Determination of Coniferous Wood’s Compressive Strength by SE-DenseNet Model Combined with Near-Infrared Spectroscopy

Round 1

Reviewer 1 Report

The authors have submitted an article that outlines the important investigation of the Coniferous Wood's Compressive Strength by SE-DenseNet Model Combined with Near-Infrared Spectros- copy. The results of this study are highly novel and are of great importance for the scientific and wood industry point of view.

The title and keywords accurately reflect the content of the manuscript. The authors of this manuscript gave us a clear introduction to the study they have done with an overall good literature review. The aim of the study is clearly defined. The materials and methods are described clearly with sufficient details of the performed measurements and the measurement techniques are appropriate to resolve the stated objectives of the study. The obtained results are well presented in an unbiased, detailed, clear and easily comparable manner where you can clearly draw the conclusion. The discussion is well written and the conclusion can be easily extrapolated from it. References consist of appropriate and relevant papers.

 

According to my opinion, the manuscript should be accepted for publication in Applied Sciences Journal.

Author Response

Dear reviewer, thank you for your recognition of our manuscript!

Reviewer 2 Report

The paper reports a study of conifer wood discrimination through near-infrared spectroscopy together with SE-DenseNet model which can realise end-to-end prediction without complex structural dimension reduction. The study is interesting and can have applications in different fields. The results are well described and clear.

Author Response

Dear reviewer, thank you for your recognition of our manuscript!

Reviewer 3 Report

The manuscript is well detailed, need just some minor corrections.

Comments for author File: Comments.pdf

Author Response

Dear reviewer, thank you for your recognition of our manuscript!

We have made some modifications according to your suggestions. Please see the attachment!

Author Response File: Author Response.docx

Reviewer 4 Report

Dear authors,

Your work increases the knowledge on the use of non-invasive methodology in the prediction of softwood strength in fields currently under study with relevance to the wood industry, construction and architecture, environmental sustainability, etc. The paper is well thought out, executed and written, however, in my opinion, some minor and major modifications, clarifications and changes are required to improve the quality of the manuscript and make it suitable for publication in Applied Sciences 2022:

50, 58, 62, etc.  Please define the first time throughout the document the abbreviations MOE, MOR, RMSEC, RMSEP, etc. for example: (modulus of elasticity), (modulus of rupture), (root-mean-standard error for calibration), (Root Mean Square Error of Prediction), etc. respectively.

92 "larch, hemlock, and mongolica" appears at the beginning while "200 specimens of larch, hemlock, and camphor pine," appears later, perhaps camphor pine forest (Pinus sylvestris var. mongolica)? Please, clarify.

120 If it is possible to enumerate traditional modelling methods.

140-141 is a repeated sentence.232 Figure 5. in the dotted box X1 to X2 process should add somewhere something like "example detail X1 to X2" i.e. it should be understood that the process is the same for all X1 to Xn.

Table 1. Place spaces correctly: Examples: 1 × 1,4c 1×1,0.5c n×n

245-246 In my opinion this statement should be better justified or briefly expanded.

Figure 12. Please, provide values for each PC on both axes (% of the variance). It is important to know the total percentage of the variance explained by components 1 and 2. Also indicate a and b in the figure.

274-435. It is necessary to add  citations and discussion with the current international publications since the discussion has been practically omitted throughout section 4: Results and discussion. This is a major change that must be solved and cannot be missed.

437-439 To my understanding this sentence is not an own conclusion of the work but rather an extension already exposed in material and methods.

440-445 Add "for the types of wood tested".

446-454 Summarise without adding specific values: these are already indicated in results. Conclusions should be clear, concise, own and extrapolable/reproducible. This is a major change.

----

Congratulations for the work done and good luck with your new research in favor of expanding knowledge.

Best regards

 

 

 

 

Author Response

Dear reviewer, thank you for your recognition of our manuscript!

We have made some major cahnges according to your suggestions. Please see the attachment!

Author Response File: Author Response.docx

Reviewer 5 Report

In general, this is a good paper. Authors need to provide more references to all the methods they use. Also, authors need to be careful regarding the chemometrics terms used. I tried to correct a few but a comprehensive review of the article is needed.

 

Line 22 – “better” - better than what? – the use of better implies comparison with something else

Line 42 – spell out NIRS

Lines 43-46 – I am not sure what that sentence mean

Line 50 – spell out MOE and MOR

Line 54 – better than what?

Line 58 – spell out RMSECV

Line 58 – R2 – 2 should be superscript. Make this edit throughout the manuscript

Line 62 – add units where applicable. Is RP R for prediction? If so, explain.

Line 62 – spell out RMSEP

Line 94 – for the standard, add more details: “In accordance with standard GB/T 1935-2009, …”

Line 102 – 5mm is most likely not the fiber diameter. Please use the right value

Line 108 – what is the vendor of SpectraSuite?

Section 2.2 – was the fiber directly against the wood? Was any optics used? Was the fiber optics probe providing illumination and collection? If not, how was it illuminated? Please add more details. How many co-adds were used per scan? What reference standard was used? 99% or lower? Please provide details about the wood testing system? Brand? Load cell? …

Lines 119-120 – “Then, the spectra are processed and done dimension reduction.” – I presume it is means “and dimension reduction performed”

Line 120 – what are the “traditional modeling methods”?

Line 129 – provide more information about the standards similar to line 94.

Line 135 – spell out PLSR

Line 140 – “Therefore Consequently” – pick one

Line 146 – MSC is multiplicative Scattering Correction, not Multiple.

Lines 147-148 – “The scattering effect is reduced by mean centering, normalization, and Vector Normalization (VN), while the centering method reduces the errors caused by matrix inverse mathematical operations.” – means centering and normalization of centering methods, not methods to reduce scatter. That is what SVN and MSC are for. Please correct.

Line 151 – please provide more information about the wavelet transform.

Lines 170 and 176 – do authors mean regression instead of correction? Same for line 266.

Line 194 – multiplies, not multiplizes

Line 228 – fix formatting

Table 1 – what are the 3 vertical dots representing?

Line 254 – spell out NER and TPR

Figure 6 – what does the red circle correspond to? The legend has a sample indicator for “ignore” but it does not seem to be present on the plot.

Lines 303-304 – “SPXY is generally superior to KS and random selection methods in data set division, so SPXY is chosen to divide the training set and test set by 4:1.” – spell out SPXY and KS. Also, explain that this step is to select samples and the SPXY and KF are sample selection methods.

Lines 305-306 – what is meant by “the mean of the evaluations is used as the final evaluation index”

Line 308 – provide more information about the standards similar to line 94.

Figure 14 – why does the x axis have decimals? The number of component is a whol;e number.

Line 356 – what is a “PLS mastery scores”

Table 5 – was the preprocessing selected in table 4 used as input to the results presented in table 5?

Line 371 – spell out RF

Figure 15a – there appears that 1 mongolica is identified as a larch sample and 2 hemlock are identified as mongolica. Please explain these results. Could it be a labeling mistake?

Line 385 – “with PLS-DA selected with a principal number of 3;” – I think authors mean the number of PLS factors. Same comment as for line 413

Figures 17 d/e/f and not needed

Section 4.2.3 – Can the regression models be used to classify the grades? Are the results better or worse than the classification algorithms?

Lines 452-454 – add units for RMSE values

 

Author Response

Dear reviewer, thank you for your recognition of our manuscript!

We have made a lot of changes according to your suggestions. Please see the attachment!

Author Response File: Author Response.docx

Round 2

Reviewer 4 Report

The manuscript has been sufficiently improved to be published in Applied Sciences, 2022.

Author Response

We sincerely thank you for your recognition and support of the manuscript！

Reviewer 5 Report

Thank you for addressing my comments. I noted a few additional items.

line 60 - squares. not square

line 286 - precision is TP/(TP+FP). I can't find a reference for what (TP+TN)/n is. but it is not precision.

line 399 - add % after 87.787

 

Author Response

We have made some minor modifications according to your suggestion:

(1):"square" has been changed to "squares".

(2):"precision" has been corrected to calculate by TP/(TP+FP).

(3):"%" has been added after 87.787.

We sincerely thank you for your recognition and support of the manuscript!