A Framework for Evaluating the Load-Carrying Capacity of Bridges without Design Document Using an AI Technique
Round 1
Reviewer 1 Report
Paper: A Framework for Evaluating the Load-carrying Capacity of 2 Bridges without Design Document Using an AI Technique
The manuscript is devoted to a topic that deserves attention due to the aging infrastructure stock in many developed countries …..
The paper is generally well organized and written, even though it needs some improvement as follows:
· In the introduction, the authors do not mention a recently developed framework related to Italian bridges. It considers four risk types conveyed in the Class of Attention of the bridge, which is a qualitative risk evaluation and is the preamble to accurate evaluations based on safety checks and detailed models. I think it is one of the most advanced frameworks for bridge management being also very recent (2020). To this end see and cite the following paper
. https://doi.org/10.3390/infrastructures6080111
· Yet in the introduction, when describing Ai techniques, it is important to underline that such techniques can be also useful for Structural Health Monitoring. To this end see and cite the paper
https://doi.org/10.1007/s42452-019-0808-6
· The main problem of the developed procedure seems to be that it is independent on the bridge structural type: RC, prestressed reinforced concrete, steel, and steel-reinforced concrete are on the same level. It is unlikely that such a general procedure can be effective with any bridge structural type.
· Section 3 starts with a figure. It is better to move the figure on.
· The sentence in lines 151-152 is very strange. The authors want to use deflection as an indicator of the actual bridge capacity. It seems a wrong approach since an increase in deflection can be representative of serviceability problems not necessarily linked to load-bearing capacity.
· Expression 3 does not result in a displacement since units give 1/L and not L for a proper displacement. what is M: a moment or a force?
· Figure 4g: what is the unit of the horizontal axis? Put the units also in the other figures
· Line 166: when talking about the collected design documents, it should be stated the bridge typologies, and the number of bridges
· Table 1: it is strange having min data larger than the maximum and avg
· Section 5.4: it seems that the ANN method tends to overestimate the target strength
· In this reviewer’s opinion a major revision is necessary mainly relating to a better characterization by structural type as mentioned above. Neglecting this aspect, it results in a blind application of AI forgetting engineering aspects that are very important in the safety evaluations of infrastructures.
· Conclusions should be modified to make the reader aware that such a procedure can be used for large-scale evaluations of an entire bridge stock. This latter, in order to prioritize more refined safety assessment activities that should be carried out through the usual in-situ testing and finite element modelling necessary for careful and precise evaluations at the single bridge level, which cannot be replaced by AI procedures
Author Response
Manuscript number: applsci-2173082
Paper title: A Framework for Evaluating the Load-carrying Capacity of Bridges without Design Document Using an AI Technique
The authors thank the reviewers for their careful reading of the manuscript and constructive remarks. The authors have taken the comments on board to improve and clarify the manuscript. Please find the below-detailed responses to all comments (reviewers’ comments in black, our replies in blue). Furthermore, the manuscript has been proofread by a native English speaker.
Sincerely,
Jin-Kook Kim, Ph.D, Professor
Seoul National University of Science and Technology
Author Response File: 
Author Response.docx
Reviewer 2 Report
The paper proposes a framework for estimating the load-carrying capacity of old Korean bridges in absence of design documents, based on the use of the artificial intelligence. In detail, with reference to simply supported decks, relationships between externally measurable geometric characteristics and design strength are established starting from 124 design documents, comparing the performance of different regression algorithms; then, an ANN is determined. The actual condition of the bridge, subjected to ageing effects, is taken into account introducing a bridge condition rating system, based on the prediction of a response ratio, defined as the ratio between the calculated deflection ant the measured deflection response. In order to predict the above parameter, 82 safety diagnosis reports for multiple bridges were collected. Again, relationships between response ratios and the bridge parameters (such as the bridge dimensions, number of lanes, service period,…).
The paper is not badly written, although the use of the English language can be improved. Also, the paper is sufficiently well organised and easy to understand. The research topic is of interest and in line with the journal aims and scope. However, the following main issues must be addressed by the authors before the paper could be accepted for publication:
- It is stated in the paper that the evaluation of load-carrying capacity is entirely de pendent on the engineers' judgment and research through on-site load tests. This is not fully correct since the conventional approach for the safety assessment of existing bridges is based on performing structural analysis and safety verifications starting from the material properties obtained from experimental tests. The use of on-site load tests is possible but usually unconventional in western Europe. This aspect must be well defined and contextualised.
- The relationship between the geometric properties of a bridge and its load carrying capacity is not so direct, especially for rc bridges, since strength and stiffness may be uncorrelated (e.g. stiffness is closely dependent on the geometry while strength is closely dependent on the amount of reinforcement). Relationships may be clearer for steel or steel-concrete composite bridges, and from a general point of view, different correlations may be found depending on the bridge typology (not only the static scheme). The Authors are invited to consider this issue and to suitably address it in their paper.
- According to the previous comment, it interesting to have a better idea of the dispersions of the predicted data with respect to the benchmarks.
- A table, or a flowchart, summarizing the input data for the ANN is kindly appreciated.
- How is the sample divided to train and test the ANN?
Minor and editorial issues:
- The sentence in the abstract “Here, no field tests were conducted” is a bit abrupt.
- Page 1, line 35: segregation usually identify the separation of cement paste and aggregates of concrete from each other during handling and placement. In the section, it seems that it’s an issue due to ageing.
- Page 9, line 247: the reference to equation 4 is not correct.
- Please define the ordinate of graphs in Figure 8;
- Page 4, line 129: please, revise the pagination.
Author Response
Manuscript number: applsci-2173082
Paper title: A Framework for Evaluating the Load-carrying Capacity of Bridges without Design Document Using an AI Technique
The authors thank the reviewers for their careful reading of the manuscript and constructive remarks. The authors have taken the comments on board to improve and clarify the manuscript. Please find the below-detailed responses to all comments (reviewers’ comments in black, our replies in blue). Furthermore, the manuscript has been proofread by a native English speaker.
Sincerely,
Jin-Kook Kim, Ph.D, Professor
Seoul National University of Science and Technology
Author Response File: Author Response.docx
Reviewer 3 Report
Please see attached
Comments for author File: Comments.docx
Author Response
Manuscript number: applsci-2173082
Paper title: A Framework for Evaluating the Load-carrying Capacity of Bridges without Design Document Using an AI Technique
The authors thank the reviewers for their careful reading of the manuscript and constructive remarks. The authors have taken the comments on board to improve and clarify the manuscript. Please find the below-detailed responses to all comments (reviewers’ comments in black, our replies in blue). Furthermore, the manuscript has been proofread by a native English speaker.
Sincerely,
Jin-Kook Kim, Ph.D, Professor
Seoul National University of Science and Technology
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Since all my concerns were addressed, the paper can be published.
