Fully Convolutional Neural Network Prediction Method for Aerostatic Performance of Bluff Bodies Based on Consistent Shape Description

Round 1

Reviewer 1 Report

The manuscript presents a deep-learning approach to assess the aerodynamic drag of objects. This method is presented as an alternative to traditionally used methods such as wind tunnels and computational fluid dynamics (CFD).

This manuscript presents a very interesting idea. It is well described and well structured. Congratulations to the authors.

Author Response

Please see the attachment

Reviewer 2 Report

The article deals with the application of ML for CFD problems. Authors propose the use of a fully convolutional neural network to predict the flow fields around a bluff body. As the inputs to the neural network authors assume the distance field combined with two additional fields Ix and Iy corresponding to x and y coordinate. The use of these additional fields brings distinguishes the proposed method from e.g. DeepCFD software. The results show that the neural network is able to predict the flow field and the aerodynamic loading coefficients with acceptable error.

I didn't found any serious errors or flaws. I would recommend add a reference to DeepCFD and to emphasize the differences from that software. Moreover, authors should remove the part of the sentence concerning "U" (line 150). "U" is not included in the equation (1), so there is no reason for U here.

Next, I was not able to understand what authors mean by "stabilizing pressure field" (line 524). Could you explain it?

Last issue concerns the Reynolds number. Does "R" at line 326 mean the Reynolds number? If so, it would be better to denote it by "Re" in order to avoid confusion with "R" in Table 1. Does the Re=4000 correspond to practical configurations?

Author Response

Please see the attachment

Author Response File: Author Response.pdf