Online SOPDT Model Identification Method Using a Relay
Round 1
Reviewer 1 Report
This article describes a method of identifying an online analog second-order plus dead-time model (SOPDT) used for autotuning a PID controller. The authors investigated previously used models, articles for plant models published in recent years, and previous identification techniques and proposed a more sophisticated SOPDT model-based controller.
This work develops a SOPDT model with an online identification method without taking the control system out of operation for a time-varying plant. The paper discussed the determination of four parameters of an overdamped SOPDT model. With the new method, the second-order inertial model with a time delay is determined to autotune the PID controller.
To determine all four parameters of the model, the identification method for the SOPDT model uses an occasional identification experiment using a relay in a closed-loop control system, while the autotuning process is completed with the supervisory controller. In addition, the article also discussed the determination of the steady-state gain, measurement of a time-delay constant, determining of limit cycle parameters, calculation of the time constants, examples of identification of the second-order time delayed plant based on simulation results, the parameter comparison with other methods, robustness, etc.
More detailed comments are given below:
Line 6: “a method of identify identifying online”, typo
Line 35, the authors mention that “in many cases this model was not adequate for the properties of the plant, which could lead to mistuning of the controller.” Evidence or references are needed for the cases of mistuning.
Line 349, Title 3.2: “Comparition with other methods". Typo
Line 411-421: The discussion regarding pros and cons of the proposed method. “more precise determination of this model parameter. possibility of dispensing in the second stage of the experiment with the measurement of the amplitude”, ”disadvantage may be the slightly longer time duration of the extended identification experiment". Suggest a data-based or tabulated comparison between models in a more explicit way. Table 1 in section 3.2 only compares model parameters. A more thorough discussion is needed to compare advantages and disadvantages to better justify the motivation for proposing the new method.
In addition, the language of this paper needs to improve. Many typos and grammatical mistakes.
Author Response
Response to the Reviewer 1.
Thank you for your kind review of my manuscript entitled “On-line identification method of SOPDT model using a relay” sent to the editors of the Journal „Applied Sciences”. The comments received served as a guide for improving my manuscript.
Line 6: “a method of identify identifying online”, typo
This was an ugly mistake. MDPI AutorService changed it to
“a (novel) method for the on-line identification…”
Line 35, the authors mention that “in many cases this model was not adequate for the properties of the plant, which could lead to mistuning of the controller.” Evidence or references are needed for the cases of mistuning.
On this, SOPDT usually captures the properties better than FOPDT is reported by Liu, Tao, Qing-Guo Wang and Hsiao-Ping Huang ( 2013). They consider these issues in detail in their comprehensive article published in the Journal of Process Control (Vol. 23, Issue 10).
I have therefore added the relevant literature reference and then amended the other references throughout the all text.
Of course, this need not always be the case. Some plants will always be modeled using first-order models, even if there are opportunities to create higher-order models.
However, I changed the word 'mistuned'. My reasoning was that if the FOPDT model is reflecting the properties of an object less adequate than the SOPDT model, then a controller tuned on the basis of any model-based rule will also control this plant worse. Worse does not necessarily mean bad, however, and 'mistuned' could be perceived as such.
Line 349, Title 3.2: “Comparition with other methods". Typo
Corrected word in Title of Section 3.2 into “Comparision”
Line 411-421: The discussion regarding pros and cons of the proposed method.
“more precise determination of this model parameter.
possibility of dispensing in the second stage of the experiment with the measurement of the amplitude”,
”disadvantage may be the slightly longer time duration of the extended identification experiment".
In industrial applications, the autotuning method is introduced in newer generation controllers (e.g. Siemens S5-1200). I do not have access to the manufacturer's data (I guess it is a trade secret), but on the basis of a study of the effects of autotuning, it can be assumed that a method of generating oscillations by means of a symmetrical relay is applied, while the settings themselves are determined on the basis of ensuring the required stability margin, i.e. directly on the basis of knowledge of limit cycle parameters. When writing about the longer duration of the identification experiment, I mean that it is extended by an additional stage of accurate measurement of oscillation parameters, which is missing in these implemented methods.
Suggest a data-based or tabulated comparison between models in a more explicit way. Table 1 in section 3.2 only compares model parameters.
Obviously, comparing parameters alone does not tell us everything about the quality of the model created. Therefore in addition to the model parameters themselves, integral ratios are also compared. I have also added a Nyquist plot to evaluate the adopted time delay. First of all the parameters estimated of this model can be used for autotuning a PID controller. I have prepared some examples of using the model obtained by this method for controller tuning. If the distinguished Reviewer finds it necessary and the Editor allows me to increase the volume of the manuscript, I will gladly include such a Section in this paper. However, in my opinion, this may diffuse the topic, as the impact of the model on the control performance also depends on the tuning method chosen.
A more thorough discussion is needed to compare advantages and disadvantages to better justify the motivation for proposing the new method.
I have tried to make the ideas behind the research on this problem better highlighted both in the text and in the conclusion.
The motivation I have given (Page 3) points to the development of a method that is simple to calculate.
And it seems to me that this is what it is. The comparison on the merits is therefore soft.
When 4 model parameters are needed to determine a more complex model like the SOPDT, more complex means are required to achieve this (such as, for example, various signal transformations including
Fast Fourier Transformation A-locus functions, calculations by numerical methods, state-space approaches etc.) I do not consider these difficulties in detail, I do not compare them with my method, where we measure four quantities and then substitute them into the formulae (18, 19) and above all, I do not question the quality of these methods.
As for the measurements, I am describing those that should be more accurate, but the standard ones do not make the method useless, they only reduce the accuracy of the resulting model, and only to a moderate degree.
In addition, the language of this paper needs to improve. Many typos and grammatical mistakes.
I believe that most of the errors have been corrected. Unfortunately I made some of the corrections after Proof Reading
best regards
Author
Author Response File: 
Author Response.doc
Reviewer 2 Report
The paper concentrates on an on-line identification method of SOPDT model using a relay. The effectiveness of the proposed method is verified by the comparative results. Overall the work is quite interesting; however, the reviewer has some concerns regarding the contributions of the paper as follows.
1. The advantages of the proposed method of this paper should be more highlighted. What is unique and novel? What is being done better?
2. The authors should clearly state the limitations of the proposed method in real applications.
3. The proposed ideas with unique features compared to existing/published papers should be more highlighted.
4. There are far too many dry formulas listed in Section 2.5. Additional corresponding explanations are suggested for inclusion.
5. Three tested systems used in the simulation are too small. The real-world systems should be used to evaluate the effectiveness of the proposed method.
6. Through additional explanations and discussions of the findings, the efficacy and benefits of the work are summarized.
7. What are limitations of the proposed method? And state some future direction, which may help other researchers.
Author Response
Response to the Reviewer 2.
Thank you for your kind review of my manuscript entitled “On-line identification method of SOPDT model using a relay” sent to the editors of the Journal „Applied Sciences”. The comments received served as a guide for improving my manuscript.
- 1. The advantages of the proposed method of this paper should be more highlighted. What is unique and novel? What is being done better?
I have slightly modified Abstract by briefly outlining the path to the model mentioned in the title of the paper. In section 2.5, I also stressed that in this method, unlike in the others, the measurement of oscillation amplitude is unnecessary. However, it may be useful in view of the calculation of the maximum allowable time delay. In Conclusion I list the features of this method which I want to obtain. I have tried to highlight them more.
.
- 2.The authors should clearly state the limitations of the proposed method in real applications.
I have added an extensive section to Chapter 3.3 Robustness which, in this context, outlines the problems involved in implementing such identification in real systems.
- The proposed ideas with unique features compared to existing/published papers should be more highlighted.
The models that were determined using the relay method ( FOPDT, critical damped SOPDT or High Order Plus Dead Time, simple integrated models have three unknown parameters. In order to determine them by taking a direct measurement of one of them, it was sufficient to determine one point of the Nyquist plot of the plant and adjust the other two model parameters to coincide with the model characteristic point. Relevant formulas to achieve this were presented in the literature. When four model parameters are needed to determine a more complex model like the SOPDT, more complex means are required to achieve this (such as, for example, various signal transformations including Fast Fourier Transformation A-locus functions, calculations by numerical methods, state-space approaches etc.) I do not consider these difficulties in detail, I do not compare them with my method and above all I do not question their quality. I am listing them for the sake of comparing the complexity of the identification method and pointing out my motivations ( Page 3) to develop a method that would not require these different techniques. And it appears that I have succeeded in doing so.
- There are far too many dry formulas listed in Section 2.5. Additional corresponding explanations are suggested for inclusion.
I have added some clarifications on the calculations carried out in this section.
- Three tested systems used in the simulation are too small. The real-world systems should be used to evaluate the effectiveness of the proposed method.
Really, they are too small. I have chosen above all those that allow the effectiveness of the method to be documented under certain conditions and also so that the results obtained can be compared with the work of other researchers. To extend this list, I have reached for the case of a high-order plant with dead time, stable zero and varying poles and, above all, with a small ratio of delay time to the dominant inertia time constant.
Obviously, the application of the method in a real system, even if not very complicated, allows a more complete verification of the proposed method. This was lacking in this work. We have plans to implement it to identify a model of a system of two/three free-flowing tanks of different shapes, but the actuator system is being replaced at the moment and we would not be able to carry out relevant tests in this regard within a reasonable time.
- Through additional explanations and discussions of the findings, the efficacy and benefits of the work are summarized.
The efficacy of the obtained results are discussed in more detail
- What are limitations of the proposed method?
About " A limitation of this method...." I report at the end of the additions made to Chapter 3.3
And state some future direction, which may help other researchers.
Some suggestions for directing further research on this issue are added at the end of the manuscript.
I also hope that most of the linguistic errors have been corrected. Unfortunately I made some of the corrections after Proof Reading
best regards
Author
Author Response File: Author Response.doc
Round 2
Reviewer 2 Report
This paper has been well revised. I have no further comments.
