24小时热门版块排行榜

返回列表

当前只显示满足指定条件的回帖，点击这里查看本话题的所有回帖

jasperecust

金虫 (小有名气)

应助: 1 (幼儿园)
金币: 1486.4
散金: 8
红花: 1
帖子: 136
在线: 69.1小时
虫号: 400010
注册: 2007-06-12
专业: 人工智能与知识工程

[求助] 被一个审稿人提了9条意见，还有戏吗？

Reviewer Comments to Author:

Reviewer: 1
Comments to the Author
The authors have applied a multi-objective optimization approach towards estimation of tray efficiencies of distillation columns. The algorithm was applied towards an industrial distillation column but tray efficiency values obtained were not validated with experimental data. The temperatures of various trays predicted by the model were shown to be in agreement with experimentally measured values but the former were obtained by minimization of differences with the latter. In this regard, there is limited evidence to support the validity of the proposed algorithm and substantially more work is required to enhance the credibility of this manuscript. As such, this manuscript is not recommended for publication in its present form and the authors are advised to refer to the following comments for improving their work.

1. The objective functions presented as equation (2) have not been properly defined. Firstly, the meanings of all variables with primes and how they differ from corresponding variables without primes have not been explained anywhere in the manuscript. Secondly, it is mentioned that the optimization procedure started with dividing the entire distillation column into p sections. However, this variable “p” does not appear anywhere in the objective function and does not seem to be related to any variable in the objective function.

2. The flow diagram provided in Figure 1 is not self-explanatory and at least a brief explanation of the procedure used for calculating tray efficiencies should be provided. Furthermore, it appears that three outputs could be derived from the procedure shown, referred to as mechanism model, industrial distillation column model and neural network model. The relations and differences between these models and how they are derived and utilized subsequently should be explained clearly.

3. As with point 1 above, the objective functions presented on page 12 of the manuscript have not been defined clearly. The difference between variables with subscript “m” and corresponding variables without the subscript should be stated. If the optimization process involves minimization of the difference between measured values of temperature and composition with those calculated from a model, this should also be stated clearly.

4. Following from point 3, it is mentioned that the model of the distillation column was built with Aspen Plus but no details of the model have been provided. Sufficient details of this model need to be provided for other researchers to be able to rebuild this model independently.

5. Similarly, it seems that the authors have constructed a neural network model to predict tray efficiencies from various operating parameters. However, no details of this neural network model have been provided other than a reference to (7). To be convincing to readers for work of such a nature, details of the network architecture, training data sets, test data sets, training method, weight values derived etc should be provided in the manuscript..

6. Are there any limitations of the model in terms of the need for accurate initial guesses? For example, how would the accuracy of final solutions and time required to reach the final solutions be affected by poor initialization? The authors may need to carry out some sensitivity analyses on this aspect of the model.

7. It is not clear how the authors selected the best or most appropriate optimization algorithms for the study of the industrial distillation column based on performances obtained with the 5 benchmark problems listed in Table 1. There does not seem to be a close relation between any of these benchmark problems with the distillation column optimization problem.

8. Figure 4 shows temperature values obtained from experiments and simulations using two different models. It is not unexpected that good agreement is observed since minimization of differences in temperature values between actual measurements and simulations was part of the optimization process. On the other hand, it would be interesting to compare experimentally determined tray efficiency values with those predicted by the model as presented in Table 2. This would be a more rigorous and thus convincing validation of the model.

9. Finally, the standard of English writing in this manuscript is unacceptable in its present form and may lead to serious difficulties in understanding the technical aspects of the work reported.

Reviewer: 2
Comments to the Author
The paper needs significant revision and clarification prior to re-review for publication. It is unclear to the reviewer how the optimization algorithms are applied to the plant data. My interpretation of the approach is that optimization algorithms are used to “fit” a mass transfer efficiency to temperature and concentration values for each major section of the tower. This approach is novel and useful and probably warrants publication but it needs to be explained much more clearly.

With regard to the actual paper content: It is unclear how the Aspen simulation is linked to Matlab. One would assume that the theoretical tray count from an Aspen RADFRAC simulation is being manipulated by the optimization algorithm implemented in Matlab but this is not stated. A revision of Figure 1 might provide much needed clarification. The tray efficiency values reported in Table 2 for the stripping section are unrealistically low and probably reflect a tower operating above its hydraulic limit. The authors need to check the percent of flood for the tower bottom section. The term “load” needs to be defined and their “distillation section” is traditionally called the “rectification section”. The assignment of a single tray efficiency for each section of the column is acceptable for the case study but would not be accurate for a column which had a large and non-linear temperature profile. No representative tray hydraulic data is given and the tray type and dimensions for the base case are not included thus making independent assessment of the results impossible. Finally, the neural net (NN) calculations are not discussed in adequate detail. Publication 7 is referenced but no details concerning the application of neural nets to the base case are given. It is unclear to the reader how the data presented in Figure 4 was obtained. Specifically, one cannot determine from the manuscript how much of the actual operating column data was used to train the neural net and how much was used to validate the fit quality. At a minimum, the column operating variables used in the NN should be given.

郁闷！不知道还有没有戏？