Forum
Dynamic Stall Errors
Hi Dr. Marten,
I’m running tests on a Darrieus rotor at a steady TSR of 0.5. The objective is to determine the effects of each dynamic stall model in this version of QBlade in contrast to that of the prior version (Beddoes Leishman in Qblade 1.0). The results of 5 plots are demonstrated in the attached image: the original polar taken directly from the extrapolation module; the results of a simulation with dynamic stalls disabled; the GOR, ATEF, and the OYE model.
The results are rather strange compared to the previous QBlade. The results of the simulation without dynamic stall should overlap directly with the original polar plot but does not do so. Next, the GOR model does not behave appropriately. Its as if the blade is stalling and recovering instantaneously. The OYE model, just flatlines and is unrealistic. The most trustable result is that of the ATEF model.
In a separate work soon to be published, we validated the old QBlade and Beddoes Leishman model against experimental results. The ATEF model presented here has magnitudes similar to that obtained by the Beddoes Leishman model at low AoA . However, I am going to test this further to see what the results look like when encountering high AoA, since you mentioned here that ATEF is not good at high AoA.
Additional info:
The simulation was performed for 200 timesteps with stepsize of 3°
- You need to login to have access to uploads.
Hello,
its hard to tell what is not working on your side. But simulations at a TSR of 0.5 have AoA oscillations around +-170°. This is beyond the ability of any DS model to predict correctly. For such an operating condition the reduced frequency is also quite low…
The oscillations that you are seeing could come from the wake elements impinging on the blade – but again I am not sure. Try running some DS tests at a larger TSR, such as 1.5 – you should get much more reasonable results.
When comparing the Cl/AoA curves you need to make sure that they are extracted from a section where Cl is not interpolated over 2 airfoils/polars – then the results from the simulation should fully align with the steady state polar data.
I have attached an screenshot as reference that is directly taken from QBlade for the SANDIA 34m turbine template that shows a comparison of the DS models and the quasi-steady (BASELINE) curve in QBlade.
BR,
David
Uploaded files:- You need to login to have access to uploads.
Hi Dr. Marten,
I’ve looked deeper into this and found the following results from figure 1 :
The GOR model was tested over the selected range of TSR in the plot. The severe fluctuations have largely disappeared because I removed the structural input file and selected ‘None’. Therefore, there is a problem in my structural file, but I’m currently unsure where the issue is. Regardless, there are still some fluctuations at TSR = 3.5 and 4.5.
Additionally, for a blade height of 1 meter, I have extracted the data from a section height of 0.471 (where an airfoil node exists), yet the ‘UA off’ and ‘original polar’ plots do not align. I am using the same airfoil along the entire blade span, and the same polar for all the airfoil sections. Do you encounter this issue also? Or perhaps I’m not considering some 3D effects…
I believe there is more for me to experiment on my side regarding the above two issues, I am probably missing something important. However, I have also discovered something quite critical in the new version of QBlade. In a separate simulation that attempted a full self-starting simulation with a single polar, I found that reverse dynamic stall is not predicted by the GOR model. I suspect that this is due to the zero value of the F function at 180° α, described here. In contrast, using the previous version of QBlade, we found that reverse dynamic stall magnitudes in self-starting VAWTs are reasonably predicted (against experimental results [1-2]).
Regarding the absence of reverse dynamic stall, I present figure 2 as attached. I can’t present a proper plot to you because my Matlab codes currently cannot process the output data from QBlade. This is because the step size is constantly changing, which shouldn’t be the case because I set it to a constant of 3° theta. I can send you the raw data files if you are interested to look at them. I only encounter this variable step size issue with a self-starting case. Again, I suspect it has something to do with my structural input file. I will update my code to process the variable timestep size, but I’ll demonstrate this figure to you first so you can have an idea of what it looks like.
Figure 2 demonstrates the lift coefficient against α where the rotor has made over 400+ revolutions which is the reason for the density of the plot. The highlights in black and blue, are where reverse and forward dynamic stalls are expected to occur respectively. The lift coefficient in the black box should be as high as in the blue box. Currently, the lift coefficient magnitude in the black box corresponds to that of the input polar. I am inclined to think that if the polar decomposition tool allow us to alter the F function at around ± 180° α, this issue may go away. However, you mentioned (Reply #2) that the GOR model does not require decomposition, so I’m unsure if that will solve the issue
[1] A. H. Lind and A. R. Jones, “Unsteady aerodynamics of reverse flow dynamic stall on an oscillating blade section,” Phys. Fluids, vol. 28, no. 7, 2016, doi: 10.1063/1.4958334.
[2] J. Hodara, A. H. Lind, A. R. Jones, and M. J. Smith, “Collaborative investigation of the aerodynamic behavior of airfoils in reverse flow,” J. Am. Helicopter Soc., vol. 61, no. 3, 2016, doi: 10.4050/JAHS.61.032001.
- You need to login to have access to uploads.
Hi,
in general, I would advise to NOT use a DS model during startup where the AoA is oscillating by +-180°. No DS model is valid for such an operating condition and the models are not representative of the aerodynamics during startup at all…
here are a few additional thoughts:
Additionally, for a blade height of 1 meter, I have extracted the data from a section height of 0.471 (where an airfoil node exists), yet the ‘UA off’ and ‘original polar’ plots do not align. I am using the same airfoil along the entire blade span, and the same polar for all the airfoil sections. Do you encounter this issue also? Or perhaps I’m not considering some 3D effects…
If you are using the “two point lift/drag evaluation” you should use the angle of attack at 0.75% chord as a reference, not the angle at 0.25% chord. Then the polar data should align.
I believe there is more for me to experiment on my side regarding the above two issues, I am probably missing something important. However, I have also discovered something quite critical in the new version of QBlade. In a separate simulation that attempted a full self-starting simulation with a single polar, I found that reverse dynamic stall is not predicted by the GOR model. I suspect that this is due to the zero value of the F function at 180° α, described here. In contrast, using the previous version of QBlade, we found that reverse dynamic stall magnitudes in self-starting VAWTs are reasonably predicted (against experimental results [1-2]).
What do you define as “reverse dynamic stall” and under which conditions should this occur? Thid would require an AoA oscillation around a mean of -180°. During normal operation this should not occur (except if the VAWT is rotating backwards). During startup, the oscillations are in the range of +-170° AoA – but there is no oscillation around a mean of -180°. The GOR model does not use the f-function at all – so this doesnt affect the results in any way.
Regarding the absence of reverse dynamic stall, I present figure 2 as attached. I can’t present a proper plot to you because my Matlab codes currently cannot process the output data from QBlade. This is because the step size is constantly changing, which shouldn’t be the case because I set it to a constant of 3° theta. I can send you the raw data files if you are interested to look at them. I only encounter this variable step size issue with a self-starting case. Again, I suspect it has something to do with my structural input file. I will update my code to process the variable timestep size, but I’ll demonstrate this figure to you first so you can have an idea of what it looks like.
What do you exactly mean by variable step size? In a time domain simulation the timestep size is always constant, it cannot be variable. That also means that the azimuthal discretization will vary if your RPM is changing, such as during startup. The “azimuthal step size” entry in the setup dialog is just valid for the rotational speed that is shown in the “RPM” entry of the same dialog.
Figure 2 demonstrates the lift coefficient against α where the rotor has made over 400+ revolutions which is the reason for the density of the plot. The highlights in black and blue, are where reverse and forward dynamic stalls are expected to occur respectively. The lift coefficient in the black box should be as high as in the blue box. Currently, the lift coefficient magnitude in the black box corresponds to that of the input polar. I am inclined to think that if the polar decomposition tool allow us to alter the F function at around ± 180° α, this issue may go away. However, you mentioned (Reply #2) that the GOR model does not require decomposition, so I’m unsure if that will solve the issue.
Again, what is the exact process of “reverse dynamic stall” maybe plot the AoA over time to see how this is changing at the section over time to get a better view on this. Also, why should the negative lift peak (when the airfoils leading edge is flying foward) have the same magnitude as the positive peak?
I hope these answers and suggestions can help a bit…
BR,
David
Hi Dr. Marten, sorry for the delay in my response. I’m currently rushing to complete a few things simultaneously.
If you are using the “two point lift/drag evaluation” you should use the angle of attack at 0.75% chord as a reference, not the angle at 0.25% chord. Then the polar data should align.
Okay, I tried this again and I found that they overlap almost perfectly, though there are still some minor discrepancies shown in the plot. I imagine these are from the 3D effects in the simulation?
What do you define as “reverse dynamic stall” and under which conditions should this occur? Thid would require an AoA oscillation around a mean of -180°. During normal operation this should not occur (except if the VAWT is rotating backwards). During startup, the oscillations are in the range of +-170° AoA – but there is no oscillation around a mean of -180°. The GOR model does not use the f-function at all – so this doesnt affect the results in any way.
The work by Hill et al.[1] shows that when the TSR is close to zero, α can reach up to ± 180° . In that regard, for a self-starting rotor which begins rotation at rest, the flow is attacking the trailing edge of the airfoil at certain azimuthal locations. Airfoils with flow attacking the leading edge have stall angles. Similarly, airfoils with flow attacking the trailing edge have stall angles too. These ‘reverse flow’ stall angles are only marginally lower than that of their ‘forward flow’ counterpart. [2]
In the context of the QBlade simulation, reverse dynamic stall is defined as when the rotor blade exceeds the ‘reverse flow’ stall angle and the lift is artificially enhanced beyond that of the original polar.
What do you exactly mean by variable step size? In a time domain simulation the timestep size is always constant, it cannot be variable. That also means that the azimuthal discretization will vary if your RPM is changing, such as during startup. The “azimuthal step size” entry in the setup dialog is just valid for the rotational speed that is shown in the “RPM” entry of the same dialog.
I see, that would explain why the azimuthal discretization is changing. I will look back at this again.
Again, what is the exact process of “reverse dynamic stall” maybe plot the AoA over time to see how this is changing at the section over time to get a better view on this. Also, why should the negative lift peak (when the airfoils leading edge is flying foward) have the same magnitude as the positive peak?
Regarding your second question in this quote, the airfoil trailing edge is flying first in the black box while the leading edge is flying first in the blue box. I meant that the magnitude of the lift in the black box should be more or less equal to that of the blue box due to the effects of dynamic stall. I’ve looked through experimental works (page 13 shown by the top left corner of [2]) and it shows that the lift enhancement caused by reverse dynamic stall can be as strong as that of forward dynamic stall.
I believe it’s becoming much more difficult to discuss this topic without proper plots. Maybe we can put this discussion on hold temporarily? Currently, we are close to submitting a second manuscript. Once the papers are published we can resume. I actually have a lot to share and QBlade has been extremely useful for the elucidation of the rotor dynamics
[1] N. Hill, R. Dominy, G. Ingram, and J. Dominy, “Darrieus turbines: The physics of self-starting,” Proc. Inst. Mech. Eng. Part A J. Power Energy, vol. 223, no. 1, pp. 21–29, 2009, doi: 10.1243/09576509JPE615.
[2] A. H. Lind and A. R. Jones, “Unsteady aerodynamics of reverse flow dynamic stall on an oscillating blade section,” Phys. Fluids, vol. 28, no. 7, 2016, doi: 10.1063/1.4958334.
- You need to login to have access to uploads.
