Integrating Bayesian Networks and Numerical Simulation for Risk Assessment of Deep Foundation Pit Clusters
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe topic of this manuscript holds substantial theoretical and practical engineering significance, offering an in-depth examination of the relatively underexplored issue of risk assessment for foundation pit clusters. The manuscript proposes a novel risk assessment framework that demonstrates strong application potential and is likely to make a valuable contribution to the field. The paper is well structured, and the research logic is clear. To further enhance the manuscript’s clarity, rigor, and overall impact, I offer the following constructive suggestions for the authors’ consideration.
- While the risk assessment of individual excavation pits has been widely studied, research on excavation pit groups or clusters remains relatively limited. A key concern is that the manuscript's exposition of the critical risk characteristics specific to pit groups lacks clarity. From a practical engineering standpoint, the risks during the excavation of pit groups can be classified into two primary categories:
（1）Buffer Zone Method: This involves establishing a buffer zone between pits. The pits flanking the buffer zone are excavated first, followed by the central pit within the buffer zone. In this configuration, the width of the buffer zone becomes the critical control parameter and a primary source of risk.
（2）Adjacent/Sequential Excavation: This typically involves excavating a pit on one side of a partition wall. Standard practice dictates that the excavation of the adjacent pit on the other side should only commence after the completion of the first pit's substructure. However, project schedules often necessitate the concurrent excavation of the adjacent pit before the first is structurally complete (a scenario of "closely spaced" or "abutting" pits). Here, the principal risk stems from the management and control of the concurrent construction conditions.
The authors appear to model both risk categories predominantly through the single variable of "pit spacing." However, the paper fails to systematically articulate why pit spacing should be considered the crucial determinant in managing both of these distinct risk scenarios. This omission makes the rationale for focusing on this parameter seem tenuous. I strongly recommend that the authors expand this section to provide a more robust theoretical and practical justification for their choice of "pit spacing" as the key analytical factor.
- The risk judgment for excavation pit groups is an inherently complex decision-making process, given that many of the underlying mechanisms are not yet fully understood, which constitutes a primary challenge in this research area. Consequently, such assessments often rely on the input of experienced experts to guide the decision-making process. This raises a critical question regarding the risk judgment matrices presented in Tables 1-5: were they developed with or validated by expert opinion? If expert consultation was indeed part of the methodology, it is crucial that the authors explicitly state this in the text, as this information is fundamental to establishing the scientific validity and numerical rigor of the values within these matrices.
- Regarding Section 3.5, the "standard model" of the excavation pit group is not clearly defined. The manuscript lacks essential information, such as the plan view of the pit group, the cross-section of the support system, and the basic construction conditions. Furthermore, Figures 7 and 8 are presented without sufficient explanation, making it unclear whether they depict a plan view or a profile view of the support structure. It is recommended that the authors supplement the text with these necessary descriptions to improve the clarity and reproducibility of their work.
- Regarding the legend in Figure 9, there is an inconsistency in the terminology used. In the section titled “Evaluation of Support Failure in Deep Foundation Pit Clusters (DFPC),” the text uses “type A” and “type B” to designate the two deep foundation pit layouts. However, the legend in Figure 9 uses the terms “In line” and “Parallel.” To maintain consistency throughout the manuscript, it is recommended that the authors either use the more descriptive terms “in-line layout” and “parallel layout” directly in the main text, or explicitly define what “type A” and “type B” correspond to upon their first mention. This will ensure clarity for the reader.
- Line 153: In Equation (5), it is recommended to move the parentheses from just the numerator to enclose the entire fraction. This would be more notationally conventional.
- It is suggested that the authors emphasize the potential comprehensiveness of their research method in predicting various failure modes. The current validation primarily focuses on two risk factors: support failure (C2-8) and pit spacing (C7). However, a complete risk system for a Deep Foundation Pit Cluster (DFPC) also involves other critical failure modes. It would be beneficial to briefly address this in the conclusion. For instance, the authors could state that while the present study provides an in-depth validation centered on support failure, the proposed “BN-AHP” framework is highly extensible. This design would allow for the future integration of quantitative analysis results from other key failure modes, such as hydraulic instability, into the Bayesian Network, thereby enabling a more holistic and comprehensive risk assessment for the entire DFPC system.
- In the model validation section, it is noted that the simulation result for the positive bending moment shows a 24.89% difference compared to the model by Zheng et al. Although the paper concludes that this discrepancy is within an acceptable range, from a technical standpoint, the argument could be strengthened by briefly speculating on the possible reasons for this relatively large difference. For instance, such a discrepancy could stem from variations in the parameter settings at the soil-structure interface, differences in the refinement of the mesh division, or subtle distinctions in the constitutive model parameters used in the respective models. Including such a discussion would further enhance the persuasiveness and technical rigor of the model validation section.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article addresses an interesting and current topic. Despite being extremely specific, the subject is important for current constructions. This study proposes a framework that integrates expert knowledge with data-driven structural learning. This framework systematically qualifies risk interactions among nearby pits via fault tree analysis. The framework also optimizes BN topology via structure learning, and the AHP is used to determine prior probabilities within the BN, combining systematic prioritization with probabilistic reasoning.
However, it is important to give the article a greater scientific structure, as the intention is to publish in a high-impact international journal. To this end, the title could be clearer and more objective. Furthermore, the review is geographically limited (basically, only Asian authors) and numerically limited. It is suggested to increase the scope of the research and include a more diverse range of authors.
Furthermore, the figures are unclear, for example, Figure 1. Much of the information is difficult to read and the analysis is simplified. The same goes for the tables, especially Tables 1, 2, 3, and 5. There's too much information for one table, with very little analysis to follow.
After all this, there's a case study that feels a bit disjointed, despite being the most interesting part of the article.
And the conclusions, due to the scientific confusion of the entire article, are completely abstract and inconclusive.
As stated, the article is interesting and deserves to be published, but it needs a complete overhaul.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors- In Table 1, why were 17 factors considered for fundamental risk? Isn't it better to have less factors?
- Why were PhD students selected as a specialists? Isn't it better to select skilled engineer instead of the students?
- In AHP, weight factors are significantly important. It is suggested to indicate method of obtaining the factors using a diagram or figure.
- Numerical model indicated in Figure 4 is not enough clarified. That is, which method was used to simulate DFP (FEM or CDM), the boundary conditions, method of analysis and loads are not explained.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper “Hybrid Bayesian Network-based Risk Assessment Framework for Deep Foundation Pit Clusters” reports a research work about the evaluation of the possible risks which lead to the collapse of DFPC by means a new approach based on Bayesian Network (BN). In general, the topic of the manuscript is interesting for both scientific community and designers also considering that is not completely covered by current literature. Furthermore, the proposed approach and the obtained results are discussed and commented in the text. However, some aspects of the paper must be improved before to consider the manuscript for publication in Buildings.
- Section 3.1: the presented numerical model must be described in detail: What boundary conditions were applied? Was a construction stage analysis performed to correctly assess the displacements and the soil stresses?
- Is the model size large enough to eliminate lateral boundary effects? Generally, three times the size of the structure is used per side (see M. Zucca, A. Franchi, P. Crespi, N. Longrarini, P. Ronca. “The new foundation system for the transept reconstruction of the basilica di collemaggio”. Proceedings of the International Masonry Society Conferences, 10th International Masonry Conference, IMC 2018, 9-11 July 2018, Milan, Italy).
- What are the parameters used in the HSS model definition?
- Table 8: What do the authors mean by positive or negative bending moment?
- In the numerical model, the support structures are implemented using beam or truss elements?
- Figure 8 is unclear.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsI believe the article is interesting and deserves to be published, as the results and achievements may be useful to other researchers. The authors did a very good job of reviewing it, trying to incorporate everything possible to improve the article. Therefore, I consider it to have met the requirements for publication.
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper can be considered for publication in present form
