Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article addresses methods for applying Digital Model-Predictive Control Systems as Intelligent Agents. Unfortunately, the manuscript does not explain why labeling Digital Model-Predictive Control Systems as Intelligent Agents provides any additional beneficial properties or functional advantages. The title should therefore be reformulated or appropriately justified within the text.

Based on the content of Section 1 (Introduction), it appears that the purpose of the paper was not to solve a clearly defined research or technical problem, but rather to demonstrate the operation of the system developed by the authors. The manuscript states that the proposed system can perform the functions of digital twins, optimize controls, provide pre-training for process operators, and facilitate mutual adaptation between operators and the Intelligent Agent. However, none of these claims are substantiated within the article; no rigorous verification process supporting these assertions is presented.

For example, Section 3.3 (Case studies) states that verification was carried out in the Honeywell UniSim® Design (USD) environment and that Figure 3 demonstrates moderate overshoots, low overall control variability, and satisfactory attainment of target trajectories. Unfortunately, Figure 3 contains only two screenshots and does not confirm any of the values or parameters mentioned. The type of simulation model of the controlled process is not described—there is not even an indication of whether it is linear or nonlinear, stationary or non-stationary. The manuscript does not mention how stability was assessed, how robustness to changing conditions was evaluated, or any similar methodological details.

The article should clearly formulate the specific research problem addressed by the authors. A review of the current state of the art relevant to this problem is required. The methodology of the conducted research must be presented precisely, in a way that demonstrates that the proposed approach meets the authors’ stated objectives. Moreover, the results should include quantitative performance measures that demonstrate that the proposed method is competitive compared to existing solutions.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents an intelligent-agent–based control concept combining identification algorithms, JITL models, a digital twin, and operator-focused modules. The topic is relevant, but the paper in its current form lacks clarity, structure, and sufficient scientific rigor. Several essential components are missing or underdeveloped.

Below are the key issues that must be addressed before the paper can be considered for publication.

Major Concerns

1. Abstract lacks substance.

The abstract does not clearly define the problem, the contribution, or the results. It reads descriptive rather than scientific. The authors should revise it completely to include:

• research gap,
• objective,
• methodological outline,
• key findings,
• relevance to sustainability.

2. Introduction is unfocused and insufficiently argued.

The introduction does not identify what is missing in existing literature or why the proposed approach is needed. The authors must:

• articulate a clear research gap,
• explicitly state research questions or objectives,
• justify the relevance of IA-based control in the context of sustainability.

3. Methodology and system structure are not rigorously described.

Key components (associative search, JITL architecture, IA functions, personal adaptation) are presented conceptually, but not technically. The manuscript lacks:

• unambiguous definitions,
• algorithmic clarity,
• implementation details.

4. Case study is inadequately explained and unconvincing.

The ball-mill example is too superficial to serve as a validation. Critical information is missing, including dataset details, evaluation metrics, and any comparative analysis. Without these, the case study does not demonstrate the effectiveness of the proposed system.

5. Operator pre-training and adaptation modules remain speculative.

Both modules are described in a visionary manner, with no methodological grounding, no measurable framework, and no validation plan. These sections must be reformulated or significantly expanded.

6. Conclusion is too general and does not reflect scientific contribution.

The conclusion should summarize the main contributions, limitations, and specific future work directions. In the current form, it does not provide a meaningful closure.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents a compelling integration of Just-in-Time Learning (JITL) models, specifically those built with associative search algorithms, into a Model-Predictive Control (MPC) framework, framing the resulting system as an Intelligent Agent (IA). The concept of using these dynamic, inductive models for control, digital twinning, and operator training is innovative and has significant potential for industrial applications. The article is generally well-structured, but several sections require clarification, expansion, and stronger empirical validation to strengthen the overall argument and contribution.

The following problems were found in reading.

• The abstract should more clearly elaborate on the existing gaps or deficiencies in the current research. The innovation of the method proposed should also be elaborated.
• The definition of "intelligent agent" in the Introduction section is rather genera. Why such systems can be considered IA? What the differences from existing IA systems in conjunction with the control system structure proposed in this article.
• For Section 2, when introducing the association search algorithm, there is a lack of explanation of the computational complexity and the real-time guarantee mechanism. Please add the feasibility of the algorithm in real-time control, especially in the high frequency data environment.
• The derivation process of equations (2) to (10) is relatively intensive. Please add textual explanations, and ensure that each symbol is to be clearly defined.
• In Section 3.2, it is mentioned that "Identifiers in feedback loops of process control systems can underlie a new type of digital twins (DT)". What the differences and advantages between it and the traditional analog DT? A complementary comparative analysis is suggested to highlight its advantages in dynamic adaptability.
• For Section 3.3 to 3.5, the content lacks a concrete implementation framework or preliminary experimental verification. Please give simplified implementation examples or simulation experiments, and analyze the results.
• The format of the manuscript (such as uniform paragraph indentation) needs to be further checked to ensure compliance with journal requirements.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have revised the manuscript in response to the reviewer’s comments. The previously raised concerns have been addressed satisfactorily, and the revisions significantly improve the technical content and overall clarity of the paper.

Most of the comments have been fulfilled. However, a few issues still require attention before final acceptance, as outlined below.

First, the authors should consider a further revision of the manuscript title. It should explicitly indicate that the primary subject of the paper is the Intelligent Agent, and that the implementation of Model-Predictive Control constitutes only one of its functions. This clarification would improve consistency between the title and the actual scope of the work.

Second, there remains an inconsistency in the notation of mathematical operators. Specifically, in line 322 the authors use a transpose operator, whereas in Equation (15) a complex conjugate (Hermitian) operator is applied. The authors are encouraged to ensure consistent and correct use of mathematical notation throughout the manuscript and to clearly state whether the variables involved are real-valued or complex.

Third, the chosen verification method—namely, the comparison of the proposed control algorithm with a PID control loop tuned using the Internal Model Control (IMC) method—is considered appropriate and technically justified. However, the presentation of the comparison results in Section 3.3.13 (Efficiency metrics) is currently difficult to read and interpret. To improve clarity and scientific rigor, the results should be presented in a structured form, such as tables or figures, enabling a clear and direct comparison of performance metrics between the proposed approach and the reference controller.

An additional comment concerns Sections 3.4 and 3.5. The authors may wish to reflect on whether the material presented in these sections optimally fits the main focus of the manuscript. In their current form, these sections appear to be less thoroughly developed from a scientific perspective than the preceding parts of the paper. The content might therefore benefit from further elaboration, potentially in a separate, more comprehensive publication.

With the above minor revisions implemented, the manuscript would be suitable for publication.

Author Response

Please see the attachement

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. Dear Authors,
   the manuscript has been substantially improved compared to the previous version, and the efforts invested in enhancing the structure, methodological clarity, and presentation of the results are clearly evident. Nevertheless, in order to fully meet the standards of the journal and further strengthen the scientific contribution of the work, the following issues should still be addressed:

   1. Clearer Emphasis on the Sustainability Contribution
      Although the benefits in terms of control stability, reduced oscillations, and decreased equipment wear are well demonstrated, their connection to sustainability remains largely implicit. It is recommended to more explicitly articulate—particularly in the Abstract and Conclusion—how the proposed approach contributes to energy efficiency, emission reduction, and extended equipment lifetime.

   2. More Explicit Positioning with Respect to Existing Approaches
      The manuscript would benefit from a clearer positioning relative to contemporary data-driven and machine-learning-based control methods. In particular, emphasizing the advantages of the proposed approach in terms of interpretability, real-time applicability, and robustness under industrial conditions would help to better highlight its novelty and practical relevance.

   3. Improved Methodological Transparency and Readability
      While the methodological framework is generally sound and well described, certain implementation choices (e.g., the selected range for the number of associations and the transition from identification performance to closed-loop control evaluation) could be explained more clearly. Minor improvements in terminology consistency, section transitions, and visual emphasis of key results would further enhance the readability of the manuscript.

   Considering these points, the manuscript has strong potential to be positioned as a valuable contribution to the field of intelligent control systems for sustainable industrial production.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

It can be published.

Author Response

We sincerely appreciate your thorough review. Your comments and recommendations were extremely helpful and allowed us to improve the quality of our manuscript.