Next Article in Journal
Development of Digital Twin for DC-DC Converters Under Varying Parameter Conditions
Previous Article in Journal
An Enhanced Fractal Image Compression Algorithm Based on Adaptive Non-Uniform Rectangular Partition
 
 
Article
Peer-Review Record

Broken Wire Detection Based on TDFWNet and Its Application in the FAST Project

Electronics 2025, 14(13), 2544; https://doi.org/10.3390/electronics14132544
by Wanxu Zhu 1,2,*, Zixu Zhong 1,2, Sha Cheng 1,2, Qingwei Li 3,*, Rui Yao 3 and Hui Li 3
Reviewer 1:
Reviewer 2:
Reviewer 3:
Electronics 2025, 14(13), 2544; https://doi.org/10.3390/electronics14132544
Submission received: 27 May 2025 / Revised: 21 June 2025 / Accepted: 22 June 2025 / Published: 24 June 2025
(This article belongs to the Section Computer Science & Engineering)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript presents an innovative approach to broken wire detection using a Time-Domain Feature Weighted Network (TDFWNet) tailored for the specific challenges faced in the Five-hundred-meter Aperture Spherical radio Telescope (FAST) project. The integration of time-domain feature analysis with Convolutional Neural Networks (CNNs) represents a significant advancement in the field of wire breakage detection, particularly in complex engineering environments where interference signals are prevalent. The authors have demonstrated through experiments and data augmentation techniques that TDFWNet outperforms traditional CNN models in terms of precision, recall, F1 score, and accuracy, which is a commendable achievement.

 

Furthermore, the application of TDFWNet in practical engineering scenarios, specifically in the bending fatigue tests of FAST drive cables, highlights its robustness and practicality. The consistency between the detected suspected wire-breakage signals and the results of post-fatigue test disassembly inspections underscores the effectiveness of the proposed method.

 

  1. While the manuscript briefly mentions existing methods for broken wire detection, a more comprehensive review of recent advancements in the field, especially those involving deep learning and advanced signal processing techniques, would strengthen the paper's foundation. Discussing how the proposed TDFWNet compares with or builds upon these recent works would provide readers with a clearer context of the research's novelty and significance.
  2. The section on data augmentation using FCDTW could benefit from more detailed explanations and visualizations. Specifically, the process of generating augmented data, including the selection of interpolation coefficients and the criteria for ensuring the validity of augmented samples, should be clearly described. Additionally, providing examples or visualizations of the augmented data compared to the original data would enhance the transparency and reproducibility of the research.
  3. The current study relies on a limited number of original FAST drive cable breakage signal samples. To strengthen the generalizability and robustness of the TDFWNet model, it is recommended to collect more wire breakage signals from different scenarios and operating conditions. This would allow for a more comprehensive evaluation of the model's performance under various conditions and potentially lead to further improvements in its accuracy and reliability.
  4. Although the authors have acknowledged some limitations of the study, such as the limited sample size, a more detailed discussion of the potential limitations and challenges associated with the proposed method would be beneficial. This could include the impact of environmental noise, the need for real-time processing capabilities in practical applications, and the scalability of the method to larger-scale systems. Furthermore, outlining specific directions for future research, such as the development of multi-dimensional interference suppression algorithms or the integration of other advanced sensing technologies, would provide readers with a clear roadmap for advancing the field.

In conclusion, the manuscript presents a promising approach to broken wire detection with significant potential for practical applications. With the suggested revisions, the paper could be further strengthened and made more impactful within the scientific community.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The research addresses a challenging and crucial issue: the detection of wire breakage in the traction cables of the FAST feed cabin. The health of these steel wires significantly influences the mechanical properties of the entire traction cable system, making rapid and effective detection essential for operational safety and performance evaluation.

The main contribution, the TDFWNet, is an innovative hybrid deep learning architecture. It creatively integrates the feature-learning capabilities of Convolutional Neural Networks (CNNs) with explicit time-domain feature analysis. By using feature probabilities derived from key temporal parameters (waveform factor, pulse factor, and kurtosis) as weighted inputs, TDFWNet significantly enhances the sensitivity and recognition accuracy of wire breakage signals and improves its ability to distinguish interference signals.

This approach goes beyond traditional CNNs that rely solely on implicit feature extraction. It significantly outperforms traditional CNN models across all key evaluation metrics: Precision – a 1.5% improvement; Recall – a 2.0% improvement; F1 Score – a 1.8% improvement; Accuracy – a substantial 16.6% improvement.

Although effectively mitigated by FCDTW, the study acknowledges that the limited number of original FAST traction cable breakage signal samples (23 sets) primarily restricts its application to the specific scenario of monitoring bending fatigue. A broader range of original samples from different breakage mechanisms or environmental conditions could further enhance the model’s generalisation capabilities.

The future work section correctly identifies the need to analyse monitoring data from the actual operational environment of FAST and to explore external interference factors. The current validation is strong for controlled bending fatigue tests but lacks explicit discussion on how TDFWNet would perform under various real-world environmental interferences that may present different noise profiles than those encountered during bending fatigue.

The article justifies the chosen weights for waveform factor, kurtosis, and pulse factor (0.6, 0.3, 0.1, respectively) based on time-domain feature analysis. While reasonable, it would be beneficial to discuss whether these weights are fixed across all cable types/conditions or whether a more adaptive mechanism could be explored in future work to optimise performance for potentially varying signal characteristics.

In the model architecture section, the calculation of feature probabilities (P_feature) refers to the "continuous data segment where the probability of wire breakage exceeds 0.5." A clearer explanation of how this "feature segment" is initially identified or windowed from the raw input signal before these probabilities are calculated would improve understanding. Is it a sliding window approach, or does the CNN output initially define these segments?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

The submitted manuscript demonstrates a high-quality implementation of a real-time fault diagnosis system for industrial control systems. The use of LSTM-based deep learning and multi-source data fusion is sound and state-of-the-art. The system architecture is clearly presented and its performance is evaluated using relevant metrics.

At the same time, I would like to suggest a more precise match between the title and the actual content of the study. Although the manuscript presents fault detection of broken wires and validates the results through disassembly and inspection, the term “Broken Wire Detection” in the title may suggest a broader or more general diagnostic framework than is currently detailed in the text. Consider revising the title to more clearly reflect the specific scope of your approach - for example: “Real-Time Fault Diagnosis for Cable Fatigue Based on Multi-Source Data Fusion and LSTM-based Deep Learning.” Additionally, it would be useful to include a brief discussion of the limitations of the proposed system, such as sensitivity to data loss, scalability in more complex topologies, or dependence on signal quality. This would increase the transparency and practical value of your contribution.

Best regards,

Comments on the Quality of English Language

Some expressions and sentence structures could be improved. The manuscript would benefit from minor language polishing by a native English speaker.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Back to TopTop