes ranging from 0.1mm to 1.5mm were applied to areas requiring higher resolution. In domains where axisymmetric clearances were present, for instance at the inner and outer diameter of the heat shield cavity, a swept mesh method was applied in order to generate higher quality elements at relatively acute angles. Prism elements were used in order to capture the fluid and thermal boundary layers. At the static fluid domain surfaces, an initial cell height of 0.1mm was used, with eight cell layers and a growth rate of 1.2. In the rotor domain, all surfaces were given an initial element height of 0.025mm, with six layers and a growth rate of 1.15. The total mesh count for the fluid domains was ~25million. Y+ values were typically seen to be <15 throughout the stage cavities. 2.3 GENERAL CHT MODEL CONSIDERATIONS Solid domains were assigned temperature dependent material properties as per internal CTT data sheets. Neighbouring solid components were modelled with flush contact interfaces. Where small clearances existed, for instance between a journal bearing and the shaft or between housing architecture, appropriate thermal resistance terms were included to prevent an over calculation of conduction across the interface using the following equation: Thermal Resistance Term = x k Ideal air properties were used for the turbine and compressor fluid domains. The Shear Stress Transport (SST) turbulence model was selected for all fluid domains except the heat shield cavity which was assumed to have free convective laminar flow. Equivalent sand grain roughness values were applied to stage wetted surfaces to represent average casting tolerances. The wheel and impeller rotating domains were modelled with the Rotating Frame of Reference (RFR) method. At the interfaces of the static and rotating fluid domains, a stage ( mixing plane ) interface type was used in order to generate an averaged downstream flow profile. Heat generation by bearing losses was not considered. The Boussinesq buoyancy model was used in the heat shield cavity as this produced the most stable solution. This model utilises incompressible air properties and generates a representative momentum source which is governed by the local temperature difference from a given reference temperature. The reference temperature used was the heat shield cavity volume average. As per the operating condition of interest, a pressure ratio and inlet temperature were applied as boundary conditions to both the turbine and compressor stage. A rotor speed was applied to the rotating wheel and impeller domain and a counterrotating wall velocity was set to the non-rotating walls. Ambient heat transfer interactions were represented by using a Heat Transfer Coefficient (HTC) and reference temperature boundary condition applied to the housing external walls. Heat transfer interactions with the oil drain/gallery walls were accounted for using a set of HTCs and reference temperature boundary conditions developed at CTT. For transient CHT models, the first time step of the simulation is initiated from a representative steady solution. It is important to therefore, select a transient starting point at an anticipated thermally balanced condition.
Commented [CSH4 arrows point to 3 STUDY 1 CHT ANALYSIS OF HEVG TURBOCHARGER A CHT model of a non-water cooled HEVG turbocharger was generated in order to predict temperatures at critical bearing housing locations under a range of steady operating conditions. A thermal survey of the turbocharger was completed on an engine test in order to understand the level of error present in the CFD model. CFD monitor points were positioned at equivalent thermocouple locations in the test in order to compare temperatures. Thermocouple locations were focussed on the regions of interest, around the journal bearings and piston seal ring. 3.1 GEOMETRY PREPARATION Figure 1 - HEVG turbocharger cad assembly following geometry preparation All housing architecture was made flush and small component details simplified. Small clearances between components not affecting the fluid cavities were also removed. The actuator was omitted from the model. 3.2 CFD MODEL COP TIP TIT Qamb BOUNDARY CONDITION CIP CIT Turbine ER 2.95 Compressor PR 1.99 TIT ( C) 600 CIT ( C) 25 TOP Oil Inlet Temperature ( C) 110 Rotor Speed (RPM) 160000 Qoil Ambient Temperature ( C) 25 Figure 2 - Boundary conditions for HEVG turbocharger at steady operating condition
Temperature ( C) 3.3 RESULTS Figure 3 - Temperature contour along mid plane of HEVG turbocharger Commented [CSH5 should be capital contour Survey CFD 250 150 50 0 TC1 TC2 TC3 TC4 TC5 TC6 Thermocouple Location Figure 4 - Temperature comparison between CHT model and thermal survey at thermocouple locations at steady operating condition Figure 4 shows that the CHT temperature predictions were an average of 1.6% higher across the piston seal ring thermocouples, an average of 7.9% higher across the turbine side journal bearings and an average of 1.8% lower across the compressor side journal bearings.
3.4 DISCUSSION The level of agreement between the CHT model and thermal survey gives a useful degree of confidence for the purpose of evaluating different design concepts under equivalent modelling assumptions. A limitation of the model includes the resolution of the HTCs used for representing oil interactions and the associated thermal exchange in the oil gallery. Implementation of multiphase CFD analysis supported by physical testing present good opportunities to increase the analysis resolution and also broaden design concept prediction in the area of oil manipulation. An additional extension to the model would be the introduction of a bearing loss term to enhance thermal interactions in the bearings. Investigating both internal and external net radiation effects would enhance the model and possibly enable the model to evaluate the thermal impact of a turbocharger s orientation with respect to hightemperature neighbouring components, such as the engine exhaust manifold. Additional temperature comparisons with thermal tests at multiple operating conditions are needed to gauge the robustness of the CHT model approach. An increased number of thermocouples would also enhance the understanding of the model s accuracy across more componenets in addition to providing the opportunity to develop boundary condition terms at the flage locationsof adjoining components not modelled. Further mesh sensitivity studies could be complete to achieve a better resolution of both the fluid and thermal boundary layer. Commented [CSH6 sense 4 STUDY 2 TRANSIENT CHT ANALYSIS OF A HOT SHUTDOWN EVENT IN A HEVG TURBOCHARGER Immediately disengaging an engine following a period of high torque/power operation without sufficient idling can result in a Hot Shutdown Event (HSE). Due to negligible cooling from oil flow, high thermal gradients between housings cause a thermal soak across the turbocharger. This can generate temperature spikes at sensitive locations around bearings and piston rings and repeated HSEs can lead to oil degradation, component wear and ultimately reduced service life. The availability of a CHT model to predict absolute temperature spikes resulting from an HSE is of significant value as it can enable the evaluation of multiple design concepts or support a DOE earlier in the design process and reduce the number of physical tests and associated costs. An additional benefit of CHT analysis for modelling an HSE is the flexibility of evaluating thermal performing under a range of operating scenarios, such as different idling periods. 4.1 CHT MODEL A number of assumptions were made in order to simplify the physics and converge to a stable solution. At the initiation of an HSE the pressure drop across both stages will transiently reduce to near zero. This spool down period was not modelled as it occurs over a timescale far smaller than of interest in terms of heat propagation across the housings. As such, at the first time step, it was assumed that the pressure drop across both stages were near zero. A negligibly small pressure gradient was maintained across both the stage cavities in order to assist model stability.
Temperature ( C) Temperature ( C) Temperature ( C) Flow characteristics in the stages were then be assumed to be governed by free convective buoyancy forces and such, incompressible air properties were used alongside the Boussinesq buoyancy model. This generates a representative momentum source governed by the local temperature difference from a given reference temperature. The reference temperature used was the respective stage cavity volume average. Temperatures throughout the turbocharger core components will typically peak within a 10 minute time period. A timestep of 0.02s was required to ensure a stable convergence. The CHT model of the HSE was initiated from the steady state solution obtained in Study 1. 4.2 RESULTS 350 350 250 250 150 150 0 50 150 250 Time (s) a TC1 TC2 TC3 TC4 TC5 TC6 0 50 150 250 Time (s) b TC1 TC2 TC3 TC4 TC5 TC6 350 Survey CFD 250 150 50 0 c TC1 TC2 TC3 TC4 TC5 TC6 Thermocouple Location Figure 5 - a. Transient temperature profiles of thermocouple locations in thermal survey b. transient temperature profile of thermocouple locations in CHT model c. maximum temperature at thermocouple locations during Hot Shutdown Event Figure 5 shows that the CHT maximum temperature predictions were an average of 9% higher across the piston seal ring thermocouples, 1.4% lower across the turbine side thermocouple locations and 9.7% lower across the compressor side thermocouple locations. Commented [CSH7
4.3 DISCUSSION Whilst Study 2 has not generated as close an agreement with the thermal survey as per the steady operating condition in Study 1, this transient CHT approach could still be applied for directional comparison of thermal performance prediction in an HSDE. Reasons which may explain the difference between the thermal survey and CHT model temperatures include potentially unsteady ambient interactions and the influence of adjoining components such as the exhaust manifold which was not taken into account in the model. Additional correlation with the thermocouples positioned around the turbine housing would provide the opportunity to develop a representative boundary condition at the turbine housing flanges to capture the effects of the turbocharger thermal inertia. In addition, net internal and external radiation effects may influence the rate of cooling. Multiple radiation models could be investigated and implemented into the CHT model to evaluate the effects. It is important to recognise that the accuracy of a transient model is heavily determined by the accuracy of the steady-state solution which the model is initiated from. The Boussinesq model which simplifies the buoyancy forces in the flow physics may be a contributing factor in the discrepancy between the temperatures. Alternative buoyancy forces based on ideal gas properties and density fluctuations could be explored. The timestep is an important value to investigate in transient analyses. Further sensitivity studies would enable the determination of the effects of reducing the Courant Number in terms of time-based accuracy. 5 STUDY 3 GENERATION OF TRANSIENT THERMAL LOADS FOR INCREASED RESOLUTION DTMF ANALYSIS OF HEWG TURBOCHARGER Damage and Thermo-mechanical Fatigue (DTMF) analysis is a Finite Element Analysis (FEA) modelling technique which leads to the failure of a material due to cyclical thermal loads and is an important stage in the design process of turbocharger components. A standard approach to performing a DTMF analysis utilises a combination of transient thermocouple data from an Accelerated Thermal Cycle Test (ATCT) and constant heat transfer coefficients in order to generate thermal boundary conditions. Thermal and structural FEA analysis is subsequently carried out over a number of ATCTs to generate inputs for the TMF predictions. The success of the TMF prediction is highly influenced by the accuracy of the absolute temperatures and thermal gradients calculated in the thermal FEA model. CHT analysis can be used to generate increased resolution thermal inputs on the stage wetted surfaces. Whereas the standard approach uses constant heat transfer coefficients, CHT analysis will export individual nodal Heat Transfer Coefficients and Wall Adjacent Temperatures as.csv files for each thermal load step. This is then interpolated onto the FEA model mesh and so the thermal effects and gradients associated with gas expansion/compression are captured. 5.1 GEOMETRY PREPARATION In order reduce the size of the CHT transient solver time, thermocouple locations at the turbine housing/bearing housing flange would be used as a boundary condition meaning the bearing housing and compressor stage could be omitted from the model.
TIT ( C) All geometric considerations for the turbine housing and wastegate components were consistent with Study 1 and Study 2. 5.2 CHT MODEL TC1 TC5 TC3 TC2 TC4 Figure 6 - Thermcouple locations for HEWG turbocharger 800 1/30 Hz 1 Hz 700 600 500 400 Period of interest 0 0 400 600 800 0 Time (s) Figure 7 Transient TIT ATCT boundary condition for HEWG turbocharger
Temperature (deg C) Temperature (deg C) In order to enhance model stability, the profile gradients for the TIT, TIP and rotor speed were reduced by interpolating the data at a frequency of 30-1 Hz. Figure 7 above illustrates how the ATCT data was used for the TIT boundary condition. As is highlighted in figure 7, the time period of interest starts at a non-thermally balanced condition. As such, the analysis was initiated ~280s earlier in order to develop a more thermally representative condition at the actual time step of interest. A steady-state CHT analysis was completed at this point and used to initiate the transient simulation. A timestep of 0.05s was used in the transient simulation. 5.3 RESULTS CHT and WAT values were generated and exported at all turbine stage wetted surfaces every 30 seconds for interpolation on the subsequent FEA analysis. Figure 8 - HTC contour plot on HEWG wetted surfaces at 540s in the ATCT 600 TC1 TC2 TC3 TC4 TC5 600 TC1 TC2 TC3 TC4 TC5 500 500 400 400 a 0 0 400 600 800 0 1 1400 Time (s) 0 0 400 600 800 0 1 1400 Time (s) Figure 9 - Transient ATCT temperatures at thermocouples for HEWG turbocharger for a. thermal test and b. CHT model b Figure 8 illustrates the high HTC gradients seen on the turbine stage walls. Figure 9 compares temperatures at thermocouple locations between the ATCT with the CFD.
LOCATION DTMF RATIO DTMF RATIO (Standard) (CHT Supported) Waste-Gate Bore to Volute 11.55 0.51 Waste-Gate Bore 16.44 0.63 Outlet (edge) 2.65 0.97 Radius Around Waste-Gate 157.40 8.27 Bore Tongue End 3.52 1.90 Volute Between Tongue Crack 1.36 0.69 and Waste-Gate Between Waste-Gate Bracket 0.71 0.17 Pads Adjacent to Waste-Gate 77.25 3.53 Bracket Pads Adjacent to Feeder 111.11 0.27 xx Cycle Prediction Within Ratio of 2 (0.5 2) xx Cycle Prediction Within Ratio of 5 (0.2-0.5 and 2 5) xx Cycle Prediction Outside Ratio of 5 Outlet Sidewall 30.00 0.76 Sidewall at Bearing Housing 15.91 0.46 Side AVERAGE 413.3 1.65 Figure 10 - Comparison of DTMF results between standard approach and CHT supported approach The DTMF Ratio is the ratio of predicted life cycle to actual test cycle. A target of >0.5 and <2 is considered good. 5.4 DISCUSSION Differences between the transient test ATCT temperature and CHT predicted ATCT temperatures are present. However, the temperature profiles largely agree and provide confidence that the thermal loads exported at the cavity wetted surfaces contain a reasonable representation of the thermal gradients. Temperature differences will exist due to the simplified boundary conditions interpolated from the test and the interpolation frequency could be further investigated to understand how sensitive this is on the housing temperatures. In addition, the thermal exchange with the bearing housing was simplified using a transient temperature across the entire flange surface. An extended model inclusive of the bearing housing could be built in order to investigate how sensitive the bearing housing is on the turbine housing temperatures. As per Study 2, further investigation of the effects of the timestep would enhance understanding on the transient temperatures across the turbine housing. Figure 8 shows contrasting areas of high and low thermal interaction at the wall associated with local flow characteristics. The objective of the CHT analysis was to generate thermal load inputs for a subsequent DTMF analysis. Therefore, the comparison of the standard DTMF analysis and the CHT Supported DTMF analysis is important. As given in figure 10 the CHT supported DTMF analysis vastly improved the predicted failure cycles and in the majority of locations are within the good ratio of 0.5-2. The average failure prediction ratio is calculated as 1.65 as opposed to 465 using the standard DTMF conditions.
6 CONCLUSION All three studies have demonstrated different applications of Conjugate Heat Transfer (CHT) analysis of a turbocharger. Study 1 and Study 2 were primarily focussed on temperature prediction capability under a steady operating condition and a transient Hot Shutdown Event (HSE). Study 3 generated thermal loads to support a highresolution Damage and Thermo-Mechanical Fatigue (DTMF) analysis. Whilst the results of Study 1 and Study 2 showed a difference in temperatures between the test thermocouples and the CHT model by <10% and <15% respectively, the results do provide confidence that adopting this analysis method can widen and enhance a design process with respect to thermal management considerations. A greater number of designs can be evaluated and optimised prior to costly physical testing. The more sophisticated the turbocharger assembly, the more attention is required to the treatment of the contact points between components. Although the studies within this paper have not included a water-cooled housing, the same approach can be adopted to model both steady and transient coolant flow. Study 3 demonstrated a more simplified geometry under transient boundary conditions derived from an Accelerated Thermal Cycle Test. The real measure of the success of the CHT model in this instance was the comparison of the DTMF analysis results between the standard approach and the CHT supported approach. The results showed a clear improvement in life cycle prediction and for turbine housings, this paper provides strong evidence to use CHT analysis for generating thermal loads. 7 REFERENCES [to be completed]