Application of Artificial Intelligence Technologies for Diagnostics of Production Structures

Round 1

Reviewer 1 Report

Broad comments. Using Artificial Intelligence methodologies for fault diagnosis is an idea that has been under discussion recently, mainly due to limitations in modeling and theoretical approaches of such complicated dynamical systems. In general, the text is well structured and has clearly defined topics, but there are some drawings regarding the motivation, impact, and innovation of the work. Authors should pay attention to describing the added value of their effort.

Specific comments. In general, the text is very well structured and has clearly defined topics. Some comments for improvement:

1. Since the work refers to a specific use case (torpedo ladle cars) the authors could consider refining their title to be more descriptive of the work.
2. The abstract is descriptive of the method and results of the manuscript. Authors should consider refining the abstract such that the approach followed is part of a generic issue.
3. The authors have made a concise overview of the topic and a brief reference to the existing literature in the introduction. As a general drawback, I could say that there is no reference to similar approaches (e.g. [1]) where the accuracy of machine learning methodologies has been performed on real datasets in vessels.

[1] Theodoropoulos, P., Spandonidis, C. C., Giannopoulos, F., & Fassois, S. (2021). A Deep Learning-Based Fault Detection Model for Optimization of Shipping Operations and Enhancement of Maritime Safety. Sensors, 21(16), 5658.

• The authors are advised to refine the first section such it addresses the following:
• What is the research question on hand?
• Importance of the issue (more or less is well described in the text)
• Who else worked on the issue and what are the limitations/borders?
• What is the approach of the present work?
• What is the innovation of the approach?

5. The authors could consider adding a paragraph describing the structure of the paper at the end of section 1.
6. It is emphasized that the motivation and impact of the work, which is believed by the reviewer to be very important, should be clearly described. The novelty should be clearly stated and described.
7. The authors could consider adding a flowchart illustrating the 6 stages of their process along with inputs and outputs.
8. A brief description of the multilayer perceptron and radial basis function network could be fruitful.
9. Authors should better justify the selection of 480 epochs since this is not directly derived from figure 2 and table 1.
10. The authors could provide some technical details regarding the communication and data fusion technologies included in figure 4.
11. Is figure 5 needed?
12. In section 5 the authors should justify the different numbers for test cases (a total of 60 for the new method and 108 for a standard diagnostic method).
13. Besides, the authors should better describe the last paragraph of section 5 (significance), since there are parameters not directly explained.
14. Conclusions should provide (other than the main outcomes) a brief overview of the work.

Author Response

Dear Reviewer.

Thanks for the comments. I have attached the answers in a separate file. I hope they answer all questions. Thank you very much.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors present a neural network based solution for detecting burnout zones of a torpedo ladle cars' lining, and also a software making use of the developed solution. 

The study presents an effective use of neural network based prediction model for an industrial problem. The error rates appear to be plausible for the problem. The neural models are straighforward and they are described clearly. However there are several drawbacks in the paper, so I'd suggest improving the paper for the following issues:

• The title of the paper is too general for the content. The titles emphasizes "Diagnostics of Production Facilities for Maritime Industry" however the proposed model generated prediction for diagnostics only for a particular problem.
• Related work section is very short and the AI methods used in these previous studies are not clearly described. Therefore the difference of the proposed neural model is not much clear. There are several papers by the co-authors of the submissions listed in the references list. It is necessary to describe the difference of the proposed work from the earlier works by the authors. 
• I'd suggest revising the literature review for similar studies in other industries involved in similar problems (in general manufacturing and steel industry, for instance)
• The use of FF neural network is plausible for the problem and the experiments give promising results. However there are recent deep learning architectures applied for image related prediction problems in the literature. So it is needed to give motivation and justification for selecting a classical NN based approach.
• In Table 4, it is not clear why ncount is different for standart diagnostics approach and the proposed method. It looks like different experiments are conducted for two cases and hence not clear whether the results are comparable or not. 
• In general, for the experiments, the data set information and the settings in terms of training/validation/test data sets are missing. For instance in Table 4, what is the setting for each line (each experiment), what is the overall result?

Author Response

Dear Reviewer.
I thank you for the serious and good work with our article. We have made corrections and attached a file with explanations.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Authors have covered previous comments, looking forward for their next work. 

Reviewer 2 Report

The authors responded my comments suffieciently. The analysis could be extended but I believe in the current content it can provide an insight for the audience. 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.