Optimizing the Neural Network Architecture for Automation of the Tailored UV Post-Treatment of Photopolymer Printing Plates

Round 1

Reviewer 1 Report

This is an interesting piece of work.  The work is generally well presented and easy to follow but requires a few minor improvements to the writing and clarifications, which are itemised by line.

 

Abstract

Line 10.  Don't express in terms of aims, just state that you used SFE data to build and optimize the network

 

Line 18.  Be careful with the claim that this approach has enabled the automation of the process.  This suggests the model is already being used to tailor the properties, is this correct, or does it merely show potential for use in this way?

 

Include some indication of the success of the technique (see later points in conclusion).

 

Introduction

Line 25 remove “the” from “the packaging..”

Line 75 – surface energy instead of surface tension?

Line 76 - are waxes carbohydrates?  maybe just better to state waxes, rather than carbohydrate compounds

Line 78 remove “the” from “the degradation..”

Line 79 remove “the” from “the UV..”

Line 84 and throughout – I think the term “wetting” would be better than “adsorption”.

Line 99 and 100 - What systems is this reference describing?  Unless you state what this is used for, it could be inferred that it is used for flexo plates, which it isn't.

 

Methods

Table 1.  I'm not sure you need Table 1 and it could be covered with text, one type was fixed at 10 min while the other was varied from 0 to 20 min in 1 min intervals  Indicate why 5 min and not 10 min for plate C.

 

Line 149 - what do you mean by standard ink?  Indicate the substrate used as this has an influence on the wetting behaviour and print quality.

General comment – It might be worth including contact angle data in supplementary material - this could be used to look at the variability (or  standard deviation) between measurements - I appreciate this is not practical for SFE - but it would be useful to gauge the reliability of the method, noting that individual SFE values jump around a bit.  The model is only as good as the data, so if SFE is subject to variability, this should be mentioned.

 

Line 159 – suggest “with repeat drops taken on a different locations” (the word parts suggests lines, dots etc.)

Line 159 – Reference not referent

Line 159 - do you need to mention drop shape?  all drops are spherical

Line 174 – Can you provide a reference for Keras

Line 179 - Can you give a reference to somewhere that explains these different models?

Line 181 – mention that models were trained separately for the three different printing plates

Line 200 – “after” rather than “for”

 

General point on the charts provided - You have 0 to 20 in 1 min intervals, so why don't you show all the data, since all the data was used in the models and this might better highlight some of the less reliable data points and overall trends.  I assume the first reading is for or 0,10 treatment, not 0,0.  Please provide the datum value somewhere.

 

Question - These plates are varied in their response to UV and A has a much lower starting SFE, why is this?  Can you provide an answer in the paper.

 

Line 215 – Better to say could be explained, rather than can be explained - I don't think you can be sure unless you verify with some measurement of surface chemistry

Line 224 - had a stronger effect

Line 236 - this relates to the previous points on the chart, but this inflection would be better highlighted if you used all the data points rather than every 5 minutes

 

Line 246 – same point as previously on carbohydrates

Lines 248-249 this mechanism should be pointed out earlier in methods when you describe the treatments used

Line 251 – “but at the same time”  this sentence is a unclear from here.  I would break it up.

Line 257 – Title for section is at the end of the page.  If this happens,, it is best to move it forward so that it is on the top of the following page

Line 259 - demonstrated outlined edges rather than had the outlined edges

Line 269 - negligible rather than neglectable

Line 270 - if it the component is negligible, how can it influence the result.  Don't you mean that the transfer is dominated by dispersive forces, with negligible polar component

 

General point on microscope images – these are not very high quality.  Can images be provided without compression, can a scale bar be added?

 

Line 278 and 298 – apparent rather than visible

Line 279 result in rather than resulted with

 

Figure 7 - I don't understand how these lines are wavy - is the plate distorted as it is too soft?  Please clarify.

 

Line 287 – resulted in

Line 298 – “in an” rather than “with the”

Line 301 – note that the crosslinking increases the hardness of the plate

 

General points - This information helps the reader interpret the effect of plate surface energy on the print, but this is of no consequence for the model - which only looks at predicting SFE from treatment parameters.  I am not suggesting you do this but the model would be useful if it could also indicate print outcome.  This is potentially more interesting as the print depends on hardness and SFE which are both affected by UV treatment yet have different effects.

 

We don't know anything about the ink and the ink properties would also have an influence on the print

 

It might be worth drawing a firm conclusion on the effect of UV treatment on print quality here.  why are the plates different to one another in their behaviour.  Only plate C seems to give "acceptable" results.

 

Section  3.3 – I have not provided many comments here as this technique is outside my areas of expertise (although I am familiar with its use).  My points mainly relate to the applicability of the findings.

 

In Figure 12 you present the best model for each plate.  My question is would it be practical to use different models for different plates if you wanted to develop a predictive tool; this looks a bit like "cherry picking".  If you use a plate that is not in this study, how would you know which model is best?  Is there one that performs generally better than the others?

 

You present MSE, but I don't get a feel for how close a prediction could be achieved with this technique.  Also, we know how sensitive print outcomes are to SFE/UV treatment, so will this technique give a recommended treatment that will give a decent print, or is it not sensitive enough?

 

Conclusion

Line 383 – the SFE effect is noted, but hardening of the plate is not, this also occurs and will affect the print quality

 

General point – The conclusion does not state how reliable this approach could be for predicting SFE, and whether this would be sensitive enough to target certain printing outcomes.  There is also the point made earlier about whether this model could be expanded by bringing in print outcomes, and different ink types, since this is really what the printer is ultimately interested in.

 

 

 

 

Generally this is well written and easy to follow.  I have presented an itemised list of minor changes in the comments.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The main question is the automation fo some processes during the production fo flexo printing forms as well as UV post-treatment itself, as a very important part of post-treatment for flexo forms.

The topic is original, with an approach and all the relevant and technical research that the scientist considers within this research.

New approach and the new way of automation especially in this field of graphic production. It was not considered earlier in this way.

Authors should consider applying some more video analysis of the results (with the two or three different measuring devices) and also some more detail statistical analysis.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf