Effect of Conical Spiral Flow Channel and Impeller Parameters on Flow Field and Hemolysis Performance of an Axial Magnetic Blood Pump

Round 1

Reviewer 1 Report

Dear Authors

I have found a lot of unproven sentences. In many places the discussion is lacking or very poor. It requires significant improvement. Please, see your originally submitted file (if you have not received it, please contact Editor). When you answer all questions and correct shortcomings, please write explanations in text as much and as clear as possible.

Below I present only general remarks as the most important and critical points concerning the quality of paper.

Introduction. In my opinion it should be extended with more literature. Paper contains only 11 References. It looks very poor so it must be changed.

General comment. Hemolysis should be Haemolysis in the whole text.

L72. Please explain the term “five-degree-of-freedom levitation”. As the sentence concerns the axial bearing it is quite not understood statement. Add it to the text.

Fig. 1. The flow direction should be added. Axial hybrid bearing is not symmetrically sketched. Why?

General comment. It would be of a great significance to add the photos of the pump and its parts, if possible, especially the impeller. It is not obvious how impeller looks like and the other parts.

Chapter 2.1. The references 8 and 9 are not used after 6 and 7. References 6 and 7 are placed further. It should be changed.

L80-81. Add dimensions to the Fig.1. The other important ones also should be placed.

Chapter 3.1. In my opinion it contains quite obvious sentences concerning common use of CFD (and particularly ANSYS/Fluent). It could be deleted completely without any loss for paper quality. I leave it to decision of the Authors (I do not insist to delete it, please decide). In case in which Fluent would be one and only software on the market able to solve the task with blood as a medium then it would be understood the justification of using it.

Chapter 3.2. 1) First paragraph presents the blood parameters. As we know Fluent does not contain the defined blood. It is important to write that this is artificial medium with blood-like parameters. Sometimes we use the parameters of glycerine to simulate behaviour of the blood. Is this the case here? Please write in the text. 2) The blood is non-Newtonian (rheological) liquid. Some models are not adequate to use it in such a case. Why did The Authors use Standard k-e model? Please justify your choice.

L146 and L154-155. In line 146 there is stated that the pressure difference was set to 13.3 kPa. In lines 154-155 there is stated mass flow inlet and outflow boundary conditions were set. This is contradictory. Additionally, the ‘outflow’ condition does not contain any value to be set in Fluent, so no pressure can be set. This is quite weird using such a set of boundary conditions, because it does not allow to ensure the pressure difference. Please explain it.

Fig.3 is unreadable. Parts of pump should be presented separately.

Fig.4. 1) How can be explained the horizontal line between 0.25 and 0.5deg. Please explain in text. 2) The highest value on vertical axis should be 120.

Fig.5. The streamline distributions are not readable in such a shape of presentation. A reader cannot conclude from these pictures. It requires zooming in and discussion about the differences. Additionally, the map is needed. Of what parameter the streamlines are coloured. It is not mentioned.

L190. Picture e) does not contain backflows. This sentence is not adequate to picture. Additionally, the caption is wrong.

L191-193. The Fig.5 does not explain to a reader why 0.75deg is the best to be applied. This must be re-arranged.

Fig.6. There is no discussion why point for 4 blades is lower. In my opinion it is very probable that all points should be located in straight line. It may be caused that computation was not iterated enough. It should be discussed and justified in the text.

L211. It is not shown in Figure 6 that “… outlet pressure was negatively correlated with the number of blades of the rear guide wheel.” It must be re-arranged. The pressure difference is also dependent on inlet pressure, which is dependent on mass flow rate (boundary condition) – here inlet pressure is not presented so this conclusion is not obvious for a reader.

L223-L227. The discussion is not proven by the presented results. It must be discussed more clearly. The same also concerns the other parts of paper.

General comment. The Authors use kind of rotational pump. It may be possible that rotor generates low areas of pressure (including cavitation), which can make red blood cells destroy. The Authors should evaluate and write if there are observed some areas of low pressure that is dangerous for blood. In my opinion it is quite important for a quality of the paper.

L273-L275. It is quite important statement. The given average value should be discussed why it “was excellent”. Additionally, the sentence should be referenced to prove why “the hemolysis performance of the blood pump was excellent”.

Fig. 10. 1) It is highly unreadable. It must be changed. 2) It should be moved to chapter 4.3.

L320. “…the clearance should be kept as large as possible.”. In L308 it is written „…when the clearance was too large, blood backflow occurred.”. It is contradictory. Please explain/revise it.

Conclusions. L336. It is not proven in the text that it is “When the hub taper angle was 0.72°, the hemolysis prediction value was lower than that of the traditional axial blood pump”. Please explain it in the text.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This study investigated a new blood pump concept supported by electrodynamic bearing, in order to minimize blood damage. This was assessed by steady state RANS-based a CFD simulations and a conventional hemolysis model. The findings showed improved results in comparison to traditional axial blood pumps.

General comments

In think the general concept/research question of the study is interesting. However, I find the CFD method to not be reliable, lacking clear motivation of the selected approach and no model verification appears to have been performed. The manuscript is not very well written, with a lot of redundant, descriptive text, and some sections lacks clarity.

Major concerns and specific comments

• You have not explained why a steady state simulation framework, using a Reynolds-averaged Navier-Stokes (RANS) turbulence modeling approach, is applicable for this flow application. For example, in patient-specific blood flow simulations, RANS-based turbulence modeling had for long time been discouraged due to evidence of their poor performance in these flow regimes (pulsatile, anisotropic, transitional and relaminarizing flows) (Ryval et al. 2004, Yoganathan et al. 2005, Taylor et al. 2010, Gårdhagen et al. 2010, Andersson et al., 2020). The underlying Boussinesq assumption (eddy viscosity concept) of these simplistic models cannot handle the anisotropic turbulent flow features present in these flows and consequently provides unrealistic viscous and turbulence-related shear stress fields, which in the end may lead to misleading conclusions. Blood damage prediction rely on accurate estimation of the local fluid stresses, which cannot be sufficiently estimated using these RANS methods.  Also, here the simulations appear to have been conducted with the standard k-epsilon model, which is one of the least recommended turbulence models for general CFD applications. Furthermore, to motivate the choice of this turbulence model, you have cited Heinonen et al. (2020), which is completely irrational/unrelated reference. How well is the standard k-epsilon model in predicting the pressure difference over the application? Please motivate why the selected CFD modeling framework is a sufficient choice for predicting the flow parameters of interest in this application.
• This study has not provided any motivation of the chosen CFD modeling settings nor any evidence that the model has been verified sufficiently, e.g., according to some common procedure within the CFD community (Celik et al., 2008). In this regard, the modeling errors (mesh-related and iterative convergence errors) on the outputs (solution field) are not known and can be significant. The CFD methods conducted in this study is therefore not trustworthy.
• Why was the hemolysis model proposed by Giseriepen et al., (1992) selected, and not some of the more recent developed models, e.g., see Faghih et al., (2019)? There are various was to compute the scalar stress term in equation 1, as e.g., noted in Faghih et al., (2019). How was this stress term computed in this study, and why?
• The overall post-processing and display of results are not very convincing and sometimes hard to interpret (e.g., the streamline plots in Figs. 5 and 7, poor mesh plot in Fig. 3). Generally, the figure captions are also too brief, lacking clarity.

References

Ryval, J., A. G. Straatman, and D. A. Steinman. "Two-equation turbulence modeling of pulsatile flow in a stenosed tube." J. Biomech. Eng. 126.5 (2004): 625-635.

Yoganathan, Ajit P., K. B. Chandran, and Fotis Sotiropoulos. "Flow in prosthetic heart valves: state-of-the-art and future directions." Annals of biomedical engineering 33.12 (2005): 1689-1694.

Taylor, Charles A., and David A. Steinman. "Image-based modeling of blood flow and vessel wall dynamics: applications, methods and future directions." Annals of biomedical engineering 38.3 (2010): 1188-1203.

Gårdhagen, Roland, et al. "Quantifying turbulent wall shear stress in a stenosed pipe using large eddy simulation." Journal of biomechanical engineering 132.6 (2010).

Andersson, Magnus, and Matts Karlsson. "Characterization of anisotropic turbulence behavior in pulsatile blood flow." Biomechanics and Modeling in Mechanobiology (2020): 1-16.

Celik, Ishmail B., et al. "Procedure for estimation and reporting of uncertainty due to discretization in CFD applications." Journal of fluids Engineering-Transactions of the ASME 130.7 (2008).

Heinonen, R. A., & Diamond, P. H. (2020). A closer look at turbulence spreading: How bistability admits intermittent, propagating turbulence fronts. Physics of Plasmas, 27(3), 032303.

Faghih, M. M., & Sharp, M. K. (2019). Modeling and prediction of flow-induced hemolysis: a review. Biomechanics and modeling in mechanobiology, 18(4), 845-881.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Please find the attached review report file. 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Dear Authors

Thank you for correcting my indications. Please see the other ones, which in my opinion should be taken into account to improve the quality of your paper (see also your revised submitted file).

References. I still sustain the number if Refs. is too small.

Fig.2 caption. Why a) part of figure is called ‘Modal diagram’? This is not a diagram.

Fig.5. Vertical axis. “Outlet flow rate (m/s)”. Flow rate in m/s?! It must be changed.

Figs. 6,8,10. The colour maps are needed (what parameters are on streamlines). The Fig.11 contains it. Why the other ones not?

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

processess-1577947-peer-review-v2: Effect of Conical Spiral Flow Channel and Impeller Parameters on the Performance and Hemolysis of an Axial Magnetic Blood Pump

General comments

Dear authors, I find the comments to most of my concerns and questions to not be very adequate/convincing. For concern:

• You have not answered the questions, but instead focus in explaining how a RANS modeling approach works. You have still not given any rationale why a RANS approach would provide reliable Reynolds stresses predictions for blood damage predictions. In fact, it seems that you have ignored most if my comments here, which suggests that you don’t understand the fundamental modeling difference between different RANS models (k-epsilon, SST, etc.) nor scale-resolving CFD methods, and it’s implication into the credibility of turbulent flow field predictions (and Reynolds stresses).
• I don’t find the model verification procedure to be very convincing. Here you have focus in quantifying the mesh density dependency on the overall mass flow rate, but investigated how the mesh-related error impact the parameters of interest, i.e., the Reynolds stresses (or blood damage predictions). Obtaining mesh insensitivity results if often very easy (forgiving) when assessing bulk flow parameters, such as overall pressure drop (or mass flow rate in this case). I also pointed out the well-known ASME verification procedure by Celik et al., (2008), which you apparently ignored.
• For the results plots, I don’t find it very fruitful and to use streamline plots to highlight region of “blood stagnation” and “turbulent field”. Just because you notice a helical pattern of flow does not necessarily imply that the flow is turbulent.

Additional concerns:

• Page 6. “… the rest of the discrete formats are first-order winward …”. First, I assume that you mean “first-order upwind”? Secondly, first-order discretization schemes are very diffusive and not generally recommended for approximate the gradient terms in the discretized Navier-Stokes equations. In fact, in many well-regarded journals within the community, running first-order schemes for CFD analysis are not allowed.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Some minor revision is needed to improve the English writing.

The authors have revised the manuscript  and answered my concerns very carefully. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx