Ball Mill Load Classification Method Based on Multi-Scale Feature Collaborative Perception

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper presents a novel RB-SwinT architecture and reports a high classification accuracy of 96.20% on a self-constructed dataset. However, the study would be significantly strengthened by validating the proposed method on a publicly available benchmark dataset or comparing it with more recent state-of-the-art methods in industrial signal processing. Additionally, the lack of open-sourced code or detailed hyperparameter settings may hinder reproducibility. It is recommended that the authors provide more implementation details or consider releasing the code to facilitate future research.

1. Since industrial ball mills primarily operate under wet grinding conditions, while the ball mill used in this study is for dry grinding, is the proposed method applicable to wet grinding conditions?
2. Why was the acceleration sensor (DH131) installed at the ball bearing location in the experimental setup? Would the proposed method remain effective if the acceleration sensor were installed on the cylinder surface or at both ends?
3. The description of the experimental setup lacks sufficient detail and should be enhanced.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript addresses a ball mill load classification method based on a multi-scale feature collaborative perception approach. The topic is highly relevant from an industrial perspective. The proposed methodology appears promising; however, the paper in its current form suffers from several technical, structural, and presentation-related shortcomings. These issues should be thoroughly addressed before the manuscript can be considered for publication. The main concerns are as follows:

1. The literature review should form an integral part of the Introduction to provide a coherent background and motivation for the study. The authors are advised to merge the sections 1 (Introduction) and 2 (Related Work) into a single, well-structured Introduction, ensuring a logical flow from general context to specific research gaps.
2. The purpose of the paper is not explicitly formulated. The Introduction should conclude with a concise problem statement and clearly defined research objectives. Additionally, the authors should highlight the novel contributions of the proposed method relative to existing approaches, outlining what distinguishes this work and why it is needed.
3. The manuscript lacks a dedicated Discussion of Results section. While results are presented, they are not sufficiently analyzed or contextualized. A separate Discussion section is essential to interpret the findings, compare them with prior studies, evaluate the strengths and limitations of the proposed approach, and provide insights into practical implications for industry. Without such a section, the scientific value and applicability of the results remain unclear.
4. Informal expressions such as “by us” should be avoided in scholarly writing. Scientific manuscripts require an objective and formal tone. A more appropriate phrasing would be, for instance, “This paper presents…” or “The proposed method…”. The authors are encouraged to revise the manuscript for academic style and tone.
5. The notation for units is inconsistently formatted. For example, “9mm” should be written with a space separating the number from the unit (9 mm). The authors should review the entire manuscript to ensure compliance with SI unit formatting standards.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This study discusses the condition monitoring of heavy rotating equipment, particularly ball mills. The goal is to design predictive maintenance measures to enhance production safety and energy efficiency by accurately monitoring the operating state of ball mills.

The Introduction presents the current state of the art in monitoring and signal processing methods.

The research questions focus on enhancing the diagnostic accuracy of ball mill load conditions and investigating innovative methods for feature extraction from vibration signals.

The authors propose a monitoring method, RB-SwinT, integrating data from time-frequency maps of vibration signals.

The methodology used to process the input data is detailed in the paper—the data acquisition methodology as well as the measurement chain.

The authors indicate and exemplify the signal processing method, and some results are presented.

However, the main goal of the paper is to identify the state of the equipment, and there is no discussion related to this issue. The results presented in the paper are not associated with any state of the machine, particularly when the machine is in a critical state.

Therefore, the authors are asked to discuss the results in relation to the functioning state and determine when maintenance is required.

This way, the benefits and shortcomings of the method can be evaluated.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

All my comments have been taken into account. Thank you.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper can be accepted for publication.