Accuracy of Mathematical Models and Process Simulators for Predicting the Performance of Gas-Separation Membranes

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you, Yousef, for your exciting work. My apology for the delayed response. The work is conducted in a good manner, but I do have several questions, and they are important to be addressed before acceptance. The manuscript should be accepted after. 

My comments/inputs are:

1. The method states that there is no accumulation across the membrane and no back mixing. This is not accurate in a spiral wound module. Back mixing is not present if a sweep gas is used to avoid no accumulation/concentration polarization, but these assumptions contradict each other. And in reality, sweep gas is used. Can this be incorporated into the calculation?

2. The stage cut, referred to here as permeate cut (the term needs to be standardized to the common knowledge) - anyway, the data reported from MTR states that the stage cut is 0.2, and why does the author consider stage cut 0.5 for the simulation? This difference can cause a lot of issues in designing a membrane process. 

3. Line 167-179: the 17% error is big. I understand that the author justified that it is acceptable - but a 17% error makes a big difference when designing a membrane system based on these simulation tools, e.g., the simulation determines 4 modules are needed to produce a good conditioning process, but in reality, it could be 5. That's a big impact on a process design. Can this be further improved? 

4. Line 180-188. I do not agree with the argument that the high error for higher chain HC is because of their low concentration in feed. This might be true for this exact case, but has the author considered the molecule competitive sorption effects, concentration polarization, other solubility effects, etc? This is a huge issue when dealing with different feed compositions consisting of higher concentrations of those HC. Many cases of fuel conditioning require a higher degree of conditioning, e.g., 93% to 97%; how can these issues be addressed? 

5. Line 189-192: Taking on GC inaccuracy based entirely on assumption is not needed. Based on this logic, if the GC is correctly calibrated, etc., how would the author address the data/simulation discrepancies? 

6. Line 207-217. The argument that taking only methane into account and then using simulation tools would perform better is not a great argument. What is the use of these tools if not addressing the other components in the fuel gas when they are the main components to be removed in a fuel gas conditioning process? A different argument and supporting data/information need to be used if the issues of inaccuracy/discrepancies (stated above) cannot be addressed. 

 

Thank you, and I look forward to having more information. Again, this is a great work. 

Comments on the Quality of English Language

English language is clear and precise to deliver the message. 

Author Response

First, I want to thanks the reviewer for his valuable comments.

Reviewer #1

The study developed a membrane module in the COCO using Scilab, enabling gas separation simulations. I have provided comments below for the authors to consider:
1. The method states that there is no accumulation across the membrane and no back mixing. This is not accurate in a spiral wound module. Back mixing is not present if a sweep gas is used to avoid no accumulation/concentration polarization, but these assumptions contradict each other. And in reality, sweep gas is used. Can this be incorporated into the calculation?

• Yes, there is a possibility of back mixing and concentration polarization but the cross-flow design has been used to minimize the effect. The continuous flow in one direction with high pressure (feed to retentate) may reduce the phenomena. Another possible phenomenon is the plasticization in which large hydrocarbon molecules can cause surface swelling.
• The following was updated:
  • Abstract (Lines 21-23).
  • Results and Discussion were updated accordingly (Lines 193-199)
  • Conclusion (Lines 235-236).

2. The stage cut, referred to here as permeate cut (the term needs to be standardized to the common knowledge) - anyway, the data reported from MTR states that the stage cut is 0.2, and why does the author consider stage cut 0.5 for the simulation? This difference can cause a lot of issues in designing a membrane process.

• The permeate cut in this study is defined as the recovery (amounts of gas A in the permeate compared to amounts of gas A in the feed). The stage cut you are referring to is the total stage cut, not for each compound. To solve the membrane system, stage cut of each component needs to be assumed. Yes it can be calculated for each compound using MTR data but the objective is to test the accuracy of simulators to determine the stage cut of compounds. If MTR stage cuts were used, the system will be solved instantly. In reality, we want to use process simulators to estimate the permeate composition.

3. Line 167-179: the 17% error is big. I understand that the author justified that it is acceptable - but a 17% error makes a big difference when designing a membrane system based on these simulation tools, e.g., the simulation determines 4 modules are needed to produce a good conditioning process, but in reality, it could be 5. That's a big impact on a process design. Can this be further improved?.

• Yes, I agree 17% is high and an error of 12% is approved by most refineries. The error could be related to plasticization and concentration polarization as the calculated methane composition is higher than the actual value. More industrial data is needed to confirm the accuracy of the simulators but only MTR provided the full data of feed/permeate/retentate composition and flow rate.

4. Line 180-188. I do not agree with the argument that the high error for higher chain HC is because of their low concentration in feed. This might be true for this exact case, but has the author considered the molecule competitive sorption effects, concentration polarization, other solubility effects, etc? This is a huge issue when dealing with different feed compositions consisting of higher concentrations of those HC. Many cases of fuel conditioning require a higher degree of conditioning, e.g., 93% to 97%; how can these issues be addressed.

• Yes, the high error could be related to concentration polarization and other effects. I believe for a higher degree of conditioning, the error will decrease because the simulators were good at estimating methane concentration.
• The statement indicating that the high error is related to the low composition has been removed.
• Abstract/Conclusion/Results and Discussion were updated accordingly (From Q1).

5. Line 189-192: Taking on GC inaccuracy based entirely on assumption is not needed. Based on this logic, if the GC is correctly calibrated, etc., how would the author address the data/simulation discrepancies.

• The section stating that the GC could contribute in error was removed.

6. . Line 207-217. The argument that taking only methane into account and then using simulation tools would perform better is not a great argument. What is the use of these tools if not addressing the other components in the fuel gas when they are the main components to be removed in a fuel gas conditioning process? A different argument and supporting data/information need to be used if the issues of inaccuracy/discrepancies (stated above) cannot be addressed.

• The section discussing the use of methane only has been removed.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The paper by Alqaheem Y. compared the accuracy of two process simulators for predicting the performance of gas-separation membranes. Process simulators were based on the commercial Uni-SIM® and the freeware COCO software. Models implemented within these software packages were validated using experimental data.

Both process simulators use the same physical parameters and models, with the difference that some equations in the COCO had to be specified manually. Thus, the convergence of the solution method used by different software packages for the same problem was actually compared. Considering that both solution methods included in software packages are applicable to solve this problem, it is not surprising that the results of simulations agree well with each other.

Nevertheless, they equally poorly (albeit at an acceptable level) reproduce experimental data. Low accuracy is not a problem of the solver, but an imperfection of the chosen physical model.

Given the above, the question arises: what exactly did the author want to show or compare?

In essence, the gas-separation membrane model implemented in a free software package was validated with a model based on commercial software. Both models have been proven to be equally applicable for predicting the membrane performance. However, the accuracy of these models cannot be linked to the software, but is primarily determined by the convergence of the solvers and the applicability of the underlying physical model of the process.

Considering that the presented results do not correspond to the title and the stated goals, I believe that the author should seriously revise the article.

Author Response

Reviewer #2

The paper by Alqaheem Y. compared the accuracy of two process simulators for predicting the performance of gas-separation membranes. Process simulators were based on the commercial Uni-SIM® and the freeware COCO software. Models implemented within these software packages were validated using experimental data.

Both process simulators use the same physical parameters and models, with the difference that some equations in the COCO had to be specified manually. Thus, the convergence of the solution method used by different software packages for the same problem was actually compared. Considering that both solution methods included in software packages are applicable to solve this problem, it is not surprising that the results of simulations agree well with each other.

Nevertheless, they equally poorly (albeit at an acceptable level) reproduce experimental data. Low accuracy is not a problem of the solver, but an imperfection of the chosen physical model.

Given the above, the question arises: what exactly did the author want to show or compare?

In essence, the gas-separation membrane model implemented in a free software package was validated with a model based on commercial software. Both models have been proven to be equally applicable for predicting the membrane performance. However, the accuracy of these models cannot be linked to the software, but is primarily determined by the convergence of the solvers and the applicability of the underlying physical model of the process.

Considering that the presented results do not correspond to the title and the stated goals, I believe that the author should seriously revise the article.:

• Thanks for the comments. There are two objectives in this study; first to test the accuracy of the available mathematical model for membrane simulation. The model assumed zero effect of back mixing, polarization concentration, and plasticization. This can be true when comparing the membrane with experimental data at the laboratory level with pure gas feeds. However, in reality with mixed feed streams, polarization concentration and plasticization are known to alter membrane permeance with time and this was investigated by comparing the simulation results with industrial data. It was found that the previously mentioned effects should be taken into consideration when modeling the membrane for industrial applications.
• The second objective was to test the capability and precision of simulators in solving the membrane equation because numerical methods are needed. These methods can vary from one simulator to another. As the reviewer stated, we found that the simulators’ solvers have a minor effect on the variation of results from field data.
• The following was updated:
  • Title (Lines 1).
  • Introduction (Lines 57-58)
  • Abstract (Lines 21-23).
  • Results and Discussion were updated accordingly (Lines 193-199)
  • Conclusion (Lines 235-236).

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author has addressed the concerns and this manuscript should be accepted. Thank you,. 

Reviewer 2 Report

Comments and Suggestions for Authors

The addition of the introduction and conclusion and a more detailed discussion of the obtained model accuracy values clarified the idea and worth of the study. Therefore, it is recommended to accept the manuscript in present form.