Integrating Process Mining with Discrete-Event Simulation for Dynamic Productivity Estimation in Heavy Civil Construction Operations

Round 1

Reviewer 1 Report

The paper proposes the interesting DES method with Bayesian updating technique to estimate the construction operation times. The updated simulation prediction shows the increase in cycle times compared to the initial static prediction. The paper is promising but requires significant improvements to improve the readability, including the following comments:

In the Introduction, the problem should be presented and the background of the problem should be outlined.  Meanwhile, we read about the problem only in lines 337 to 342.

Literature review: Section 2.1 and 2.2 should be combined into subsections: 2.1 Process mining and simulation modeling in Architecture, 2.2 Process mining and simulation modeling in Engineering, 2.3 Process mining and simulation modeling in Construction

After literature review, the conclusions including pross and cons of existing models, methods should be given, rather than giving the conclusions in the Introduction, including also Fig.1 and 2.

Methogology:

Please present the data processing process to eliminate noises and inconsistencies.

The methodology presented is very limited, instead the authors refer to Rashid and Luis, 2019 for a deep learning technique, long short-term memory based recurrent neural network to identify activities of the equipment. Process discovery algorithms are used to discover the process from event logs which is presented in Rozinat et all. 2009. This makes the presented methodology difficult to understand.

Fig. 7 presents the number of times each of these activities occurred during the data collection process. Should hauling and returing also be depicted?

Please explain how α-algorithm is used to generate the Petri net. A numerical example for achieving the graph demonstrating inter-class dependencies between the event classes is strongly needed. How were the values of the dependency graph achieved?

The exemplary relationship between activities need to be explained in Fig. 12. What the difference between process (square), connector (circle) and card in the block diagram? What are numbers, values?....

The strong value of the paper is data driven example, and the interesting conclusions for the stockholders.

Extensive editing of English language and style required

Author Response

Dear Reviewer,

Please find our responses to your comments in the attached pdf document.

Thank you,
Joseph Louis

Author Response File: Author Response.pdf

Reviewer 2 Report

Integrating process mining and simulation for performance analysis is an interesting and up-to-date research topic. The research was applied to the real-world example which is also a strong advantage of the paper.

In line 286, ? = ?(?) is the universe of event logs (what is ??) while in line 316 L is  is the likelihood of observing the experimental outcome. I suggest not using the same notation for two different meanings. 

Although the discovered process in Fig. 9 is quite simple, the alpha algorithm is not the proper choice for process discovery. The alpha algorithm is a toy example algorithm which has so many disadvantages, even in the simple cases such as short-loops, self-loops, omitting some events, and it does not take into account frequencies, so it just should not be used for real-life purposes. If you used ProM for mining, at least use any of the extensions of the Alpha algorithm (Alpha++/Alpha#), but it will be better to stay with just Heuristic Miner for Petri Net discovery or try some other state of the art miners such as ILP miner or Inductive Miner.

In the references, only 5 of them are from the last five years. As the process mining topic is quite popular and up-to-date, the authors could extend their bibliography with more recent positions. Especially, there have been many papers concerning forecasting future performance of the processes (maybe not in heavy civil construction operations, but in other fields, and they should be elaborated in the paper), e.g.
- Teinemaa, I., Dumas, M., Rosa, M. L., & Maggi, F. M. (2019). Outcome-oriented predictive process monitoring: Review and benchmark. ACM Transactions on Knowledge Discovery from Data (TKDD), 13(2), 1-57.
as well as papers concerning process mining in underground mining:
- Brzychczy, E. (2019). An overview of data mining and process mining applications in underground mining. Inżynieria Mineralna, 21.

Some minor issues:
l. 18: "This paper proposes a framework that integrated performance forecasting with process mining..." => "This paper proposes a framework that integrates performance forecasting with process mining..."
l. 20: "further transformation to into a DES model" => "further transformation into a DES model"
l. 109: "demonstrated that that event logs" => "demonstrated that event logs"
l. 301: "the duration of all the activities are calculated" => "the duration of all the activities is calculated"
l. 584: The references [17] and [18] are the same paper.

Author Response

Dear Reviewer,

Please see the attachment for our responses to your comments.

Thank you,

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors took into account the comments of the reviewers. I recommend the paper for publication.

Reviewer 2 Report

The authors answered my questions and made suitable improvements to the previous version of the paper. I have no further comments and I think the paper is suitable for publication in its present form.