A Holistic Framework for Developing Expert Systems to Improve Energy Efficiency in Manufacturing
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript presents a well-structured study on expert systems for improving energy efficiency in manufacturing. By integrating expert knowledge with automated data analysis, the proposed framework enhances scalability and adaptability. The inclusion of the Expert System Shell for Energy Efficiency (ESS4EE) strengthens its practical applicability. The case study in a metalworking production line effectively demonstrates feasibility. However, some aspects require further refinement to improve clarity and practical relevance:
- The paper includes a valuable comparative discussion of existing expert system development methods. It could be improved by more explicitly highlighting how the proposed holistic framework advances or diverges from these prior methods in terms of novelty, scalability, or efficiency gains.
- The connection between the personas (Table 1) and their specific contributions to each phase of the methodology could be clarified further, perhaps with a more detailed RACI matrix or similar tool.
- Expand the description of Figure 4 (component diagram) in the text. Explain how modules like the Inference Engine and Process Interface interact dynamically to enable energy efficiency recommendations.
- The paper should address potential challenges in knowledge acquisition and representation, particularly how to handle conflicting expert opinions or knowledge gaps during the development of the rule base.
- Supplement the 94.3% algorithm accuracy claim with statistical metrics (e.g., precision, recall) or comparative analysis against baseline methods (e.g., manual expert analysis). Discuss limitations of the validation scope (e.g., single factory case study).
- While the paper demonstrates the framework’s application in one model factory, consider discussing how easy or difficult it is to extend the framework to larger or different types of manufacturing facilities (e.g., batch vs. continuous processes). Real-world scalability often hinges on integration with enterprise IT systems and data pipelines, which could be a useful topic to address.
- Energy-efficiency initiatives in industry usually require some level of investment, even if mostly in organizational resources. Including a high-level cost-benefit analysis of implementing the presented expert system—estimating required man-hours, software overhead, or hardware—could help industry practitioners and researchers gauge feasibility.
- Many manufacturers already have some form of ERP (Enterprise Resource Planning), MES (Manufacturing Execution Systems), or EnMS (Energy Management Systems). A brief discussion or a case illustration of how the expert system would integrate with or complement these systems would strengthen the paper’s applied relevance.
Author Response
Thank you very much for taking the time to review this manuscript! Please find the detailed responses below:
Comment 1: The paper includes a valuable comparative discussion of existing expert system development methods. It could be improved by more explicitly highlighting how the proposed holistic framework advances or diverges from these prior methods in terms of novelty, scalability, or efficiency gains.
Response 1: We aimed to address the novelty, scalability and efficiency gains to prior methods in the introduction (Section 1) and in the summary (Section 6). "While previous research has demonstrated the potential of ESs in this domain, their development remained largely unsystematic, application-specific, and difficult to generalize, limiting scalability and transferability. This work addresses these limitations by introducing a methodology, that enables structured development while remaining adaptable to different manufacturing environments". More details are provided in Section 1.
Comment 2: The connection between the personas (Table 1) and their specific contributions to each phase of the methodology could be clarified further, perhaps with a more detailed RACI matrix or similar tool.
Response 2: Thank you for what we consider to be particularly valuable input! We have integrated the concept of the RACI matrix into our framework. For this purpose, we specified the different degrees of involvement in Section 3.2 and also visualized them in Figure 3. (page 7-9, section 3.2, line 266-308)
Comment 3: Expand the description of Figure 4 (component diagram) in the text. Explain how modules like the Inference Engine and Process Interface interact dynamically to enable energy efficiency recommendations.
Response 3: Thank you for pointing this out! Previously, in Section 4, we focused solely on the structure of the architecture. We added a paragraph to describe their interaction in the deployed state (page 9, section 4, line 331-338).
Comment 4: The paper should address potential challenges in knowledge acquisition and representation, particularly how to handle conflicting expert opinions or knowledge gaps during the development of the rule base.
Response 4: Challenges in knowledge acquisition and representation are indeed critical aspects in the development of ESs. Handling conflicting expert opinions is addressed by the previously introduced RACI matrix, according to which there can only be one accountable person per activity who holds the final authority for decisions (page 9, section 3.2, line 306-308). Gaps in knowledge can be compensated for by using a fuzzy rule base, which provides results even if the knowledge is not completely accurate or only partially available.
Comment 5: Supplement the 94.3% algorithm accuracy claim with statistical metrics (e.g., precision, recall) or comparative analysis against baseline methods (e.g., manual expert analysis). Discuss limitations of the validation scope (e.g., single factory case study).
Response 5: The accuracy mentioned here is the quotient of the number of correct assignments and the number of total data points. This also corresponds to the accuracy term of ML models. Furthermore, we added the information in the manuscript that the algorithm achieved this accuracy against a manual expert analysis for a representative production day for the same production line (page 13, section 5.4, line 438-440).
Comment 6: While the paper demonstrates the framework’s application in one model factory, consider discussing how easy or difficult it is to extend the framework to larger or different types of manufacturing facilities (e.g., batch vs. continuous processes). Real-world scalability often hinges on integration with enterprise IT systems and data pipelines, which could be a useful topic to address.
Response 6: Thank you for the suggestion! We have also applied the ES, but not exactly according to this framework, for batch and continuous processes (please see our own references mentioned in the manuscript). In fact, we have already successfully tested this framework in an industrial environment. We recorded data using mobile measurement devices and subsequently analyzed it using an ES. We are currently aiming to investigate the integration into a company's own IT infrastructures in the next step.
Comment 7: Energy-efficiency initiatives in industry usually require some level of investment, even if mostly in organizational resources. Including a high-level cost-benefit analysis of implementing the presented expert system—estimating required man-hours, software overhead, or hardware—could help industry practitioners and researchers gauge feasibility.
Response 7: We acknowledge the importance of evaluating the feasibility of implementing the expert system from a cost-benefit perspective! In general, through our approach we try to reduce the continuous involvement of experts, increase their efficiency in analysis and expand the knowledge transfer in this domain. This is combined with the aim of reducing the man-hours required to improve energy efficiency. Using the ESS4EE presented in Section 4, we intend to minimize the effort required to implement the software. The remaining effort can vary greatly depending on the complexity of the use case. In addition, the potential absolute benefit can also increase significantly for energy-intensive companies.
Comment 8: Many manufacturers already have some form of ERP (Enterprise Resource Planning), MES (Manufacturing Execution Systems), or EnMS (Energy Management Systems). A brief discussion or a case illustration of how the expert system would integrate with or complement these systems would strengthen the paper’s applied relevance.
Response 8: We have added a technical specification in section 3.1: "From a technical perspective, ESs can cover all levels of the automation pyramid. At the field level, they record energy and process data, which are subsequently visualized at the supervisory level. Additionally, at this level, signals can be transmitted via the human-machine interface of the ES back to the control system, which, in turn, can influence machines at the process level through actuators. At the operational level, the collected data enables the calculation of EnPIs, which serve as a basis for recommendations to optimize energy usage. Ultimately, at the enterprise level, these insights can be integrated into long-term strategic planning. The technical systems with which ESs can interact include sensors, programmable logic controllers, Supervisory Control and Data Acquisition (SCADA) systems, Enterprise Resource Planning (ERP) systems, and Manufacturing Execution Systems (MES)" (page 7, Section 3.1, line 248-258).
Reviewer 2 Report
Comments and Suggestions for AuthorsA framework is proposed in this paper for developing expert systems to improve manufacturing energy efficiency, filling a systematic ES development methodology gap and demonstrating practical applications through both a metalworking case study and the proposed ESS4EE platform. The following questions for the authors to clarify.
- This validation relies on a single case study in a specific metalworking environment. However, is the framework still applicable to other metalworking environments?
- The authors briefly describe the fuzzy logic rule base (Table 2), but the process of defining membership functions, linguistic variables, and rule priorities lacks transparency. This limits reproducibility. A step-by-step workflow for rule creation is recommended.
- The authors describe the roles (Table 1), but lack operational details. A workflow diagram is recommended to illustrate the interactions between roles during ES development.
- The article uses appropriate technical vocabulary and relatively accurate grammar. However, some sentences, such as the one in the fifth paragraph of the introduction, are too long and complex.
The English could be improved to more clearly express the research.
Author Response
Thank you very much for taking the time to review this manuscript! Please find the detailed responses below:
Comment 1: This validation relies on a single case study in a specific metalworking environment. However, is the framework still applicable to other metalworking environments?
Response 1: While this validation focuses on a specific metalworking environment, the framework is designed to be adaptable to other metalworking contexts. We have also applied ESs—though not strictly following this framework—to batch and continuous manufacturing processes (as referenced in our manuscript). Additionally, we have successfully tested this framework in an industrial setting, where data was collected using mobile measurement devices and analyzed via the ES. As a next step, we aim to investigate its integration into existing company IT infrastructures to further evaluate its scalability and applicability across different manufacturing environments.
Comment 2: The authors briefly describe the fuzzy logic rule base (Table 2), but the process of defining membership functions, linguistic variables, and rule priorities lacks transparency. This limits reproducibility. A step-by-step workflow for rule creation is recommended.
Response 2: We acknowledge that the execution related to the definition of membership functions, linguistic variables, and priority numbers is not described in detail in the manuscript. The specific assignment follows a similar approach to the selection of suitable EnPIs, relying on expert knowledge in this field, which is elaborated in detail in Ref. [46]. The complete implementation of our use case can be found in Ref. [47].
Comment 3: The authors describe the roles (Table 1), but lack operational details. A workflow diagram is recommended to illustrate the interactions between roles during ES development.
Response 3: Thank you for what we consider to be particularly valuable input! We have integrated the concept of the RACI matrix into our framework. For this purpose, we specified the different degrees of involvement in Section 3.2 and also visualized them in Figure 3. (page 7-9, section 3.2, line 266-308)
Comment 4: The article uses appropriate technical vocabulary and relatively accurate grammar. However, some sentences, such as the one in the fifth paragraph of the introduction, are too long and complex.
Response 4: We appreciate the reviewer’s feedback regarding sentence complexity. Unfortunately, we were unable to identify any sentences that were clearly not reader-friendly. However, we have tried to take this into account in all further additions and changes to the manuscript.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article is interesting in that it uses an old and well-established methodology such as expert systems in contemporary concerns such as the issue of reducing CO2 emissions. Its specificity in relation to the chosen German sector somewhat affects the eventual generalization of its use, despite this the potential for applications is somewhat significant.
A regenerative AI approach would be interesting, this would add to the points of view.
Author Response
Thank you very much for your thoughtful review and constructive feedback! We have focused on the German industrial sector because it exemplifies the pressing and complex challenge of improving energy efficiency. However, the proposed framework is applicable in any industrial setting where energy efficiency improvements can be justified on economic and/or ecological grounds.
We chose the expert system approach because its inference mechanism is transparent and easily comprehensible, facilitating knowledge transfer and user acceptance compared to more complex AI techniques. As noted in the summary and conclusion, we recognize the potential of advanced AI methods, particularly large language models, for knowledge formalization and the expansion of the knowledge base.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThis version can be considered for publication.