Influence of the Approach Direction on the Repeatability of an Industrial Robot

Round 1

Reviewer 1 Report

The article deals with the important problem of repeatability analysis, which is one of the most important parameters of robots. The authors presented a good and critical review of the literature, and the work was based both on theoretical analysis and experimental research.

 

Comments:

It is not clear to me for what purpose the analysis of the influence of weight, radius etc on repeatability was made and mixed with the study of the influence of the direction of motion. Shouldn't only the analysis of the measurement data mentioned in Table 5 be performed? Only then can the influence of other factors be examined separately, eg minimum and maximum load, etc. Then the work would be ordered. Now, I have a feeling it's all mixed up.

As I understood from the content of the article, the direction of movement is related to the starting point. Is the starting point number given in the second column of tables 4-8 or is it different? If these are the starting points, why do graphs 10 and 11 show the results for measurements 14 and 16 where there are different masses. Please explain. And be sure to specify what the "points" column in tables 4-8 means, because all my doubts arise from it.

It seems that the citations [26, 27] concerning modal analysis and rotating disk do not relate to the subject of the article. In my opinion, they are an attempt to increase the number of citations.

Author Response

Dear reviewer,

Thank You for the paper assessment. In the beginning, the authors defined a set of measurements to perform. The authors were curious how the other parameters will influence the repeatability, e.g., how the results will differ. The paper proved the hypothesis, that the pose-repeatability is influenced by the approach direction. We believe that pose repeatability should be increased only in those few targets, in which a high level of precision is required. By increasing the pose-repeatability in a defined target, the level of precision will be higher than if the pose-repeatability was increased in general (in the whole workspace of a robot). However, if precision is increased for a specific target, this is valid only for the target and the defined conditions (speed, payload, etc.). Also, this measurement should be repeated after two years, as any high precision device must be calibrated.  

This is the reason why authors performed a large set of measurements under different conditions. We wanted to demonstrate that our methodology and stance is valid in general, for any motion conditions. Also, it is extremely difficult to determine the influence of every single factor (speed, payload, amortization, etc.). A six-axis industrial robot is too complicated. This article aims to the system – state determination and precision increase regardless of the degree of influence of individual factors.

The starting point number is defined in the second column of tables 4 – 8.  As the starting points are divided on the hemisphere by spherical Fibonacci lattice formula, the number of points represents the measurement precision (the higher the number of points, the more comprehensive the result). On the other hand, if the number of points is high (160, for example) the measurement process is time-consuming.  The authors believe that in the article, the verification of dependence of the number of points on a measured result is required.

We have performed a lot of measurements, only a few could be shown in this paper, as we would like to keep it readable and legible. The authors decided to display measurements 14 and 16. There is no specific reason for this, any measurement could have been described in the paper.

It may seem, that only one table of measurements could be performed for this paper. The authors wanted to be sure, that our statements are not wrong or influenced by some “specific conditions”.

The use of the 3D DIC measurement system correctly is not an easy task, especially for long-term measurement. By the citations mentioned, the authors wanted to show that they are not “beginners” in this field. Also, a citation of the methodology used in our article had to be included, as we did not invent the DIC principle. If, in your opinion, it is inappropriate, the authors will replace the citations [26, 27] with a different reference.

 

Yours faithfully

Michal Vocetka, on behalf of all the authors

 

Author Response File: Author Response.pdf

Reviewer 2 Report

I recommend to check the phrase at lines no. 390-392.

In one sentence you state "drift is not significant". In next you conclude "pre – warm-up routine ... as necessary". Can you improve logics?

Also from my point of view shafts' deflection creates more deviation then deflection of of the arm body.

Author Response

Dear reviewer,

Thank you for the paper assessment. The phrase is formulated differently, but its logic should stay the same. Drift is a problem that must be avoided, if possible. The basic precondition on how to avoid the influence of the drift is to operate the manipulator on its operational temperature, without any significant changes of both the manipulator and ambient temperature. This basic precondition was followed and so the influence of drift during our measurement is not significant.

Thank You for the remark, that shaft deflection creates significant deviation, it was added to the text.

 

 

Yours faithfully

Michal Vocetka, on behalf of all the authors

 

Author Response File: Author Response.pdf

Reviewer 3 Report

The mechanical system of a robot is formed by a configuration of rigid bodies, the elements of the system, connected to each other successively by rotating joints and translations. The modelling of the mechanical system of the robots represents the basic stage for the elaboration of the axle controls, in accordance with the movement objective.

The presented article represents a step forward in increasing the performance associated with these types of industrial robots and manipulators.

 

Please consider the following observations:

1. A better explain the measured target in relation to TCD, WCS and SCS (line 223 ... 228).
2. Do the results obtained by the experimental tests refer only to the payload of 2,538 kg (table 4, 5 and 6 refer only to 2,538 kg)? If so, why were 0,04 kg and 4,647 kg not tested?
3. Line 315. A figure containing the part of the experimental stand formed by the Q-450 Dantec Dynamic System must be presented. Figure 1 should be taken and supplemented with explanations of the DIC elements used.
4. I recommend removing lines 353 ... 374 because, in [37], page 9, the steps are presented almost identically and are not the subject of the article, especially since the Q-450 system does the calibration automatically.
5. The conclusions do not explain very well the validation of the experimental data obtained from robot drive resolvers and the data obtained through the Q-450 system.
6. Please draw a practical conclusion for end users of robots (for example for a robot used for welding).

 

Note: please explain (even if there are elements known to those who deal with this domain) the abbreviations used before being used: DIC; TCD; WCS.

Example: line 34 TCP (Tool Center Point) and line 227 TCP (Tool Coordinate System).

Congratulations to the authors for all the experimental data processed.

Author Response

Dear reviewer,

Thank you for the paper assessment.

1. The terms are explained in the new version.
2. All the measurements described in tables 4 – 8 were performed. The six-axis manipulator is too complex to define an influence of a single factor that affects the precision in a level of precision we would like to deal with, the only way is to measure and increase precision in a specific target, whatever the causes of inaccuracies are. We measured the precision under different conditions, to verify that our conclusions were correct.
3. Thank you for this advice, the new position for the figure is more suitable.
4. The Q450 calibration is not performed automatically. It is automated but operator assistance is still necessary. It is true, that the lines are not essential for the paper, and were therefore removed.
5. The data from drive resolvers were used for hypothesis verification (and verification that the error does not occur randomly). Data processed from the DIC are the data of the final deviation of the robot tool, in the measured target. For this reason, the resolver result doesn’t have to be explained in detail in conclusion, we believe that only the information written above should be mentioned there. The important result is the DIC data.
6. A simplified procedure is presented in conclusion, this method would be more suitable for high-precision assembly. Welding applications are precise, but their problem is the input part; the method presented above can improve the repeatability of the robot, but cannot precisely measure the deviation of welded parts that are usually not positioned with high precision and so increased repeatability of the torch would be useless. A seam tracking method is therefore more appropriate.    

 

Yours faithfully

Michal Vocetka, on behalf of all the authors

 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I accept all the explanations of the Authors with satisfaction and understanding.

I think the Authors may add a paragraph to explain such a rich set of experimental data, similar to how they did in response to my review.

Thanks to the Authors for their answers.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The problem of robot repeatability and accuracy is important because of many robotic applications which require high precision.

The presented hypothesis is obvious, as is very well known that the robot TCP accuracy and repeatability is related to the direction of the movement. Robot position error can be a result of many factors including: control system errors, clearances, thermal expansion, different deflection of robot arm caused by different rigidity in the x,y,z axes, and others.

However, this article is interesting because it presents in-depth research and provides a deeper insight into the nature of the problem.

There are some questions which require clarification and some corrections are needed for improvement of the readability of the paper.

Question 1

What about the second robot (inclined by an angle of 40)? Was this robot used for experiments? If so, for which experiments was it used? Comparing the results for both robots would be interesting.

 

 

 Question 2

As the vision measurement system (DIC) only allows measurements in two axes, how were the measurements in the third axis obtained?

(equations 1 and 2)?

 

Comments

The paper requires some corrections, as it was not done accordingly to the template.

Fig. 6, 15, 19 - description of the axis is required

Fig. 10-14 and  20-21 are very small and hardly readable

Reference [34] is old and presents only mechanisms with closed kinetic chains, but manipulators and robots have open kinetic chains.

A spell check of the English language is required.

Author Response

Dear reviewer,

thank You for the paper assessment, we really appreciate it. We have made all the formal corrections, that were necessary. We hope that you will find these corrections sufficient and convenient.

The mentioned reference is old, but the original DH principle is still valid, and the reference is appropriate. However, as this mathematical principle has been generally known for years, the reference is not necessary, so we removed it.

Q1:         The second robot was not used for the experiment (authors considered the common measurement, but the decision was to not perform it) because these results would not be interesting at all. Robot pose-repeatability was measured, and this value is valid not only for a specific robot but also for specific conditions (target position, robot (axis) temperature, speed, robot amortization, etc.). So, the measurement results are useable only for these conditions and cannot be used for any different robot or conditions and are valid for a limited period.

Both the robots are used now during new long-term measurements focused on heating of the robot structure under different conditions. Pose – repeatability evaluation and correction are going to be the next steps of this subsequent research.

Q2:         The authors used a 3D DIC measurement system (two angled cameras were used) so the shift and rotation were evaluated for all the axis. To make it more evident, the description in the paper has been extended.

Yours faithfully

Michal Vocetka, on behalf of all the co-authors

Reviewer 2 Report

Dear Authors,

Greetings, I must say that I greatly appreciated the reading of your article because of the adequate dose of detail, accompanied by graphics and results of step-by-step experiments; without overwhelming it.

On the other hand, I found that the techniques used in the different stages allow an orderly progression towards the verification of your hypothesis, the repeatability of the robot at the target based on the direction of approach. 

It is also important your evidence on the use of external devices such as DIC cameras that allow a more accurate measurement when evaluating repeatability.

Best regards,

 

Author Response

Dear reviewer,

thank You for Your kind words, authors are very pleased that you found our research interesting and well written.

Yours faithfully

Michal Vocetka, on behalf of all the co-authors

Reviewer 3 Report

The title of the paper is about robot repeatability but if there is a thorough bibliography on robot accuracy, there no definition of repeatability, no mention of papers about industrial robot repeatability, no citation of the standards about repeatability estimation. There is a contradiction between the title and the lines 132-135; the same in line 164 where accuracy and repeatability seem to be the same thing: line 230 "repeatable accuracy"?; line 253 "absolute inaccuracy" ? The authors absolutely need to clarify the concept of repeatability and accuracy, which have a complete different meaning in the field of industrial robot precision.

So it seems that in this paper, there is a constant confusion between accuracy and repeatability.

The authors have accumulated of lot of interesting experimental data but it seems that the analysis of the data are done on the basis of the accuracy estimation not on repeatability. The influence of the other parameters varying during the measurements is not done. For instance in Fig.21 the influence of drift due certainly to thermal effect is not analysed.

If considerable amount of data have been acquired, their analysis and representation must be seriously improved. Most of the figures are not readable (see for instance fig.8, fig.11, fig.21).

Conclusion : The authors should consider the definition of repeatability given in the standards (ISO9283 or ANSI) and reorganize their analysis on this basis to see if their conclusions are still the same. They have to be more careful about the use of vocabulary to avoid giving the impression that for them repeatability and accuracy is the same concept.

 

Author Response

Dear reviewer,

thank you for the paper assessment. The authors appreciate your analysis, which will certainly improve this article. It is true that in the introduction, scientific works focused on measurement and evaluation of the robot accuracy predominate. It is caused by the fact that there are almost no papers that deal with repeatability. Our article is going to be one of the few. This finding was surprising for us. However, one valuable article was cited, and we added some additional now. In the authors’ opinion, the introduction is beneficial, because the methodologies described in those papers could be used for the accuracy measurement, as well as for the repeatability measurement.

The constant confusion that you mentioned was caused by an inappropriate translation to English. The terminology has been unified and we hope that the paper is no longer confusing. The research deals with repeatability only. So, our conclusions are still valid.

Many influences affect the repeatability value, the authors are acquainted with them. However, the purpose of this paper is not to describe in detail every single influence, but to measure the real robot state, under the defined conditions and deduce conclusions that would help to increase repeatability. Thermal expansion will always affect any high-precision device. For this reason, the robot was warmed-up before every measurement and kept at a constant temperature as precisely as possible. This is a standard routine for every precise robot- programming task. Before the robot targets are precisely calibrated, the robot is kept in the movement for some time, to warm up its structure in the same way it will be warmed during the production cycle. During the measurement, the robot was not moving by long TCP distances, so none of the robot gearboxes (the greatest sources of heat in the case of ABB IRB1200-5/0.9) did not heat up by more than 3°C. Authors have this information from long-term measurements focused on heating of the robot structure under different conditions; this is going to be a part of subsequent research.

 

Yours faithfully

Michal Vocetka, on behalf of all the co-authors

Round 2

Reviewer 1 Report

All the formal and substantive corrections that were necessary to improve the paper have been made.

There is still a small issue related to the analysis of the results:

Line 443 “For example, in measurement 20 (Fig. 25), the difference from the (zero, exact) value is +0.019 (max) and +0.106 (min).”

It should be checked, as the given value is outside the range shown in Figure25 and maybe it should refer to Fig 24.

Reviewer 3 Report

The major drawbacks of the paper are still not answered. There is still a need fot a proper definition of repeatability considering the Standards ; the authors still need to clarify the difference between repeatability, accuracy, error...

A lack also in the statistical analysis of the differences observed in the results considering the variability of the random variable "repeatability" when a 10-sample is used (instead of 30-sample indicated in the ISO Standard). If the ISO Standard is considered, there is a high variability of the repeatability for a 30-sample, and the variability is even higher for a 10-sample. This issue is not adressed in this paper.

The proposed corrections do not answer the main problems previously detailed about the paper. These are just minor corrections.