On Numerical Simulation of Casting in New Foundries: Dynamic Process Simulations
Round 1
Reviewer 1 Report
The paper deals with the numerical simulation of dynamic casting processes is the evolving domain scheme which has been introduced to overcome problems with simulation of dynamically growing systems (e.g. continuous casting process).
In the proposed scheme, the change of the mesh, boundary and energy sources within casting process simulations have been presented and its profound effects on the computational time and resources have been examined.
The paper is appropriate for the pubblication in the Journal.
The content is interesting and new in the field of the numerical simulation of dynamic casting process. Moreover the paper is clear and well organized.
I suggest to publish as it is.
Author Response
Many thanks for the thorough evaluation of our manuscript. We are very grateful for time and efforts spend by you and the journal reviewers on careful review of our manuscript and the resulting final comments.
---------------------------------------------------------------------------------------
The paper deals with the numerical simulation of dynamic casting processes is the evolving domain scheme…The paper is appropriate for the publication in the Journal. The content is interesting and new in the field of the numerical simulation of dynamic casting process. Moreover, the paper is clear and well organized. I suggest to publish as it is.
Thanks for the review and encouraging comments.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Authors are presenting an evolving domain scheme in CFD modeling of casting processes. While the work has some elements of novelty, which may be interested to a broader readership of Metals, significant changes are needed to bring it to the level of publication.
1. The Abstract should put more emphasis on the work performed in this paper and its achievements, rather than a long intro, which should be reserved for the intro section.
2. More general background to numerical modeling in metal solidification and casting processes should be given, also why is numerical modeling important in these processes , ie. harsh conditions, prohibited cost of experiments in plants, computational challenges particularly with multiphysics coupling etc.
3. More references from this area should be cited, particularly from multiphysics background such as :
Prediction of cracks in continuously cast steel beam blank through fully coupled analysis of fluid flow, heat transfer, and deformation behavior of a solidifying shell. Metall. Mater. Trans. A 31, 2000
Numerical Method For Heat Transfer, Fluid Flow, And Stress Analysis In Phase-Change Problems. Numer. Heat Transf. Part B Fundam. 42, 437–459, 2002
Multiphysics Model of Metal Solidification on the Continuum Level. Numerical Heat Transf. Part B Fundam. 58, 371–392, 2010
Multiphysics modeling of continuous casting of stainless steel. Journal of Materials Processing Technology, 278 , 116469, 2020
4. The authors should write the governing equations they solve ie. energy equation with advection ? How the commercial code such as LS-DYNA solve it ?
5. How is the Figure 2 related to this work. It seems to be some recently published work from one of the authors in the confluence of AI and modeling. Has any of this hybrid approach used in this work, if yes, provide more details, and if not remove the figure 2.
6. More details should be given about the DMT algorithm written in a python script that works with the commercial code LS-DYNA to provide directional boundary evolution scheme, that is the major novelty of this work, provide a block diagram how it works and interacts with LS-Dyna.
7. All the computational performance is given for a single computational node. How about dual or multi-nodal performance, LS-DYNA works with MPI and it should scale wider on multi-nodes. It would be interesting if any penalty will be incurred due to inter-nodal communication cost through interconnect. If the authors don’t have access to an HPC cluster with multi computational nodes, they should say it.
Author Response
Many thanks for the thorough evaluation of our manuscript. We are very grateful for time and efforts spend by you on careful review of our manuscript and the resulting final comments. After receiving these comments, we have tried to revise the paper according to these comments, resulting in subsequent revisions within the text and also presentation of the paper. Here is a detailed description of revisions in accordance to reviewers’ comments:
---------------------------------------------------------------
Authors are presenting an evolving domain scheme in CFD modeling of casting processes. While the work has some elements of novelty, which may be interested to a broader readership of Metals, significant changes are needed to bring it to the level of publication.
Thanks for the thorough review of the manuscript. Authors appreciate the time and efforts by the reviewer for an in-depth review. Although the thermal-mechanical simulation of casting processes, presented in this paper, can be linked into a CFD simulation of melt flow and solidification, the CFD modelling of the casting process is out of scope for this manuscript. Discussions about the multi-phase CFD simulations (melt delivery, mould filling and solidification) require a full-size paper and the intention of authors in this paper is to present the new evolving technique (and its computational performance) which has been implemented on a finite element solver (LS-DYNA).
- The Abstract should put more emphasis on the work performed in this paper and its achievements, rather than a long intro, which should be reserved for the intro section.
The point about the abstract is taken and abstract text has been modified accordingly.
- More general background to numerical modeling in metal solidification and casting processes should be given, also why is numerical modeling important in these processes, ie. harsh conditions, prohibited cost…
The introduction section has accordingly been improved and a paragraph is added to the text to clarify the need for the numerical simulations.
- More references from this area should be cited, particularly from multiphysics background such as …
Three more references have accordingly been added to the reference list (at the end of the manuscript) to make the paper richer in terms of literature review for readers.
- The authors should write the governing equations they solve ie. energy equation with advection ? How the commercial code such as LS-DYNA solve it?
As it has been stated before, the research work has been focused on implementation of the new dynamic method on the finite element solver LSDYNA (as thermal-mechanical simulation, clearly described in section 4) and no explicit fluid aspects of the process are directly addressed in the work. Although, authors are working within a research team which also deals with the thermal multi-phase CFD simulation of casting processes, the work reported in this manuscript covers the finite element implementation of the technique. The long theoretical and mathematical development of the new dynamic technique and its associated equations are available from the references (by authors) which has been provided in the text and to limit the length of the manuscript, the attention is more paid into the numerical implementation and computational performance of the technique.
- How is the Figure 2 related to this work. It seems to be some recently published work from one of the authors in the confluence of AI and modeling. Has any of this hybrid approach used in this work, if yes, provide more details, and if not remove the figure 2
The point is taken and a new paragraph was added to the text to elaborate more about the use of hybrid framework on estimation of heat transfer coefficients which have been used for the casting simulation.
- More details should be given about the DMT algorithm written in a python script that works with the commercial code LS-DYNA to provide directional boundary evolution scheme, that is the major novelty of this work, provide a block diagram how it works and interacts with LS-Dyna.
Thanks for the watchful comment. The lack of diagram for the solver implementation in the text has been mended and “Figure 3” has been modified to include a diagram for a step-wise implementation of the dynamic technique.
- All the computational performance is given for a single computational node. How about dual or multi-nodal performance, LS-DYNA works with MPI and it should scale wider on multi-nodes. It would be interesting if any penalty will be incurred due to inter-nodal communication cost through interconnect. If the authors don’t have access to an HPC cluster with multi computational nodes, they should say it.
Although, the results of the benchmark study, reported in “Table 2”, is using a single node computation, all of the other results reported in Figures 7, 9, 10 and 11 are obtained using multi-nodal computations. In fact, all of the computational work for this manuscript have been performed using cluster computing with the specifications which are clearly mentioned in “Table 3”.
Author Response File: Author Response.pdf
Reviewer 3 Report
General thoughts:
1. Why was the impact of mold surface quality (roughness) on the cast not analyzed?
2. Why was the impact of mold surface coverage (coating type, coating thickness) on the cast not analyzed?
3. Does the gap formed at the joint affect the change of simulation parameters and the results obtained?Editorial notes:
Figure 9: Different scales on the Y axis.
Figure 11: Different scales on the X and Y axes
Author Response
Many thanks for the thorough evaluation of our manuscript. We are very grateful for time and efforts spend by you on careful review of our manuscript and the resulting final comments. After receiving these comments, we have tried to revise the paper according to these comments, resulting in subsequent revisions within the text and also presentation of the paper. Here is a detailed description of revisions in accordance to reviewers’ comments:
- Why was the impact of mold surface quality (roughness) on the cast not analyzed?
The focus of the manuscript is on computational performance of the new dynamic technique which has been implemented on a finite element solver for casting processes. Although, aspects of mold contact with billet during casting process and its effect on heat transfer coefficient (and billet surface quality) are important, for the work presented herein these effects have only been considered implicitly in the thermal heat transfer from billet to mold.
- Why was the impact of mold surface coverage (coating type, coating thickness) on the cast not analyzed?
As stated in the last response, the discussion about the detailed casting process specifications and its effects on the quality of billet, although important, is out of scope for this manuscript. The paper’s main intention is to describe the computer implementation of the new evolving technique for casting processes and its computational performance against the conventional methods. Hence, it cannot address all details of casting processes which are followed in the foundry practices.
- Does the gap formed at the joint affect the change of simulation parameters and the results obtained?
The mechanical gaps which appears during casting of billets can affect the calculation of contact heat transfer coefficient which would ultimately be represented in the thermal energy transfer during the simulation. In this sense, the gaps are affecting the estimation of the HTCs which are used as a boundary values for the numerical simulation during the casting process.
Figure 9: Different scales on the Y axis.
“Figure 9” shows the variation of the element processing time (in seconds) for different billet lengths and different number of computing nodes. The Y axis on these graphs show the solvers element processing time for mechanical and thermal calculations. Different time durations on these axes are due to different number of degrees of freedom for mechanical and thermal elements.
Figure 11: Different scales on the X and Y axes
The numbers on X and Y axes for these graphs are corresponding to different billet lengths in meters and Input / Output times in seconds, respectively. Since we have different units on these axes, naturally, values at different orders of magnitude can be expected.
Best Regards,
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
No further suggestions, authors have adressed all of concerns.
