Review Reports - Defect Recognition of Roll-to-Roll Printed Conductors Using Dark Lock-in Thermography and Localized Segmentation

Round 1

Reviewer 1 Report

Zheng et al. reported a method for the defect detection of the printed conductors, which is semi-automotive and may provide a powerful tool for the quality control. The results are well demonstrated and discussed. I recommend the publication of this work with minor revision.

What is the efficiency of this method compared with other method as the segmentation localization may take a longer time?
Author only present the result from conductors with the same length, how the length of the conductor affects the result?
Line 335 the correction of the caption is required.
Author separately show the result from fully cracked and partially cracked conductor. How it will be when two types of cracking coexist in one view?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This work focuses to visualize the defects of R2R printed silver conductors on flexible plastic substrates using dark lock-in infrared thermography inspection system. The authors have proposed two-state automated defect recognition method to detect and localize the electrical defect caused by the breakage on the printed conductive wires. The manuscript is well written. This version of the manuscript is suitable for the publication in Applied Sciences.

Author Response

The authors would like to thank the reviewer for the comments and recommendation.

Reviewer 3 Report

The manuscript by Zheng et al. reports on a DLIT inspection system whit automated defect recognition (ADR) methodology to detect and localize electrical defects based on localized segmentation and thresholding methods.

The strength of this work is the development of a sequence of algorithms for the post-processing of the DLIT thermal images (noise reduction and signal amplification) to efficiently identify defects.

Nevertheless, the manuscript is more like an EU Project Report than a scientific paper. The algorithms used in this work are not novel and are not automated as claimed in the title. Moreover, in this work there is no quantitative information regarding the defect size, shape and there isn’t any correlation with the resistance/conductivity of the printed tracks nor with the fabrication process, as required for the finally targeted applications (OPVs & OLEDs). Additionally, clear results regarding the efficiency and repeatability of the recognition process are missing.

Furthermore, is mentioned, “the thermography inspection system will be finally implemented and validated on a roll-to-roll (R2R) electrical testing line at VTT (called TESLA line)”, which is not presented in the paper. Instead, the presented study is about the ex-situ visualization of defects of R2R printed silver conductors on flexible plastic substrates. The fact that the defect inspection is carried out on R2R fabricated samples does not prove the capability of the tool for use during the R2R production. The movement of the web in a R2R process will arise a huge number of new issues that should be addressed before the system and the methodology of this work would be able to provide real time data and therefore, many aspects claimed in this work are missing.

Taking in to account the above, I suggest that the manuscript and its content are insufficiently novel to warrant its publication.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The authors demonstrated a dark lock-in infrared thermography method-based inspection system to visualize the defects of roll-to-roll printed silver conductors on flexible plastic substrates. This system shows exactly the position of the defect on the printed patterns when they are electrified. To achieve this goal, the authors developed a two-stage defect recognition methodology based on localized segmentation and thresholding. I recommend accept the work for publication after improvement.

The following points are for the authors to further enhance their work:

It is recommended that the title of the color card should be added in Fig. 7, Fig. 8 and Fig. 10.
Page 6, Line 160-161, “Figure 7 (a) and Figure 7 (b) show the results of Test Sample 1 with and without a complete wire broken.” should be corrected as “Figure 7 (a) and Figure 7 (b) show the results of Test Sample 1 without and with a complete wire broken, respectively”.

Line 164-165, “Figure 8 (a) and Figure 8 (b) show the results of Test Sample 2 with and without a complete wire break” should be corrected as “Figure 8 (a) and Figure 8 (b) show the results of Test Sample 2 without and with a complete wire break, respectively”.

If the thickness of the local area on the printed pattern is too thin or the pattern is partial printed, then local hotspots will be appeared. And how do you distinguish these two cases from the DLIT amplitude image?

Author Response

Thank you very much for your comments and questions. Below are the answers and replies to your questions:

Point 1: It is recommended that the title of the color card should be added in Fig. 7, Fig. 8 and Fig. 10.

Response 1: Thank you for suggestion. We have added the title ‘DLIT amplitude in colour scale’ in Fig. 7, Fig. 8 and Fig. 10. As graphically illustrated in Figure 2, DLIT amplitude can be obtained from sine image and cosine image by calculating .

Point 2: Page 6, Line 160-161, “Figure 7 (a) and Figure 7 (b) show the results of Test Sample 1 with and without a complete wire broken.” should be corrected as “Figure 7 (a) and Figure 7 (b) show the results of Test Sample 1 without and with a complete wire broken, respectively”.

Response 2: Thank you. We have corrected these sentence (L161 and L165 on Page 6).

Point 3: If the thickness of the local area on the printed pattern is too thin or the pattern is partial printed, then local hotspots will be appeared. And how do you distinguish these two cases from the DLIT amplitude image?

Response 3: Thank you for question. The partial wire breaks can be decrease in thickness or width of printed patterns, which can both result in a local increased resistance consequently a higher temperature to be detected by an infrared camera, which is so-called hot spot. It seems difficult to distinguish these two types of defects as they are both shown as hot spots in thermal images.

When we simulated defects on test samples, we are only able to simulate partial wire breaks by scratching on the wires to decrease the width. We are currently numerically simulate more defect scenarios by FEM. We can simulate variances of thickness and width on different printed patterns. The simulation results will be used to develop machine learning based ADR method. This work is under development and will be published in the future when we achieve good results.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Dear Authors,

I would like to thank you for the response to my comments. Nevertheless, as stated in my initial review in the article there is a significant lack of scientific results which are claimed to be presented in future work.

Taking in to account the above, I suggest that the manuscript at the present form is insufficiently novel to warrant its publication and should be reconsidered for publication only after implementing the proposed methodology in in-line real-time measurements.

Author Response

We would like to thank you for your comments.

Although there were some similar works conducted by other researchers, to the best of our knowledge, the proposed methodology based on two-stage localized segmentation in this study is completely novel. This study clearly shows succseeful defect detection and localization results.

The objective of this paper is to report our works on DLIT labaratory tests and defect recognition by using the proposed innovative localized segmentation algorithm, rather than our future work on in-line implementation, as we clearly stated in the conclusion session.

We might have not clearly resolved all the comments in the previous answers so we would like to make further revisions on the following points:

Point 1: The algorithm is not fully automated

Response 1: As we explained in the Section 5 - ADR Results and Discussion (Line 354): ‘the current version of ADR is semi-automatic as it still requires the operators to input a coarse ROI arrangement of wires’. We changed the title of this paper to ‘Defect Recognition of Roll-to-Roll Printed Conductors Using Dark Lock-in Thermography and Localised Segmentation’.

Point 2: There is no quantitative information regarding the defect size.

Response 2: We added quatitative information of defect size of wire breaks at Line 176: “The defect sizes of these two wire breaks are around a half of the wire widths, which are 250-500 μm.”

Point 3: Regarding the efficiency and repeatability of the recognition process

Response 3: More works will be carried out in the next stage to achieve these, and we modified the last sentence in the conclusion: “ The future works will also include more DLIT trials with various printing pattern, material, and defect scenarios. Hence, the repeatability and robustness of the proposed method can be extensively studied in the next stage.”

Point 4: The presented development were to be implemented, but necessary to be validated in the additional testing.

Response 4: Yes, the in-line implemenation of DLIT on the R2R testing line is under test. We have stated the plan in the conclusion: “In the next stage, the DLIT tool will be upgraded to provide fully automated operation and implemented on R2R testing line at VTT to continually inspect the printed conductive thin films.”

Author Response File: Author Response.pdf