Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. The validation R² of 0.99 is unrealistically high for rock mass prediction and indicates likely overfitting. The authors must provide learning curves and test the model on independent tunnel projects with different geological conditions. Regularization techniques and their effects should be explicitly reported.

2. sing only an 80/20 split for 512 samples is insufficient. Implement k-fold cross-validation (k=5 or 10) and report mean and standard deviation of all metrics. Include confidence intervals for all reported performance measures.

3. "Empirical methods" for hyperparameter selection is inadequate. Document the search space and use grid search or Bayesian optimization. Explain why ResNet50 was chosen over other depths through ablation studies.

4. Model descriptions, training procedures, and metric definitions are repeated multiple times. Consolidate methodology into one section, merge similar tables (5-6 and 8-9), and define R² and RMSE only once. Section 3 should compare models while Section 4 focuses on temporal analysis.

Comments on the Quality of English Language

The manuscript contains grammatical errors and inconsistent article usage throughout. Verb tenses shift between past and present within sections. Many sentences are too long and should be divided for clarity. Technical terms like "back analysis" need consistent hyphenation. The abstract contains run-on sentences that obscure key findings. Passive voice is overused where active voice would be clearer. Prepositions are occasionally incorrect, particularly "in" versus "on" for data inputs.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper “Deep-Learning Parameter Identification for Rock Masses from Excavation-Induced Tunnel Deformations” reports a research work about the evaluation of rock mass parameters useful to correct design underground structures by means of the application five machine learning algorithms. In general, the topic of the manuscript is interesting although there are already several works in the literature regarding the evaluation of rock mass parameters. Furthermore, the paper appears well-organized in its different sections. However, some aspects of the manuscript must be improved before to consider the paper for publication in Applied Sciences.

- The original aspects of the research work must be highlighted in detail at the end of the introduction and in the conclusions.

- Section 2.3: Why was the Mohr-Coulomb failure criterion used and not the Hoek-Brown one? The latter seems more appropriate in the presence of rock masses.

- Section 2.3 (line 191): How was the factor of 1.1 due to the presence of bolt anchors evaluated?

- Table 3 is not readable.

- The Tables references and the numbering are not correct.

- lines 34-36 consider as reported in 10.1016/j.engstruct.2020.110497

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The review file is included in the attached document.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have made substantial improvements to the manuscript in response to reviewer's comments. The additions regarding overfitting mitigation, cross-validation, and methodological clarity significantly strengthen the work.

Reviewer 2 Report

Comments and Suggestions for Authors

The paper can be considered for publication in present form