Thermal Degradation Studies and Machine Learning Modelling of Nano-Enhanced Sugar Alcohol-Based Phase Change Materials for Medium Temperature Applications
, Srinivasnaik Mukuloth 4, Ragavanantham Shanmugam 5, Venkata Charan Kantumuchu 6, Sirisha Bhadrakali Ainapurapu 7 and Muralimohan Cheepu 8,9,*
Round 1
Reviewer 1 Report
In this paper the authors study the problem of thermal degradation studies and machine learning modelling of nano-enhanced sugar alcohol-based phase change materials for medium temperature applications. Studying the kinetic properties of nano-enhanced PCM samples for application in medium-temperature thermal storage devices is the primary focus of 180 this investigation. The purpose of this research 182 is being broadened to incorporate the creation of machine learning models to predict the stability of pure PCM and nano-enhanced PCMs.
The topic is quite interesting, but I have comments. The introduction should place the proposed approach on the background of existing and known solution presented in literature. Also the importance of the research field should be stressed.
The approach proposed in this article is interesting, but not very effective in practice. This is a beautiful theoretical mathematical model. The authors of the article need to show the applicability of the proposed model using a specific example. Then readers will understand how reliable data is obtained. Also in the article it is necessary to calculate by what percentage the accuracy increases when using the method proposed in the article? By what percent will the efficiency increase?
The presented report is at a very high scientific level. I believe that the present study has a significant scientific and applied contribution, which is strongly emphasized in the basically reporting volume. A slight clarification can be made in the abstract part, where the quality of the research can be enhanced. In the conclusions, it is necessary to describe in what is the economic effect of using this method? What is the payback period for this approach?
Author Response
Point 1: The introduction should place the proposed approach on the background of existing and known solution presented in literature. Also, the importance of the research field should be stressed.
Response 1:
The main objective of this research is to use various model free integral kinetic models to evaluate the kinetic characteristics of pure PCM and nano enhanced PCM samples for medium temperature thermal storage systems. This can be accomplished by performing non-isothermal TGA experiments on pure PCM and nano enhanced PCMs at different heating rates of 5, 10, 15 and 20 oC/min. This current research is being expanded to include the development of various machine learning models for predicting the stability of pure PCM and nano enhanced PCMs. The type of PCM, heating rates, and temperature at corresponding conversion rates were all considered as input data. As output data, the mass loss of the corresponding input data set is used. The developed machine learning models are effectively predicting experimental outcomes with the highest R2 value.
The novelty and advantage of the current research is broadly discussed in the introduction section.
Point 2: The authors of the article need to show the applicability of the proposed model using a specific example. Then readers will understand how reliable data is obtained. Also, in the article it is necessary to calculate by what percentage the accuracy increases when using the method proposed in the article? By what percent will the efficiency increase?
Response 2:
Authors thank the reviewer’s suggestions. From the literature studies, it reveals that the thermal degradation kinetic models are implemented to understand the reaction mechanism and thermal stability of the coal, bio mass and flammable materials. This information will useful for design of boilers and storage containers. However, phase change materials are most
commonly used to store the thermal energy in the form latent heat. Developing efficient methods for storing thermal energy has become an increasingly crucial aspect of modern infrastructure. For this, it is necessary to determine the thermophysical properties as well as thermal stability of the PCM samples. The addition of nano particles enhances the thermophysical properties. Meanwhile, it influences the thermal stability of the PCM samples. The current investigation can help to understand the thermal stability of PCM samples with addition nano particles.
The authors performed degradation kinetics of the nano enhanced PCM samples with the help of existing integral modal free kinetic methods. As per authors knowledge, no investigation has been performed for degradation kinetics of sugar alcohol based PCM samples particularly no literature is available for degradation studies of nano enhanced PCM samples.
Point 3: The presented report is at a very high scientific level. I believe that the present study has a significant scientific and applied contribution, which is strongly emphasized in the basically reporting volume. A slight clarification can be made in the abstract part, where the quality of the research can be enhanced. In the conclusions, it is necessary to describe in what is the economic effect of using this method? What is the payback period for this approach?
Response 3:
Authors thank the reviewer’s suggestions. The economic analysis is needed when the prepared samples are incorporated in to the thermal energy storage systems. This study only deals with thermal degradation studies of prepared PCM samples. In future, the authors are planning to conduct real time experimentation with prepared samples. Also, planning to conduct economical viability study of thermal storage systems with incorporation of these materials.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Dear author,
I revised your manuscript and I will reconsider it after major revisions. I have prepared a file, you can find it attached, where are described all my doubts and suggestions.
I wish you good work.
Comments for author File: Comments.pdf
Author Response
Point 1: Line 21: “Thermogravimetric analysis”. Please, add here the name “TGA”; then, you can use this acronymous in line 32.
Response 1: The authors have followed the reviewer’s advice and the acronymous of TGA is included in the revised manuscript.
Point 2: PEPCM was not defined. Please, write the entire name before the acronymous.
Response 2: The authors removed the acronym word “PEPCM” and instead of that used the word “pure PCM”.
Point 3: Line 47: reference [1] [3]. I suggest checking them and rewrite in a correct way, such as [1,2].
Response 3: The authors have followed the reviewer’s advice and the reference citation style has been changed and updated in the revised manuscript.
Point 4: Line 57: PV cells. Please, explicit the meaning of PV.
Response 4: The authors have followed the reviewer’s advice and the term PV is expanded as photovoltaic.
Point 5: The introduction is well organized and well described. However, from a general point of view, I would like to ask if the phase change materials described in this manuscript (D-mannitol and D-mannitol with 0.5% and 1% of multiwalled carbon nanotubes) were used in other works. If yes, where? I am interested to know if these materials had some applications in the past or if they will have an application for the future. Basically, I want to be sure that this study can be useful not just theoretically but also practically.
Response 5: Typically, D-mannitol phase change material is most commonly used organic phase change material for the medium temperature thermal storage applications. Jame, L., et al (2021) used D-mannitol PCM material for thermoelectric power generator for improving the electrical efficiency.
The thermal conductivity of the D-mannitol PCM material needs to be increased in order to achieve the desired properties of changing and discharging. This can be accomplished by distributing the carbon nanotubes out in a more even distribution. Nevertheless, the presence of these nanotubes has an impact on the PCM material's thermal stability. As a result, it is essential to conduct research on the thermal stability of the PCM with the addition of nanoparticles at a variety of heating rates. This study will provide further information regarding the temperature range at which PCM embedded thermal storage system can be safe.
Jame, L., Kumar, J. R., Perumal, S., & Moorthy, M. (2021). Experimental investigation of thermoelectric power generator using D-mannitol phase change material for transient heat recovery. ECS Journal of Solid-State Science and Technology, 10(6), 061005.
Point 6: Lines 86 – 90: I am sure is all true but adding some references would be very useful
for the readers.
Response 6: The authors have followed the reviewer’s advice and included more references in introduction section.
Point 7: Lines 134 – 135: be very careful with the subscripts in Al2O3. I suggest checking the entire manuscript because I have spotted a lot of these minor errors. The same is, for example, in line 190 with the “°” symbol. Line 199 it also misses the C of Centigrade. Line 224 the superscripts “cm-1”. These are just few examples.
Response 7: The authors have followed the reviewer’s advice and all typo errors have been fixed.
Point 8: Lines 144 – 145: “There is a statistically significant difference of about 0.5% in activation energy between the two models”. What does exactly mean? Did you make a hypothesis test? Otherwise saying 0.5% is irrelevant.
Response 8: The authors have followed the reviewer’s advice and hypothesis test was not performed for these models. Thus, the above sentence was removed and updated with new sentence as “Kinetic analysis suggests the two-dimensional phase boundary reaction model best describes palm oil oxidation”
Point 9: Lines157 – 163: Please, rewrite this paragraph reviewing the English.
Response 9: The authors have followed the reviewer’s advice and authors responded by revising the suggested paragraph with thorough consideration.
Point 10: Lines 166 – 167: “This enables the mathematical features of artificial models in which people process information”. Please, rewrite this sentence if it is important for you. I couldn’t find the meaning.
Response 10: The authors have followed the reviewer’s advice and authors responded by revising the suggested sentence with thorough consideration.
Point 11: Line 202: “Make: VB ceramics etc…” I am not sure about the use of the word “Make:”. I suggest directly using the model of the machine without writing “make”. Take this in account also for the other time that “make” was used with the same
purpose.
Response 11: The authors have followed the reviewer’s advice and included the model details in the revised manuscript.
Point 12: Equation 4: I suggest using “∙” symbols instead of “x”. And I suggest checking the font of all the equations. It must be the same. For example, the “A” present in the equation 3 and 4 has a different font. Furthermore, please describe always all the equation symbols. For example, equation 3: what are Rg, Ea? I know, what they are but still the meaning must be explicated. Moreover, about the T. It is necessary to write witch kind of Temperature is (°C, K, ?).
Response 12: The authors have followed the reviewer’s advice and described all the symbols.
Point 13: Lines 263: I think that combined the equations 4 and 5 it cannot be obtain the eq. 6.
Is there the possibility that “t” (eq. 6) must be “T”? Otherwise, I cannot explain the capital “T” in the eq. 6.
Response 13: The typo error has been rectified in eq. 6.
|
(6) |
Point 14: Line 285: Typo (E_a)
Response 14: The typo error has been fixed.
Point 15: Lines 309 – 316: Is it necessary to explain known things?
Response 15: It will make the information easier for the reader to comprehend.
Point 16: Lines 315 – 316: “This model is not very reliable because of its simplistic nature. Multiple linear regression models can be described in their broadest sense by the expression.” In my opinion it is not true that this model is not very reliable. If it is correctly applied, it is reliable. Obviously in this case it is not reliable since the phenomenon is not linear. Then, why did you use it?
Response 16: Authors agreed with the reviewer opinion. The linear regression model is reliable and simplistic nature.
Point 17: Lines 324 – 326: “For this reason, original characteristics of varying polynomial
degrees are used to generate polynomial features with the goal of improving the prediction accuracy of the linear regression model.” Have you used this method? Could you please describe it properly?
Response 17: Authors observed the polynomial regression model was exhibiting superior performance than the linear regression model which observed from the literature (Mavromatidis, L. E. et al (2013)).
Mavromatidis, L. E., Bykalyuk, A., & Lequay, H. (2013). Development of polynomial regression models for composite dynamic envelopes’ thermal performance forecasting. Applied energy, 104, 379-391.
Point 18: Lines 300 – 331: “The kernel is a radial basis function…”. Why is it radial?
Response 18: In order to establish the best possible support vector regression model, hyperparameter tuning is an essential step. The prediction performance of the SVR model is subjected to being influenced by parameters such as kernel functions, which can include polynomial, linear, and radial basis functions. The radial basis function is the one that delivers the best results out of all of these kernel functions, as seen in the table. 2.
Point 19: Line 331: “…from an easily calibrated origin”. What does it mean for you “easily”?
Response 19: Authors clearly explained about the significance of the kernel function and discussed the advantage of RBF kernel function.
“The kernel function is a technique for processing raw data and transforming it into the desired format for further analysis. The radial basis function is a generic kernel that can be applied to any dataset without any prior information.”
Point 20: Eq. 12 please describe the symbols.
Response 20: The authors have followed the reviewer’s advice and the symbols used in the eq. 12 is described properly.
Point 21: Line 345: The random forest (RF) IS an ensemble classifier.
Response 21: Reviewer's recommendation is implemented.
Point 22: Line 347: Could you explain what is a “decision three”, please?
Response 22: Decision tree stands for a collection of constraints or conditions that are organised hierarchically and that are carried out in the correct order, beginning at the tree's origin and progressing outward until they reach the tree's final node or leaf. When compared with artificial neural networks, the most significant benefit of utilising a hierarchical tree structure is that the structure is visible, which makes it much simpler to understand. To construct the DT from a collection of input datasets, an evaluation measure is applied to each evidential attribute to boost the internode's level of variability. To get the DT, multiple regressions on the dataset as well as recursive partitioning of the dataset are utilised. The technique for data splitting is carried out in an iterative fashion beginning at the root node and continuing until an established termination condition is met. Only specified terminal nodes or leaves are included in the scope of application for a simple regression model. To boost the tree's generalisation capacity, reduce its structural complexity. When pruning, the total occurrences in a node can be considered. The method known as random forest is one that gathers the results of multiple decision trees into a collective pool and then chooses the most optimal one from among those pools.
Point 23: Lines 349 – 351: “The reason that RF is being utilized in this work is due to the fact that it is less likely to succumb to over-fitting and that it has previously demonstrated good regression results”. Previously, when?
Response 23:
As per the reviewer suggestion, the following sentence is reframed and justified with the help of literature support.
“The use of RF is justified by the fact that it is robust to overfitting and has shown promising regression outcomes in the previous literature (Ahmad, M. W. et al (2018)).
Ahmad, M. W., Reynolds, J., & Rezgui, Y. (2018). Predictive modelling for solar thermal energy systems: A comparison of support vector regression, random forest, extra trees and regression trees. Journal of cleaner production, 203, 810-821.”
Point 24: Lines 351 – 353: “The random forest regression is given a sample that has already
undergone pre-processing consisting of n samples.” What kind of pre-processing was used?
Response 24: Typically, the specified input and output parameters have a large range, which limits the predictability of the model. Using eq. 25, a model normalisation or pre-processing technique is used to improve the prediction efficiency of the model. The random forest regression is given a sample that has already undergone pre-processing consisting of n samples.
Point 25: Line 356: “The results of past regressions suggest…” which results?
Response 25:
The authors determined that this sentence does not make sense and rewrote it so that it does make sense. The following corrected sentence are updated in revised manuscript.
“From the literature (Ahmad, M. W. et al (2018)), it is clear that the RF regression model outperforms than the decision tree model for such data. The RF has a higher accuracy than the decision tree and less of a danger of overfitting the data. Also, it has outperformed other classifiers because it is easy to interpret, non-parametric, and performs well on a large dataset.”
Ahmad, M. W., Reynolds, J., & Rezgui, Y. Predictive modelling for solar thermal energy systems: A comparison of support vector regression, random forest, extra trees and regression trees. Journal of cleaner production, 2018,203, 810-821.
Point 26: Line 358: “…the fact that in certain instances…” which instances?
Response 26:
The authors determined that this sentence does not make sense and rewrote it so that it does make sense. The following corrected sentence are updated in revised manuscript.
“From the literature (Ahmad, M. W. et al (2018)), it is clear that the RF regression model outperforms than the decision tree model for such data. The RF has a higher accuracy than the decision tree and less of a danger of overfitting the data. Also, it has outperformed other classifiers because it is easy to interpret, non-parametric, and performs well on a large dataset.”
Ahmad, M. W., Reynolds, J., & Rezgui, Y. Predictive modelling for solar thermal energy systems: A comparison of support vector regression, random forest, extra trees and regression trees. Journal of cleaner production, 2018,203, 810-821.
Point 27: Line 362: “For highly nonlinear problems…”. Your problem is not highly nonlinear. It is for sure nonlinar but not highly nonlinear.
Response 27: The authors have followed the reviewer’s advice and the sentence is updated the term as “nonlinear”.
Point 28: Eq. 14: I suggest describing better the magnitudes.
Response 28: Authors thank the reviewer suggestion. The terms used in the eq. 14 doesn’t involve the magnitude.
Point 29: Eq. 15: Is Y an arbitrary distribution with those parameters or does it have a specific
functional form? ?2 is the variance?
Response 29: The term ?2 is known variance data points.
Point 30: Eq. 19, 20, and 21: please, describe properly the process of these equations.
Response 30: The equations are described properly and updated in the revised manuscript.
Point 31: Line 387: about the ANN. In general, an ANN has more than 3 layers. You are using just 3 of them so ANN has 3 layers, yes, but only in your specific case. Thus, please, correct the sentence.
Response 31: The authors have followed the reviewer’s advice. As per reviewers suggestion, the authors corrected the sentence.
Point 32: Eq. 22: it does not appear Xj. I Understood that you are using Zj but you must
explain it in the text.
Response 32: Authors thank the reviewer suggestion. The term Zj is explained in the text and updated in the revised manuscript.
Point 33: Lines 416 – 418: If there is the optimal calculation of the number of neurons, why
do you use trial and error method?
Response 33: Authors thank the reviewer suggestion. It is possible to determine the appropriate number of hidden neurones in the hidden layer by utilising the eq. 24, which is a general thumb rule that the researchers use to follow. The optimal neurones found by solving this equation can't be considered to be the ultimate solution. The neuronal independence test has been carried out with a sufficient number of neurones, falling somewhere within the optimal range.
Author Response File: Author Response.pdf
Reviewer 3 Report
Thermal degradation studies and machine learning modelling of nano-enhanced sugar alcohol-based phase change materials for medium temperature applications
Q: 1) According to the progress in the field, it would be better for showing the advantages and novelties of your work, the advances to the field according to your work, and the scientific benefit for the community.
Q: 2) There are many methods for studying the kinetic properties. Line 278-279, this work told that “FWO, KAS, and Starink methods are the most often used integral methods for determining kinetic parameters.” What are any reasons for using Ozawa, KAS and Starink methods to study the kinetic parameter such as the activation energy?
Q:3) The full name should be given before the abbreviation name (at line 47 and abstract part for PCM).
Q:4) Figure 1 and Figure 2, Inconsistency name in the caption and label diffraction peaks. The consistency name should be given.
Q:5) Figure 2, The major peaks and the % transmittance do not label in image, not clear enough to observe. The FTIR spectrum of MWCNT does not clear, why? The FTIR spectra of PCM and PCM+MWCNT are quite the same and why the FTIR spectrum does not show the characteristic peaks of MWCNT?
Q:6) line 230-231, this work explains that the peak at 876 cm-1 corresponds to O-H bending. Could you please explain more detail how the O-H can bend?
Q:7) Equation 2, could you please change the symbol of ma and mL to mα and mf ? normally, many papers in solid state kinetic use mα and mf..
Q:8) Equations 3-7, the meaning or the definition of symbols do not given.
Q:9) Please check your manuscripts for typos and grammar again (e.g. E_a should use Ea instead, line 285).
Q:10) Inconsistency in writing the unit of temperature.
Q:11) According to Ozawa, KAS and Starink methods, are there the linear fitting methods? Figure 7-9, the R2 should be given in Ozawa, KAS and Starink plots.
Q:12) Could you please explain how you calculate the activation energies from Ozawa, KAS and Starink methods?
Q:13) Figure 7d, why does the activation energy show the highest value at α = 0.2? In my opinion, the trend should be inclined.
Q14): No references for the kinetic equations.
Author Response
Point 1: According to the progress in the field, it would be better for showing the advantages and novelties of your work, the advances to the field according to your work, and the scientific benefit for the community.
Response 1:
The main objective of this research is to use various model free integral kinetic models to evaluate the kinetic characteristics of pure PCM and nano enhanced PCM samples for medium temperature thermal storage systems. This can be accomplished by performing non-isothermal TGA experiments on pure PCM and nano enhanced PCMs at different heating rates of 5, 10, 15 and 20 oC/min. This current research is being expanded to include the development of various machine learning models for predicting the stability of pure PCM and nano enhanced PCMs. The type of PCM, heating rates, and temperature at corresponding conversion rates were all considered as input data. As output data, the mass loss of the corresponding input data set is used. The developed machine learning models are effectively predicting experimental outcomes with the highest R2 value.
The novelty and advantage of the current research is broadly discussed in the introduction section.
Point 2: There are many methods for studying the kinetic properties. Line 278-279, this work told that “FWO, KAS, and Starink methods are the most often used integral methods for determining kinetic parameters.” What are any reasons for using Ozawa, KAS and Starink methods to study the kinetic parameter such as the activation energy?
Response 2: Authors agreed with the reviewer’s opinion. Kinetic models are categorized into two approaches that is, model fitting methods and model free methods. In model fitting models, the best statistically fitted model was identified from among various fitted kinetic models. This best-fit kinetic model can determine activation energy and pre-exponential parameters. Typically, using these models, a single heating rate is sufficient to calculate the kinetic characteristics of selected samples. The main disadvantage of this approach is that it does not account for changes in activation energy as a function of heating rate, which can be overcome by using iso-conventional kinetic methods (also known as model free kinetic methods). These methods have a number of advantages, including being simple to use and estimating the reaction model with a small number of errors. These iso-conventional kinetic methods are further classified as differential methods, such as Friedman, and integral methods, such as KAS, FWO, Vyazovkin, and Starink method, based on the methodology. As compared with the differential methods, integral methods such as KAS, FWO, and Starink method are became more popular due to following advantages.
- This kinetic approach (FWO) is a simplified and linear technique.
- A simplistic temperature integral approximation has been considered to derive the FWO kinetic model.
- The primary assumption in this FWO kinetic technique is that the activation energy maintains stable from the beginning of the reaction to the degree of each conversion point.
Point 3: The full name should be given before the abbreviation name (at line 47 and abstract part for PCM).
Response 3: The authors thank the reviewers’ suggestions. The term PCM is expanded and the same is updated in revised manuscript.
Point 4: Figure 1 and Figure 2, Inconsistency name in the caption and label diffraction peaks. The consistency name should be given.
Response 4: The labelling is given in the XRD and FTIR plots (i.e., Figure 1 and 2) in the updated manuscript.
Point 5: Figure 2, The major peaks and the % transmittance do not label in image, not clear enough to observe. The FTIR spectrum of MWCNT does not clear, why? The FTIR spectra of PCM and PCM+MWCNT are quite the same and why the FTIR spectrum does not show the characteristic peaks of MWCNT?
Response 5: The labelling is given in the FTIR plot. From the figure 2, it is clear that the FTIR peaks of MWCNT particles doesn’t have any functional groups. The addition of MWCNT particles in to pure PCM does not change the functional groups of the PCM which indicates that the nano enhanced PCM particles showing excellent chemical stability.
Point 6: line 230-231, this work explains that the peak at 876 cm-1 corresponds to O-H bending. Could you please explain more detail how the O-H can bend?
Response 6: Authors thank for the reviewer’s suggestions. The authors performed a thorough analysis of the FTIR spectra and determined that the FTIR peak located at 876 cm-1 represents the =C-H bond. The authors apologise for this mistake, and corrected spectral representation is updated in revised manuscript.
Point 7: Equation 2, could you please change the symbol of ma and mL to mα and mf? normally, many papers in solid state kinetic use mα and mf.
Response 7: The authors have followed the reviewer’s advice. The reviewer’s suggested nomenclature is updated in revised manuscript.
Point 8: Equations 3-7, the meaning or the definition of symbols do not given.
Response 8: The authors have followed the reviewer’s advice and the symbols in the eq. 3-7 are denoted and updated in the revised manuscript.
Point 9: Please check your manuscripts for typos and grammar again (e.g. E_a should use Ea instead, line 285).
Response 9: All typo and grammar error has been fixed.
Point 10: Inconsistency in writing the unit of temperature.
Response 10: The unit of temperature is properly given in updated manuscript.
Point 11: According to Ozawa, KAS and Starink methods, are there the linear fitting methods? Figure 7-9, the R2 should be given in Ozawa, KAS and Starink plots.
Response 11: At a number of different conversion points, the KAS, Ozawa, and Starink techniques all produce R2 values that are greater than 0.92, suggesting that the calculated activation energy values are within accepted range. (Balasubramanian, K. R. et al (2022)).
Balasubramanian, K. R., Ravi Kumar, K., Sathiya Prabhakaran, S. P., Jinshah, B. S., & Abhishek, N. (2022). Thermal degradation studies and hybrid neural network modelling of eutectic phase change material composites. International Journal of Energy Research, 46(11), 15733-15755.
Point 12: Could you please explain how you calculate the activation energies from Ozawa, KAS and Starink methods?
Response 12: The eq. 8, 9 and 10 are used to determine the activation energy of the prepared PCM samples using KAS, FWO and Starink kinetic models. The degradation temperature of the prepared PCM samples is determined at each degree of conversion. Plot the graphs between , Vs and at each conversion points. From slope of these plots, the activation energy value can be determined. The procedural steps for determine the activation energy value is discussed in section 2.2.
Point 13: Figure 7d, why does the activation energy show the highest value at α = 0.2? In my opinion, the trend should be inclined.
Response 13: The addition of MWCNT particles shows whimsical behaviour with PCM particles. The addition of carbon particles will slow down the degradation of PCM at lower conversion points. However, higher mass fraction of the MWCNT can enhances the porosity which leads to improves degradation of PCM samples with respect to degree of conversion (Balasubramanian, K. R. et al (2022)).
Balasubramanian, K. R., Ravi Kumar, K., Sathiya Prabhakaran, S. P., Jinshah, B. S., & Abhishek, N. (2022). Thermal degradation studies and hybrid neural network modelling of eutectic phase change material composites. International Journal of Energy Research, 46(11), 15733-15755.
Point 14: No references for the kinetic equations.
Response 14: The references are given for all kinetic equations.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Dear Authors,
I would like to thank you for your work and for taking my suggestions into consideration. I appreciated it very much. However, I do not know why, in the first round all the comments I had written were not present. Therefore, you had not the chance to respond. I think those comments were very important; thus, I decided to bring them back in this second round of review.
Warning: for these revisions consider the lines from the first version of the manuscript (first submission). Thank you.
1. Eq. 24: where did you find this equation?
2. Eq. 25: please, define the symbols.
3. Line 428: “The following figures…”. Which ones?
4. Line 431 – 432: “As was mentioned before…”. Where?
5. Line 433: “As can be seen from the numbers…”. Which numbers?
6. Line 435: “This was seen in the works”. Which work? In addition, please, use the symbol “°” for degrees.
Please, read carefully and write slower.
7. Lines 450 – 503: I truly believe that would be better to put all these numbers in a table. It is very complicated, for a reader, to follow the discussion in this way.
8. Table 1: How many times did you repeat the thermogravimetric analysis on the same kind of sample? It would be preferable to add the experimental error. Furthermore, could you explain why did you average the values?
9. Line 603 – 607: “Finding the exact values of the hyperparameters is a vital component of developing a successful model, especially for a parameter-sensitive algorithm like a Random Forest, support vector machine, Gaussian process regression, or ANN model. This is because establishing proper hyperparameter values is an essential component of evaluating a model's performance.” Translation = It is essential because it is essential.
Please, revise the sentence.
10. Line 615: “It was shown…”. Where?
11. Line 646 – 649: I would like to know in which sense are you matching the results? These models should be able to predict new results, with your explanation it seems you just matched the data you had. In my opinion the article lacks a clear explanation of the choice of Train, Validation, and Test set, as well as their dimension and exact composition.
In the end, from the article is not clear if the predictive power of these models was tested in new experimental conditions (i.e., heating rate, and/or percentage of MWCNT not used in the training).
Thank you. I wish you good work.
Author Response
Reviewer-2
Point 1: Eq. 24: where did you find this equation?
Response 1: The authors have followed the reviewer’s advice and the eq. 24 is sited properly in the revised manuscript.
Point 2: Eq. 25: please, define the symbols.
Response 2: The authors have followed the reviewer’s advice and the eq. 25 are defined properly in the revised manuscript.
Point 3: Line 428: “The following figures…”. Which ones?
Response 3: The authors thank the reviewer’s suggestions and the figures are properly cited in the revised manuscript.
Point 4: Line 431 – 432: “As was mentioned before…”. Where?
Response 4: The authors thank the reviewer’s suggestions and the sentence is reframed and deleted in the revised manuscript. The experimental details haven’t discussed before.
Point 5: Line 433: “As can be seen from the numbers…”. Which numbers?
Response 5: The authors thank the reviewer’s suggestions and the sentence has been rewritten in a more meaningful manner and updated in revised manuscript.
Point 6: Line 435: “This was seen in the works”. Which work? In addition, please, use the symbol “°” for degrees.
Response 6: The authors thank the reviewer’s suggestions and the sentence has been rewritten in a more meaningful manner and updated in revised manuscript. Also, authors replaced the word degree Celsius with the term “oC”.
Point 7: Lines 450 – 503: I truly believe that would be better to put all these numbers in a table. It is very complicated, for a reader, to follow the discussion in this way.
Response 7: The authors thank the reviewer’s suggestions and with the aid of table numbers, the discussion was held in an appropriate manner.
Point 8: Table 1: How many times did you repeat the thermogravimetric analysis on the same kind of sample? It would be preferable to add the experimental error. Furthermore, could you explain why did you average the values?
Response 8: Authors thank the reviewer suggestion; authors didn’t repeat the thermogravimetric analysis on the same kind of sample. Various kinetic models (i.e., KAS, OFW and starink) are applied to the prepared PCM samples. The representation of table.1 is updated in the revised manuscript for better understanding of readers. The experimental error details were updated in the revised manuscript.
Authors thank the reviewer suggestions; the authors believe that it is unnecessary to calculate the average value of activation energy values in this study.
Point 9: Line 603 – 607: “Finding the exact values of the hyperparameters is a vital component of developing a successful model, especially for a parameter-sensitive algorithm like a Random Forest, support vector machine, Gaussian process regression, or ANN model. This is because establishing proper hyperparameter values is an essential component of evaluating a model's performance.” Translation = It is essential because it is essential.
Response 9: The authors thank the reviewer’s suggestions and the sentence has been rewritten in a more meaningful manner and updated in revised manuscript.
Point 10: Line 615: “It was shown…”. Where?
Response 10: The authors thank the reviewer’s suggestions and the sentence has been rewritten in a more meaningful manner and updated in revised manuscript.
Point 11: Line 646 – 649: I would like to know in which sense are you matching the results? These models should be able to predict new results, with your explanation it seems you just matched the data you had. In my opinion the article lacks a clear explanation of the choice of Train, Validation, and Test set, as well as their dimension and exact composition.
In the end, from the article is not clear if the predictive power of these models was tested in new experimental conditions (i.e., heating rate, and/or percentage of MWCNT not used in the training).
Response 11: The authors thank the reviewer’s suggestions; the model accuracy test has been conducted with validation data set which is obtained from conducting the TGA experiment on PCM + 0.5 % MWCNT sample with heating rate of 7.5 oC/min. Among all, gaussian process regression model effectively predicting the validation data set with lowest RMSE value of 1.0632 and highest R2 value of 0.9997. The prediction performance of the developed machine learning models are clearly discussed in revised manuscript.
Author Response File: Author Response.pdf
