Database for Research Projects to Solve the Inverse Heat Conduction Problem

Round  1

Reviewer 1 Report

In the manuscript a model for random HTC function

generation is presented. 

Major remarks:

- Computational Mechanics, 52, 2, pp. 287-300, 2013 - could be mentioned

- 'the exact value of the Heat Transfer Coefficient' you can never find 'exact' HTC

- can we extend this scheme for non isotropic HTC ?

- what is the role of 'Smoothing' scheme for data ?

- in Fig. 4 HTC can be < 0 ?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In order to address the ill-posed problems in the inverse problem of HTC, the present work shows a FEM-ML hybrid method. The motivation of the work is good, and the it is written in straight fashion, so easy to follow. Although I don’t like the figure draw by Excel, I would suggest the publication of this work considering following reversions:

1. The introduction is not comprehensive enough, especially for ill-posed problems. Besides the GA, ANN, PSO… these machine learning problems, the sparse representation is also crucial for title issue, the recent development of the sparse representation for ill-posed problems can be found in some works like  Mechanical Systems and Signal Processing 126 (2019) 341–367. The development of GPU for FEM and ML should also be addressed.

2. GPU accelerated simulations could be considered as the novelty of work, as it has been widely used, like the work wave propagation of laminated composite plates via GPU-based wavelet finite element method, and some simulation based on LISA/FDM on CUDA. I’d like to suggest the author highlight the contribution on the solution of inverse problem.

3. FDM is not the best numerical solution due to the boundary problem (the jagged grid), I am not sure if this problem will affect the accuracy of samples. Does the staggered grid used in present simulation? Too few details are provided for your models.

4. What’s more, the details of your ANN should be presented too. The method is parameter-sensitive.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round  2

Reviewer 1 Report

Minor remarks:

- some problems with citations are included e.g.

in answers:

ANNs  have already been used by researchers of the field,  there are superior results  in  reducing

 the material parameters for selected  material functions [Sumelka,2013].:

in the text:

ANNs have already been used by researchers  of the field, there  are superior results in

 reducing the material parameters for selected material functions [? ]:

- so e.g. reference [Sumelka,2013] is not included ?

also in l.40, l.212

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The work is systemic revised, however, some points are careless: 1. Page 2, Line 39-40, I cannot see the refs [?] in References. Also for Line 77-78. 2. The dimension of some figures should be reduced for publication. 3. Fig. 1, please locate the label of x-axis and y-axis on common positions. 4. Pay attention with your typeface in Figures. It is suggested to use “Palatino Linotype” for this journal. 5. Pay attention of your figures, the legends and labels of axes are missing. 6. What do you mean by [12? 13] in Page 8, Line 212. 7. Fig. 3. The label is too small. Please give me a magnifier for checking. It needs further modification.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf