Uncertainty Analysis of Seismic Effects on Cultural Relics in Collections: Integrating Deep Learning and Reinforcement Strategies

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents a well-structured and technically strong approach to assessing seismic damage risk for cultural relics. The integration of deep learning, graph attention mechanisms, and reinforcement learning is thoughtful and clearly advances beyond many existing methods. The methodology is generally well motivated, and the experimental results provide convincing evidence that the proposed approach can outperform prior techniques.

That said, there are several points that would benefit from clarification, particularly regarding the composition of the dataset and the interpretation of the correlation analyses in Figures 4 and 5. These issues are related to clarity and interpretability, and addressing them would be important to strengthen the paper.

1. Use of Real Data versus Replica (Simulated) Data

The manuscript emphasizes the difficulty of collecting real seismic damage data for cultural relics and motivates the use of replica relic experiments combined with data correction. However, it is currently unclear:

• How many records in the CR-SDD dataset correspond to real cultural relics?

• How many records are derived from replica relic experiments?

• How many distinct replica relics were produced, and how many experimental runs were conducted per replica?

While Section 3.1 explains the correction of simulated data using integrated learning, the paper does not clearly state the relative proportion of real versus corrected simulated data in the final dataset.

Providing a concise summary (e.g., a short table or paragraph) of the dataset composition would improve transparency. In addition, a brief justification of why the available data are sufficient—for example, in terms of attribute coverage, validation against real cases, or robustness of learned patterns—would help readers better assess the empirical grounding of the study.

2. Interpretation of Correlation Analyses (Figures 4 and 5)

Figures 4 and 5 play an important role in factor screening and ontology construction, but their interpretation is currently somewhat unclear.

(a) Definition of the Correlation Coefficients

For example, in Figure 4, variables such as cause of cultural relic damage are shown with correlation coefficients (e.g., 0.614). However, it is not explicitly stated:

• What variable these coefficients are computed against.

• Whether the dependent variable is consistently the seismic damage level across all plots.

• Whether some damage-related attributes are being correlated with damage severity itself (middle plot in fig. 4)

If the y-axis in all subplots represents seismic damage level, this should be clearly stated in the figure captions and axis labels.

(b) Self-Correlation and Nonlinear Plots

If seismic damage level is indeed the dependent variable in all subplots, then at least one plot appears to show damage plotted against damage. In such a case, one would expect a near-linear (45-degree) relationship. However, the corresponding plot does not exhibit this behavior.

This raises questions about whether:

• different damage encodings or discretizations are being used,

• the plotted variables represent different damage definitions,

• or self-correlations are unintentionally included.

Clarifying this point—or excluding self-correlations altogether—would help avoid confusion.

(c) Logical Interpretation of High Correlations

Some of the reported correlations are not immediately intuitive. For instance, in Figure 5, the name of the area (unit address) appears to have a relatively high correlation with damage, potentially higher than certain physical protection factors such as whether display cases are fixed to walls or floors.

If the “name of area” is acting as a proxy for other influential factors (e.g., seismic intensity, building age, structural type, or regional construction practices), this should be explicitly explained. Without this context, such correlations may appear arbitrary or misleading to the reader.

3. Figure Presentation and Readability

To improve clarity, I recommend the following enhancements to Figures 4 and 5:

• Explicit axis labels identifying both the attribute and the dependent variable.

• Clear statements in the captions specifying:

  • the dependent variable,

  • the correlation metric used (e.g., Pearson or Spearman),

  • and whether variables are categorical or numerical.

• Consider separating attribute–damage correlations from correlations involving damage-related variables.

 

4. Clarifying the Intended Takeaways from the Correlation Analysis

Finally, the paper would benefit from a paragraph explicitly summarizing the intended insights from Figures 4 and 5, such as:

• Which factors are considered direct physical drivers of seismic damage,

• Which factors serve as contextual or proxy variables,

• How these correlations informed the ontology layer design and subsequent graph attention weighting.

Making these takeaways explicit would help bridge the gap between statistical results and domain interpretation.

 

Author Response

Response to Reviewer 1

Dear Editors and Reviewers,

RE: [applsci-4086151]

Title: Uncertainty Analysis of Seismic Effects on Cultural Relics in Collections: Integrating Deep Learning and Reinforcement Strategies

Firstly, we would like to thank the Editors and the Reviewer 1 for the comments and suggestions. We have studied all the comments carefully and have made corresponding corrections, which we hope can meet with your requirements for your approval. We elaborated on our responses to these comments and suggestions, which have been included in our revision (Response to Reviewer 1.pdf).

We appreciate the Editors and Reviewer 1 warm work earnestly, and hope that these corrections will meet with your standards for approval. If you need any further information and clarification, please do not hesitate to contact us (xuzy@sari.ac.cn and yxf@sari.ac.cn).

With best regards,

Corresponding author: Zhengyi Xu and Xiaofei Yang

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Manuscript number: Applsci-4086151

Title: Uncertainty Analysis of Seismic Effects on Cultural Relics in Collections: Integrating Deep Learning and Reinforcement Strategies

Due to the unpredictability of seismic and the complexity of collection environments, significant uncertainty exists regarding their impact on cultural relics. Moreover, existing research on the causal analysis of seismic damage to cultural relics remains insufficient, thereby limiting advancements in risk assessment and protective measures. To address this issue, this paper proposes a seismic damage risk assessment method for cultural relics in collections, integrating deep learning and reinforcement strategies. The proposed method enhances the dataset on seismic impacts on cultural relics by developing a deep integrated learning-based data correction model. Furthermore, it incorporates a graph attention mechanism to precisely quantify the influence of various attribute factors on cultural relic damage. Additionally, by combining reinforcement learning with the Deep Deterministic Policy Gradient (DDPG) strategy, this method refines seismic risk assessments and formulates more targeted preventive protection measures for cultural relics in collections. This study evaluates the proposed method using three public datasets in comparison with the self-constructed Seismic Damage Dataset of Cultural Relics (CR-SDD). Experiments are conducted to assess and analyze the predictive performance of various models.  Experimental results demonstrate that the proposed method achieves an accuracy of 81.21% in assessing seismic damage to cultural relics in collections. This research provides a scientific foundation and practical guidance for the protection of cultural relics, offering strong support for preventive conservation efforts in seismic risk mitigation.

The article is written concisely and summarizes the results of a substantial volume of information. The statements are correctly supported with results and explanations.

I recommend the authors to complete the article with information about:

• Please consider looking also at seismic vulnerability assessment methodology that considers the cultural value, or consider the losses of architectural-artistic assets that are not structurally connected with the main building, as the damage pattern could be considered similar to the one investigated;
• Please provide a short synthesis of the state-of-the-art findings, that could be easily understand, by a table or a chart;
• Please provide a clearer high-level framework, schematic, to clarify how the raw data is collected, how the numerical data is corrected based on the real data, how the weighting is considered into the risk assessment;
• Please clarify how the corrected values are supported by physical or conservation-based justification;
• Please better justify why performance on the analysed datasets are considered valid in the context of the proposed correction method for heritage applications;
• Please discuss Figure 15 in relation to seismic vulnerability knowledge for museum objects and clarify if the reinforcement learning-based recommendations are consistent with current museum practice or standards?;
• Please highlight the novelty and opportunity of your research;
• Please state the limitations of the research work.

I recommend the paper for publication after considering the major revisions because it presents important original information.

Author Response

Response to Reviewer 2

Dear Editors and Reviewers,

RE: [applsci-4086151]

Title: Uncertainty Analysis of Seismic Effects on Cultural Relics in Collections: Integrating Deep Learning and Reinforcement Strategies

Firstly, we would like to thank the Editors and the Reviewer 2 for the comments and suggestions. We have studied all the comments carefully and have made corresponding corrections, which we hope can meet with your requirements for your approval. We elaborated on our responses to these comments and suggestions, which have been included in our revision (Response to Reviewer 1.pdf).

We appreciate the Editors and Reviewer 2 warm work earnestly, and hope that these corrections will meet with your standards for approval. If you need any further information and clarification, please do not hesitate to contact us (xuzy@sari.ac.cn and yxf@sari.ac.cn).

With best regards,

Corresponding author: Zhengyi Xu and Xiaofei Yang

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Esteemed authors,

Thank you for considering all the recommendations. I believe the revised version of the manuscript is much better than the original one. Although I appreciate all responses, there is still one point that the authors should address more carefully. The following comment should be considered in more depth, because introducing only a short paragraph cannot be considered enough. 

Comment: Please consider looking also at seismic vulnerability assessment methodology that considers the cultural value, or consider the losses of architectural-artistic assets that are not structurally connected with the main building, as the damage pattern could be considered similar to the one investigated.

To clarify this comment, the authors should also include multidisciplinary studies that are using different methodologies (even empirical) to discuss the effect of the earthquakes on architectural-artistic assets, and even the loss probability. 

Thank you

Iasmina Onescu

Author Response

Dear Editors and Reviewer,

 

RE: [applsci-4086151]

 

Title: Uncertainty Analysis of Seismic Effects on Cultural Relics in Collections: Integrating Deep Learning and Reinforcement Strategies

 

Firstly, we would like to thank the Editors and the Reviewers for the comments and suggestions. We have studied all the comments carefully and have made corresponding corrections, which we hope can meet with your requirements for your approval. We elaborated on our responses to these comments and suggestions, which have been included in our revision.

 

We appreciate the Editors and Reviewers warm work earnestly, and hope that these corrections will meet with your standards for approval. If you need any further information and clarification, please do not hesitate to contact us (xuzy@sari.ac.cn and yxf@sari.ac.cn).

 

With best regards,

 

Corresponding author: Zhengyi Xu and Xiaofei Yang

Author Response File: Author Response.pdf