Acoustic-Based Machine Condition Monitoring—Methods and Challenges
Round 1
Reviewer 1 Report
This review on machine acoustic diagnostics is interesting and may be useful to some researchers and engineers working in the field.
The paper is well structured, but it lacks some considerations about appropriate microphones, including type, characteristics, arrangements, and recommended data acquisition rates.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
An interesting article presenting the challenges of modern extensive diagnostic systems
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The methods and challenges of acoustic signal-based mechanical condition monitoring are described in this paper. Firstly, anomalous acoustic signals can be extracted by three different machine-learning neural networks, namely autoencoder-based anomaly detection, Gaussian mixture model-based anomaly detection, and Outlier Exposure-based anomaly detection. Secondly, feature-based machine learning methods and acoustic image-based deep learning methods were introduced as classification methods for abnormal acoustic signals. Thirdly, several publicly available databases that can be used for acoustic signal mechanical condition detection are presented. In addition, the problem of denoising in acoustic signals, domain shift, and domain generalization in machine learning. As a review of the existing technology review paper, this paper is not comprehensive enough and lack of deep analysis of the existing literature. The article's primary contribution is unclear, and the invention does not satisfy the journal's requirements. In addition, there are some problems with the content and format.
In the reviewers' opinion, the manuscript should be rejected for publication. The specific comments are as follows:
1. This paper only briefly describes the theoretical part of the abnormal acoustic signal extraction and feature classification methods and lacks a description of the specific application of the algorithms.
2. The description of the specific process of the abnormal acoustic signal extraction methods needs to be clarified, and the essence of the methods cannot be introduced through the description in the form of a table.
3. Anomalous acoustic signal extraction can be considered from the time, frequency, and time-frequency domains. Still, in this paper, the features of different fields are described in the same table, which is confusing.
4. The abnormal acoustic signal classification method is only a brief introduction to the algorithms that can be used. The specific algorithms for different mechanical devices should be described in detail.
5. The introduction of datasets that can be used for mechanical condition monitoring only briefly describes the types of faults that are of interest to the public dataset and how the signals are acquired, lacking a description of the practical applications of the datasets.
6. This paper describes the challenges of acoustic signal-based mechanical detection. Still, it needs to explicitly describe the research work to address the existing problem. In addition, it needs to propose a development plan to address the current issues.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
As a review paper for acoustical-based machinery condition monitoring, the authors did not mention the acoustic imaging -based condition monitoring and fault diagnosis methods with microphone array measurement. These works which are highly related to the topic of this paper should also be cited and reviewed.
By the way, I still think this paper did not make much contribution to this filed. As a review paper, it should provide the future research trend and give some inspiration for the scholors in this field.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
