Multi-Zone Integrated Iterative-Decoupling Control of Temperature Field of Large-Scale Vertical Quenching Furnaces Based on ESRNN
Round 1
Reviewer 1 Report
High precision and high uniformity temperature control of large-scale temperature field is of great importance for the thermal treatment process and has gradually attracted the attention of researchers in related fields. This paper proposes a strategy to control the temperature of a large scale vertical quench furnace, which fits well with the theme of this journal.
The method is based on a spatial temporal dimensional extrapolation method that can predict the workpiece temperature, and then combining an iteration control based on self-incentive and an integrated intelligent decoupling control based on a neural network to realize high precision and uniformity temperature control in the furnace. The problem is interesting and the manuscript is well organized. It provides a reliable and novel idea for intelligent temperature control.
However, there are still some issues that need to be addressed before this article can be published. The suggestions are as follows:
(1) There is a temperature model of the workpiece. It is a set of ODE in Eq.(6) that can get the temperature distribution of the workpiece. Is it the dynamic equations that describe the dynamics of the system? or, can we say it is the transfer function matrix of the controlled system? If not, what is the difference?
(2) What is the rationale for the decoupling approach? As far as it is noticed from Fig.1, the working chamber is separated/isolated from the heating chamber. This leads this reviewer to question what is the purpose of parameter auto-tuning of controllers based on local models;
(3) The PID controller in the paper is of ideal type -- no filter is used for the derivative? Is it possible in practice? Will PI controller be enough for a good control?
(4) Compare the performance of an intelligent PID control system with a conventional PID, the result only could be one, the conventional will be worst, but it is very easily to implement and cheaper. Comment and try to demonstrate that the increase performance of an intelligent PID control can reduce the coast (what is the payback) of the industrial process (price of the equipment for the both system control).
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Line 83: Please provide the reference on which zonal model is used.
Line 138: Paragraph dedicated to the structure of the paper is unnecessary and can be ommitted.
Lines 35 and 60: Authors could include the information about radius of the furnace from line 60 in the line 35, to help readers with initial understanding of the investigated furnace.
Line 170: Please explain the parameter petting module, might be translation error.
Please better explain the sentence starting at line 196, not sure if the non-participant air is proper term to use there. Should it refer to leaking hot air through the furnace seals? Or does it refer to hot air being considered fully transparent, thus not affecting radiative heat exchange? Please consider restructuring that entire paragraph to better clarify this.
Please add word heat to sentence at line 199, to clarify it is a heat conduction model through a solid workpiece.
Line 211: A reference is missing in the brackets.
Line 212: Is the equation 4 interpolation function or eigenvalue matrix? Would you please explain how the equation 5 is derived? Does it originate from the equation Tp*Np = sum(Tneighbour*Nneighbour)?
Line 219: brackets around (1)
Please better explain the sentence at line 253. Does it refer to a faster convergence of the results with coarser time step?
The authors go directly to the simulation results from the theoretical background. While the experimental results are presented in the results, an intermediate section on models validation and verification would be nice. Even if compares already presented simulations data and esitmates its deviations from experimental. This would requiere insertion of a new heading, and slight rework of results section.
In figure 7 there are two peaks in zone 7, one of them being after the system enters quasi-stationary state. Could you discuss the origin of these deviations? I found explanation at line 521, but some initial details should be given when discussing figure 7, so the reader would understand why this peak occured in place where it was relatively unexpected.
In figure 8, instead of using words Traditional and Proposed it would be more useful to put PID and ESRNN in legend headings.
Headings of sections 6 and 7 overlap: Results and discussion, and Discussion and conclusion. Please consider removing the discussion and leaving it as Conclusions.
Question - no need to replay: Do the authors need to include both funding sources in Acknowledgements also, or stating them in Funding is sufficient?
Fort the most part the language quality throughout the manuscript is good. However, some editing should be done, mostly considering the structure of sentences, and in few places use of improper terms.
Please mind the use of prepositions in some of the sentences.
In several places sentences feel like they were directly translated to English, thus making them sound slightly odd.
Please consider using "During the temperature rising period" instead of the "In the temperature rising period", as those sentences often refer to a concurent processes, and it refers to a range in time.
Would a term "control function" or "control data" be a better suitable instead of "control law"?
The sentence in line 193 should be split in two. Also the second sentence should start with "The heat is provided by" or some similar structure. "The heat of which comes" is incorrect.
However is improperly used in sentence in line 194.
Line 218: Does the "discrete meshes" refer to "individual nodes" or "individual cells"
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
In this paper, an adaptive control algorithm is presented based on AI methods, to be employed at multiple heating zones of vertical quench furnaces. The decoupling problem of multiple control loops is redefined into a multi-loop integrated optimization control problem. A novel RNN architecture named: “eigen-vector self-update recurrent neural network (ESRNN)” is developed to tune the “Conventional PID controller” parameters.
Line 153: You can bring couple of examples for these external influences.
Line 211: the reference is missing, should be added.
Line 230: Add a reference for “Euler polyline” please.
Line 245: Add a reference for “Romberg algorithm” please.
Line 261 & 264: Can you give more explanation of the R (specific value?), what is the basis for selection of q=3 and R=2?
Line 295: Add a reference for “Lipschitz continuity” please.
Line 299: What is the reference of Lemma 1? What is MPPD and its reference?
Line 301: the MPPD “Mimetic Pseudo Partial Derivative” needs more clarification
Line 324: How this optimal condition is defined?
Line 374: Add a reference for “Ky Fan theorem” please.
Line 476: The furnace door accidentally falls. Why didn’t you carry out the measurements for another batch without such an incident? What’s the explanation for temperature drop at 14700s?
Line 539: Table 2. What is Average Tuning Time? And what is 104s?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
