Metamodel-Assisted Multidisciplinary Design Optimization of a Radial Compressor^†

Round 1

Reviewer 1 Report

The paper showed that a bounded kriging is more suitable than 'conventional' kriging for radial compressor optimization application. Overall nice read, some questions/discussion points however that would improve the content:

1) The main reason why 'bounded' kriging was better rel. conventional kriging was the fact that it was better in handling the design space region where many designs failed due to convergence issues.
The 'bounded' kriging offers one solution to the problem, but an alternative approach would be to change the problem formulation such that the convergence failure is minimizied.
To understand whether such an approach would be possible, the authors could elaborate more around this.
The reader might think that many problems arise due to bad mesh - but in normal applications if the design space is relatively well defined, the mesh should not be a problem.
Other issues could potentially be due to the boundary conditions applied, but it is difficult to judge if this is a likely issue. For instance: Is the OP2 close to the choking margin? If so, many designs in the design space
will prematurely choke when running OP2 - which is physically correct but will result in a 'non-convergence' failure point in this set up since the author uses an outlet mass-flow as boundary condition.
Is the OP1 close to the stall margin? Does is also have an outlet mass flow as BC?

It is not clear for the reader why designs close to the 'optimal' would result in non-convergence/failure.
If true, the 'optimal' design will not be robust (as small variations will cause unfeasible designs).
If many designs fail, isn´t the selected design space, or the selected run setup that is the root-cause?

I would appreciate a more clear case description, to avoid the above discussion.

2) Minor error in Equation 20 - same index used for the two efficiencies

Author Response

see attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a well written paper on the use of metamodel-assisted evolutionary algorithms supported by bounded kriging. The method is applied to a radial compressor case by considering aerodynamic and structural objectives; from this point of view, the authors consider this as a multidisciplinary application.

Technically, my only suggestion to the authors is to eliminate (or at least reduce a lot) the section on (standard) kriging. Also, I see no reason for including mathematical cases.

I have though a most important comment to make: I believe this paper has been submitted to the wrong journal. Since the new thing presented is the implementation of bounded kriging (I am unable to say at which extent this contribution is original!), it would be better for the authors to redirect their submission to a paper on Evolutionary Computations, where hundreds of similar papers (on kriging etc) can be found. In such a case, the mathematical cases will be welcome. Since I understand that the application is in the field of turbomachines, I will go for “Accepted with Minor Corrections” and is up to the Editor to consider my previous suggestion.

Author Response

see attachment

Author Response File: Author Response.pdf