Development of an Automatic Elastic Torque Control System Based on a Two-Mass Electric Drive Coordinate Observer

Round 1

Reviewer 1 Report

A practical engineeing problem is concerned using the proposed solution, it is interesting, some comments should be considered.

1.The contribution of this paper should be more clear. 

2.The proposed control solution should be discussed with existed results such as Wang Junxiao, Zhao Lei, Yu Li. Adaptive Terminal Sliding Mode Control for Magnetic Levitation Systems With Enhanced Disturbance Compensation.IEEE Transactions on Industrial Electronics,
2021,68(1),756-766.

3.The figures and description should be improved.

Author Response

1. The contribution of this paper should be more clear.

The following paragraph has been added to the introduction to explain the contribution of the paper. 
Study presented herein substantiates and implements a concept of developing algorithms that solve specific problems and are readily implementable on the existing equipment without need for additional computing devices.  The contribution of the paper consists in stating and solving the problem of developing and testing an automatic elastic torque control system for the shaft of a heavy-duty rolling mill. This system has been implemented as algorithms in the software of the existing industrial controllers (PLCs). It is simple and performs well. It does not need additional sensors or computers to be implemented; nor does it rely on complex computational algorithms. Such algorithms are based on computational tables that require a priori data on numerous process parameters. In our literature review, we have not come across any industrial implementation of such algorithms on hot-rolling mills.  
 In addition, Section 5 has been substantially rewritten. More paragraphs have been added to Section 7. Conclusions to explain the results of the study.

2. The proposed control solution should be discussed with existed results such as Wang Junxiao, Zhao Lei, Yu Li. Adaptive Terminal Sliding Mode Control for Magnetic Levitation Systems With Enhanced Disturbance Compensation.IEEE Transactions on Industrial Electronics,
2021,68(1),756-766.

To address this point, Wang Junxiao’s and Yu Li’s contributions to the theory of automatic control systems is now duly noted in Section 1 Introduction. The paper now refers to their works and analyzes them briefly. We comment how these authors have carried out excellent research and made important findings. However, their findings do not apply to the existing rolling mills without more research and adaptation of the algorithms. This is due to the complexity of the rolling process. A priori data on numerous process parameters, which are subject to change, would be required to implement the proposed control algorithms.

3. The figures and description should be improved.

Figures 5, 6a, 9, 10, 14 have been improved. Designations altered for Figures 3 and 15. Figure captions have been drastically improved.
Texts have been enhanced.

Reviewer 2 Report

Dear Authors,

In this manuscript, a 2-mass electric drive coordinate observer has been developed for an automatic elastic drive coordinate observer. After a close review, the points to consider and clarify have been listed below:

1- The manuscript is generally well written and the problem in question has been discussed in detail. However, it would be better to address the existing observers used for 2-mass systems. Also, most of the cited papers are a bit old; therefore, it would be better to cite recent papers.

2- The authors state that “All described research show the results of mathematical modeling and experimental tests run on special laboratory benches. These control algorithms are of scientific interest; however, they cannot be used to directly control industrial electromechanical systems that sustain shock (impact) loads.” In this perspective, it would be better to compare the proposed observer with a few observers used for the same purpose in terms of computational complexity and estimation performance. This clearly improves the quality of the manuscript. I think, except for AI-based observers, there is no problem with the computational load thanks to today's low-cost and high-capacity chips. For example, Kalman filters, which suffer from computational complexity, can execute within a few hundred microseconds.

3- All parameters (mechanical or electrical) have been assumed as constant in the developed observer. As known, these parameters may change during the operation. How do you guarantee stability for all operating conditions?

4- What is the platform used for the implementation of the proposed system? What is the specification of the experimental setup? How did you get the results? What is the sampling time? Considering all and more, it would be better to give more information about simulations and experiments.

5- In the second plot of Figure 8(a), the y-axis should be extended, because the results are cropped. Also, some parts of a few figures are of low quality. Please replace them with better quality ones.

Considering the above, I think that the manuscript is not ready for publication and should be revised in line with the comments.

Author Response

1. The manuscript is generally well written and the problem in question has been discussed in detail. However, it would be better to address the existing observers used for 2-mass systems. Also, most of the cited papers are a bit old; therefore, it would be better to cite recent papers.

Corrections have been made to the text. References to outdated papers removed. Section 1 now comments on Wang Junxiao’s and Yu Li’s contributions control theory and the development of observer-based automatic controls. References to newer papers added, their contents duly reviewed. These are papers [4], [5], [16]–[27], and [48] in the new bibliography. They deal with the existing observers used in 2-mass systems.

2. The authors state that “All described research show the results of mathematical modeling and experimental tests run on special laboratory benches. These control algorithms are of scientific interest; however, they cannot be used to directly control industrial electromechanical systems that sustain shock (impact) loads.” In this perspective, it would be better to compare the proposed observer with a few observers used for the same purpose in terms of computational complexity and estimation performance. This clearly improves the quality of the manuscript. I think, except for AI-based observers, there is no problem with the computational load thanks to today's low-cost and high-capacity chips. For example, Kalman filters, which suffer from computational complexity, can execute within a few hundred microseconds.

The proposed observer is compared to its counterparts covered by Wang Junxiao, Zhao Lei [21–25].  References to other publications [42, 43] added, papers [42–47] noted. 
Sections 1 and 7 have been complemented as follows. 
Advantages of the proposed development are noted. These include: ease of implementation, the ability to be implemented on existing rolling mills without using additional computing devices or sensors. Known developments tend to use computational algorithms based on computational tables. Their use requires a priori data on numerous rolling process parameters. However, these parameters tend to change in real-world industrial settings. Changes may occur when switching to rolling another set of products or even between passes (compressive parameters, metal stiffness, rolling speed, temperature, etc. are all subject to change). Under the proposed concept, all the required readings are sampled from the sensors of electrical parameters, which are part of the frequency converters. The algorithms are fully implemented in programmable logic controllers (PLCs) that are already present on the mill. This is an undisputable advantage of the solution.
The proposed solution has been pilot-tested on a state-of-the-art plate mill, Mill 5000, designed and constructed with assistance from SMS Group.  On the other hand, the authors have not come across any reports on the industrial implementation or testing of known algorithms on existing hot-rolling mills. This is another proof of the practicality and prospects of industrial implementation of the authors’ solution.

3. All parameters (mechanical or electrical) have been assumed as constant in the developed observer. As known, these parameters may change during the operation. How do you guarantee stability for all operating conditions?

The authors totally agree with the Reviewer’s comment that mechanical and electrical parameters tend to change in the rolling process.  To address this comment, Section 5 has been complemented with the following text.
When configuring the observer and the automatic control system based thereon, the parameters of the 2-mass system were assumed to be constant. Change in such parameters in the rolling process was not considered. These assumptions are valid due to the following properties of the system.  
1. The system under considered is a 2-mass system where the first mass (the motor’s rotor) does not change its inertia.
2. The stiffness of the elastic shaft (spindle) does not change either. This parameter depends on the length, diameter, and the properties of the metal that the shaft is made of. These parameters do not change even when the spindle is replaced.
3. 2nd mass inertia depends on the mass of work and backup rolls, which is constant as well. The inertia of the rolled bar is 5% to 15% of the total 2nd mass inertia. For configuration, use the mean bar inertia that deviates by 1.7% to 5%. Such deviations are commensurate with the error of the instrumentation sensors. Thus, they do not cause significant error in the configuration of the control algorithms. 
Paper [40] presents an estimate of the system parameters. Successful testing and operation of this newly designed automatic elastic torque control system on an existing rolling mill is what proves the stability and reliability of the results.

4. What is the platform used for the implementation of the proposed system? What is the specification of the experimental setup? How did you get the results? What is the sampling time? Considering all and more, it would be better to give more information about simulations and experiments.

To address this comment, Section 5 has been complemented with the following text.
The system has been implemented in VME BUS and LogoCAD. Cycle time to calculate the stand motor signals is 5 ms, can be reduced to 1 ms. The system was developed and configured using a method covered in paper [49] by the authors. Using the method, we sampled data on an industrial unit using signals from the stand motor control system (i.e., the PDA system). In pilot tests, we used signals from the torque meter, whose sensor is shown in Figure 2, for accuracy testing. Figures 3, 8, 15 show the oscillograms of these signals. The signals were recorded by the IBA PDA system. Then they were exported to MatLab for analysis and observer modeling. Debugged MatLab models were used to write a program to execute the developed algorithms in the real-time computing system based on VME BUS and running VxWorks.

5. In the second plot of Figure 8(a), the y-axis should be extended, because the results are cropped. Also, some parts of a few figures are of low quality. Please replace them with better quality ones.

Y-axis in Figure 8a has been extended. Figures 5, 6a, 9, 10, 14 have been improved. Designations altered for Figures 3 and 15. Thus, the figures are now of better quality.

 

 

 

 

Reviewer 3 Report

Dear authors,

you have to drastically improve your use of English, especially with standard terms. Also, you do not possess unified designations in the paper and combined with previous, you make big confusion for readers. Digital twins are not presented as it should be, as well as identification of parameters according to recorded measurements. The paper has a big potential, since the experiments are performed in industry on high power plant, but big corrections in presentation should be made. 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Nice revision, it could be accepted.

Author Response

The authors appreciate careful reading of the manuscript. 

Reviewer 2 Report

Dear Authors,

Thank you for your response to my comments. In the revised manuscript, all my questions have been considered. I'll recommend the Editor to accept the manuscript for publication.

Author Response

The authors appreciate careful reading of the manuscript.

Reviewer 3 Report

Dear authors,

I have more comments in order to help you to improve your paper, but the use of English has to be greatly improved.

Best regards

Comments for author File: Comments.pdf

Author Response

The authors agree with the comments and thank for a close reading of the manuscript. The corrections have been done by yellow highlight (see the manuscript).