1. Conjugate heat transfer analysis for gas turbine Film-cooled Blade
Thank you for inviting me to the Jet Engines and Gas Turbines Symposium.
I am honored to be here representing ATEC and Honeywell
The Topic of my presentation is illustrate the strengths and weaknesses of using CHT analysis to predict metal temperatures in a film cooled blade.
Bruno Aguilar
Conjugate heat transfer analysis for gas turbine Film-cooled Blade
November 17th, 2016
15th ISRAELI SYMPOSIUM ON JET ENGINES & GAS TURBINES

2. Agenda Introduction and Motivation Geometry and Computational Domain
Model Mesh Details
Numerical Set Up and Boundary Conditions
Model Convergence
Results and Discussion
Comparison against thermocouple data
Conclusions and Further Investigation
My presentation will follow this agenda.
I plan to illustrate the motivation to perform CHT analysis. This is not only applicable to the Aerospace industry but to all industries where themal-fluid calculations are relevant.
Then, I will describe some details about setting up a CHT model such:
What components to include – what is the relevant geometry that needs to be included?
Meshing – element type, Refinement, etc.
Boundary conditions – heating and cooling fluid sections,
Model convergences – when did the model achieved convergence?
Finally, I will compare the results obtained from the CHT model against thermocouple data including a sensitivity study when using different turbulence models such as:
Shear Stress Transport (SST) w/Transition from laminar to turbulent regime and simple SST
RNG k-epsilon Model
K-epsilon
Spalart-Allmaras (SA).

3. Introduction and Motivation
Modern gas turbine engines operate at high temperatures and speed to obtain high power. High turbine inlet temperature can lead to excessive thermal stress and reduced component life.
In order to correctly calculate cyclic thermal stress and component life, blade metal temperature predictions within the accuracy range of ± 4K (7°F) are required.
CHT Tm calculation is within hth=5-10% which implies DTm=20-73K (50-140oF). This may cause 10-20x error in life prediction.
Accuracy of CHT in film cooled blades will be explored in this presentation.
Gas Turbine Engines follow the Brayton cycle. Thermodynamics tells us that high efficiencies are achieved at high pressure ratios and inlet temperatures
High turbine inlet temperatures create a challenge to the durability of the engine hot section components. High temperatures causes distress due to corrosion and oxidation attack. In addition, high stress and temperatures can cause rotating flow path components to fail due to creep or Thermal Mechanical Fatigue. Thus, advanced cooling is necessary to allow this components to survive under this strenuous conditions.
Accurate metal temperature predictions are very important in order to accurately predict component durability and life. When calibrating a thermal model we target accuracies within ~7F. If you think about it, this is vey ambitious if you consider engine flow path temperatures are in the upper 2000 F. However, this is justified by the dramatic impact temperature has on creep life. In most Nickel allows, an increase of 25 F in metal temperature causes creep life to reduce by 50%.
The accuracy of CHT analysis has dramatically improved in the past years. The CFD industry such as CFX, Fluent and Star CCM+ continue to make improvements at a fast pace. In this presentation, I will show how CHT analysis can provide metal temperatures within 5-10% when advanced cooling schemes are using in airfoils. This accuracy improves as the cooling scheme becomes simpler. Accuracy of F has great implications to the durability of the component which can cause the prediction to be off by as much of 20X
Calculating Blade Temperature Is Challenging

4. Common Thermal Analysis Process
CFD
CHT CFD
Blade
Tm Validation No
Cycle Sheet
Materials
Secondary Flows
HTC
DTm > 4K
Yes
Calibration Factors
DTm ≤ 4K
Conduction Thermal Model
Blade
Tm Validation No
Now I will describe two common analytical thermal analysis processes. Other variants might exists.
One path is to construct a conduction thermal model where fluid-metal interfaces are established using heat transfer coefficients and fluid average temperature. HTCs can be obtained from correlations or using CFD analysis with wall temperature assumptions.
The second method is to use CHT analysis and allow the program to calculate metal and gas temperatures. . The limitation is that there is very few knobs to calibrate the thermal model, if thermal date exits.
On the other hand, Conduction models are very flexible to calibrate since calculated boundary conditions such as HTCs can be scaled to match data, if available.
CHT models are quickly to build while conduction thermal models are more time consuming. However, accuracy is crucial for a safe design.
DTm > 4K
Yes
DTm ≤ 4K
Stress Analyses
Life Calculations

5. Geometry and Computational Domain
Now, let’s talk about setting up a CHT model.
In this example, we are analyzing a filmed cooled blade. Flow path boundary conditions are highly depended on the upstream and downstream components.
Metal cooling depends on the forward and aft cavities flows.
Including too much flow and metal domain causes the CHT model to require high computational power.
On the other hand, including not enough components can cause the CHT model to be inaccurate in its metal predictions.
For this model we reached the following compromise:
The model consist of the fully featured blade with flow path and cooling air delivery system. The blade includes film cooling holes on the blade leading edge, pressure side, suction side and internal cooling discharge at the blade TE pressure side.
We are including the flow path boundary conditions at the inlet of the blade (Total Relative Pressure, Temperature and flow angle) and Relative static pressure at the exit of the blade. This boundary conditions are obtained from a full engine CFD model using Fine Turbo.
For cooling, we are including TOBI exit flow, total temperature and flow angle. The TOBI is a device that turns the flow into the direction of the rotating disk..
TOBI tangential on-board injector inlet.

6. Model Mesh Details
Meshing is an art, and requires some effort to guarantee that accuracy is not comprised. Increasing mesh density increases accuracy but requires more computation power.
In this case, we did a mesh sensitivity study where we analyzed the model sensitivity to Y+ (Dimensionless wall distance). We found that the optimum results were obtained with a Y+ less than 3.
Also, the mesh is a combination of tetrahedral and prims elements. Prisms, conformal mesh, is used on the boundaries between fluid and solid to improve accuracy.
At the end the model ended with 9.8 Million cells. From which 1.2 million correspond to the solid domain and 8.6 million to the fluid domain. This model was ran in a computer cluster allocating >100 processors to obtain results in hours.
Grid sensitivity was completed with different Y+ values.
Y+ values near wall is less than 3.
A combination of tetrahedral (Unstructured) and prism elements are used.
The CFX CHT model consisted of 9.8 million elements:
1.2 million cells in the solid domain (Blade) and 8.6 million cells in the fluid domain.
Conformal contact interface between solid and fluid region. Prism are wrapped around the blade surface to resolve the boundary layer (higher accuracy).

7. Numerical Set Up and Boundary Conditions
ANSYS CFX 14.0
Steady State RANS
Turbulence model
Turbulence intensity factor of 15%.
Shear stress transport (SST) turbulence model with automatic wall function.
Sensitivity study completed for 4 turbulence models: SA, k-ε, RNG and SST w/ transition.
Heat transfer
Total energy and viscous work term
Wall function
Automatic
Static structures assumed adiabatic
The model was created in ANSYS CFX. The CFD solver used was RANS (Reynolds Averaged ), higher accuracy can be achieved using LES or/and a combination of RANS + LES which is not covered in this presentation.
Different Turbulance models were reviewed and compared to thermocouple data to determine accuracy. Models studied include SST (shear stress transport) w/Transition, SST, RNG, K-Epselon and SA. Wall function was left automatic where applicable.
Total Energy and viscous effects were included. Static structures were assumed adiabatic.
We are including the flow path boundary conditions at the inlet of the blade (Total Relative Pressure, Temperature and flow angle) and Relative static pressure at the exit of the blade. This boundary conditions are obtained from a full engine CFD model using Fine Turbo.
For cooling, we are including TOBI exit flow, total temperature and flow angle. The TOBI is a device that turns the flow into the direction of the rotating disk..
Boundary Conditions

8. Numerical Set Up and Boundary Conditions
The inlet gas temperature boundary condition was obtained from a full engine CFD analysis using Fine Turbo. The Inlet gas temperature can be obtained from a Harmonic or Mixing Plane solution.
The Harmonic solution provides results at different clocking positions of the blade relative to upstream/downstream nozzle.
The mixing plane solution provides the average gas temperature that the blade experiences.
The plots shows the results of a tangential averaged harmonic solution against a mixing plane solution. The results are nearly identical. The mixing plane inlet relative total temperature distribution was used in this analysis
Inlet total gas temperature obtained using Fine Turbo.
Harmonic and mixing plane yielded similar absolute radial total temperature profiles.
Mixing Plane Tt(r) is Used

9. Model Convergence CHT steady state solution achieved
RMS residuals between 10-4 and 10-5
Mass flow imbalance ≤ 0.05% of core flow.
Blade temperatures at different locations stable.
Model convergence was monitored by the following parameters:
The residuals for the momentum components. Residuals 10-4 and 10-5 is considered to be an appropriate convergence.
Mass Flow imbalance of less than 0.05% is considered to be an appropriate convergence.
Blade metal temperatures imbalance at different blade locations shown in the plot to the right.
Steady state was achieved in 8000 iterations.

10. Results – External Gas Flow
Low pressure region in the throat due to supersonic flow.
50% radial span
Migration of hot gas towards the pressure side due to flow accelerating on the suction side.
Low temperature layer covering most of suction side.
Now jumping into the results for the external gas flow. The results shown on the slide represent fluid behavior at 50% blade radial span.
The first plot on the upper left corner shows the normalized static pressure distribution.
Notice the suction side shows low pressures and suction side high pressures. This tells us the results make sense.
The next plot shows Mach numbers. Yes, notice the exit Mach number is supersonic. This explains the very low static pressures at the blade exit plane. Also, this is an indication of high load/work turbine blade row. Notice the wake at the TE quickly collapse due to the high velocity flow surrounding the wake area.
The lower right plots show static and total relative temperatures. On the static temperature, notice the migration of hot gas toward the pressure side due to the accelerating flow on the suction side. The total relative temperature plot shows how the film cooling protects the airfoil.
Why is the suction side so cool relative to pressure side? We will explore this later.
50% radial span
Why is suction side so cool relative to pressure side?

11. Results – External Gas Flow
Secondary flow vortices roll the end wall flows from hub and shroud toward mid-span in suction side.
These vortices also disperse the flow to the entire span on pressure side.
In this plot we see the effects of secondary flow in the blade passage. Secondary flows are the migration of flow from the pressure side towards the suction side of the airfoil. On the pressure side the disperse the flow outwards. On the suction side they roll the flow inwards towards 50% radial span.
This behavior alters how film cooling interacts with the airfoil surface. This effect will be clearly seen in a later slide.
Also there is some blade tip hot flow migration from PS toward SS
In a conduction only thermal model. Calibration would be difficult if the analyst lacks a CFD model to visualize these effects.
Secondary flows impact film cooling effectiveness

12. Results – External Gas Flow
Secondary flow vortices gradually roll blade hub and tip cooling flows as flow progresses from the blade leading edge towards the trailing edge.
The cool flow migration causes the trailing edge metal temperature radial gradient to worsen.
Temperature gradients are directly promotional to stress.
This is another representation of how secondary flows disrupt flow path gas temperature distribution and film cooling flows.
At the inlet, the relative total gas temperature does not have a tangential gradient. As it marches through the flow path, it is possible to appreciate the effect of the secondary flow causing the flow path gas to migrate from the pressure side into the suction side. Creating a tangential temperature gradient.
Notice the TE PS temperatures are elevated hot. This helps to explain why suction side TE is running cooler at 50% span.
TE durability will suffer due to secondary flow effects

13. Results – Internal Cooling Flows
This slide shows results for the flow inside the airfoil.
Notice the pressure drop due to entrance loses from fwd cavity to the blade inlet passages it is ~10%. However, this loses is mostly recuperated by the blade pumping effect.
The next plot shows internal flow Mach numbers. High Mach numbers increase heat transfer. The plot also shows some recirculation zones that at the TOBI exit, seal plate and blade internal passages. This recirculation zones reduce heat transfer.
Finally, the last plot shows how static temperature increases as air travels through the blade. Air temperature almost doubles from the source to the trailing edge exit. This is another sign that TE resultant metal temperatures might not be adequate.
Pressure drop due to entrance losses compensated by pumping effect.
Recirculation zones detected inside the blade core, seal plate cavity and Tangential on board ejector (TOBI). This will adversely affect heat transfer and cause pressure loses.
Cooling air temperature increases as it travel through the serpentine. The cooling air temperature in the trailing edge region might not be adequate.

14. Results – Internal Cooling Flows
Secondary flow vortices disrupt the cooling effectiveness from the film cooling. This effect is mainly seen on the suction side.
Last row of cooling flows on the pressure side seem to have some ingestion.
These plots show how effective is the film cooling being injected into the flow path. In this plot the effect of secondary flows is more evident.
Film cooling on the pressure side is disrupted outwards on the pressure side and inwards on the suction side. 50% span on SS TE will be cool and the opposite on the PS
In a conduction only thermal model. Calibration would be difficult if the analyst lacks a CFD model to visualize these effects.
Also the last row of cooling holes on the pressure side show some evidence of ingestion, meaning air is flowing into the blade instead of flowing out, hot gas is flowing into the airfoil. This will cause the TE to run hot.
Blade tip cooling flows migrate to the suction side having little effect on the pressure side.

15. Results – Metal Temperatures
Blade tip hot due to cooling flow rolling over from pressure side to suction side. Also, internal flow recirculation affected some zones.
Trailing edge too hot due to last row of pressure side cooling holes not providing adequate cooling.
Heat tinting field experience validate these results.
This plot shows airfoil metal temperatures. Some observations:
Tip metal temperatures are high due to the effect of hot air migrating from the pressure side toward the suction side. Blade tip cooling discharge quickly migrates toward the suction side on the LE.
Secondary flows greatly disrupt film cooling by rolling the cooling flow to the middle of the airfoil on the suction side –notice the suction side tip area which is heated by two effects: migration of film cooling air towards 50% radial span and migration of hot gas from pressure side.
TE temperatures on the pressure side are affected by hot gas ingestion in the last row of film cooling holes and disruption of the film cooling outwards.
Notice a fielded blade corroborates the results from the CHT mode. Heavy discoloration on the pressure side TE and tip.

16. Results – Comparison Against TC Data
CHT
This slide shows the CHT results compared against TC data
Three blades were instrumented with rotating TCs at 50% span. Each blade had 12 thermocouples, results variation was within 2%.
The CHT prediction compared to the TC data
Most locations are within 50 F, except in the LE suction side where speed is high and there is film cooling. It seems CFD is not doing a good job in predicting flow behavior in this region.
It is important to realize that in the absence of TC data the life prediction will be off by 2^3 or by a factor of 12.
Even though CHT is off, it is off on the conservative side. The risk is to over design the airfoil cooling scheme, but safety is #1! It is always good to trade safety over engine performance.
Three blades instrumented at 50% radial span. 7 TCs on the suction side and 5 TCs on the pressure side. Blade to blade variation within 2%.
Results within 20K (~50F) except the locations 1, 2, 3 on the suction side where DTm can be as high as 73K (~140oF).

17. Results – Comparison Against TC Data
This is another representation of CHT metal temperature prediction vs TC data at 50% blade span.
Comparing the location of the TC measurements against the CHT predictions. Notice that using the SST turbulence model does a decent job predicting metal temperatures. Except on the suction side where you have high speeds with film cooling. CHT fails in locations of high speed combined with film cooling.
In the next slide we will see how metal temperature predictions change by using other turbulence models.
CHT model with SST with transition does an adequate job predicting metal temperatures except on a few points on the suction side.

18. Results – Sensitivity to other Turbulent Models
All models failed to accurately predict the metal temperature on the suction side close to the leading edge. The best prediction came from the SST with transition model.
SA fails to compute fields that exhibit shear flow, separated flow, or decaying turbulence.
K-epsilon resolves for turbulence KE and rate of dissipation, but It does poorly when computing flow fields that exhibit adverse pressure gradients, strong curvature to the flow, or jet flow.
RNG is a little bit of improvement to K-Epsilon.
SST has a significant advantages for non-equilibrium turbulent boundary layer flows and heat transfer predictions.
The SST is economical as k-ε, but it offers much higher fidelity, especially for separated flows.
Overall the SST + Transition was the best overall turbulence model. The KE was the worst on the LE and TE.
However, in the present there might be more robust turbulence models.
All models fail to adequately predict metal temperatures for a few points on the suction side.SA being the worst.
Best model to predict metal temperatures on LE and TE is SST + Transition. K-ε being the worst in these locations.

19. Conclusions
CHT reasonably predicts metal temperature for a filmed cooled blade. However, the metal temperature accuracy might not meet the life prediction requirements.
Although temperature accuracy may not meet the life prediction requirement, CHT results can still provide detailed guidance for model calibration.
Therefore, the CHT simulation could be a valuable design and analysis tool during the preliminary analysis or trade-off studies.
Further investigation:
Investigate potential issues with combustor profile.
Explore different turbulence models with better fluid flow transition prediction capability.
Apply an unsteady inlet boundary condition for the CHT simulation to obtain the blade temperature response.
CHT modeling does a reasonable job predicting metal temperatures, but not accurately enough for lifing components.
CHT does not have many knobs to facilitate calibration against available data.
However, CHT can provide guidance on the flow behavior to accurately calibrate a conduction model which is based on correct flow physics such flow migration and secondary flow effects.
By all means, this is not a perfect model and many additional improvements can be made to improve accuracy. Some of this improvements could be:
Run a thermal survey with a know combustor radial profile.
Use a better CFD technique as LES or/and combination of LES +RANS.
Finally, perform an unsteady CHT analysis including the Combustor and turbine rotating group.

20. Thank you !